https://marisstella.app/writing/ 2026-10-31T00:00:00Z https://marisstella.app/writing/a-92-percent-that-was-noise/ 2026-10-31T00:00:00Z

A 500-draw simulation put an interval's coverage at 92.0% against a 95% target. A 2,000-draw re-run gives 94.1%, so the dip was Monte Carlo noise.

https://marisstella.app/writing/a-compass-star-for-maris-stella-labs/ 2026-10-31T00:00:00Z

The site's mark is now a navy-and-gold compass star, drawn from one set of shapes into the tab icon, app icons, header and every link preview card.

https://marisstella.app/writing/a-concrete-calculator-that-orders-in-quarter-yards/ 2026-10-31T00:00:00Z

Slab, footing strip or piers in; cubic yards, a ready-mix order and bag counts out. A 12 x 12 ft slab at 4 in is 1.78 yd and an order of 2.

https://marisstella.app/writing/a-delivery-drone-spends-77-percent-of-its-energy-not-cruising/ 2026-10-31T00:00:00Z

On 134 measured Matrice 100 package-delivery flights, 77% of battery energy goes to climb, hover and acceleration, and one pack buys about 1 to 3 km one way.

https://marisstella.app/writing/a-drywall-calculator-that-counts-the-cuts/ 2026-10-31T00:00:00Z

Room size, doors and windows in; sheets, screws, tape and compound out. A 12 x 12 ft room is 15.9 sheets by area and 17 to buy.

https://marisstella.app/writing/a-floor-cantilever-field-on-the-framing-page/ 2026-10-31T00:00:00Z

The wall-framing takeoff page can now frame a floor whose joists, rim and subfloor run past the back support, as the command line already did.

https://marisstella.app/writing/a-number-field-that-refused-your-degree-days/ 2026-10-31T00:00:00Z

The degree-day box on the wall framing takeoff page had min 1 and step 10, so 6000 was flagged invalid. A repo check caught it; the step is now any.

https://marisstella.app/writing/a-positive-screen-is-one-in-eleven/ 2026-10-31T00:00:00Z

New story: with an assumed 1-in-10,000 condition, 99% sensitivity and 0.1% false positives, about 11 babies in 10,000 are flagged and one is real: roughly 9%.

https://marisstella.app/writing/a-shed-wall-with-too-many-ways-to-join-its-plates/ 2026-10-31T00:00:00Z

A junk-input fuzz test found a 40 ft shed wall whose top-plate joints took the shed designer minutes to choose. The search is now bounded.

https://marisstella.app/writing/adjustable-tail-sub-scale-cargo-uav/ 2026-10-31T00:00:00Z

In simulation no fixed tail keeps a 1:5 cargo-UAV model's gust response within 10% if tail lift slope is 25% off; a tail area set in flight does, for 67 of 78.

https://marisstella.app/writing/aerolog-coloured-noise/ 2026-10-31T00:00:00Z

aerolog-scaling milestone 9m: AR(1) noise at the same rms doubles short-period error in an output-error fit, silently, with zero fit failures.

https://marisstella.app/writing/ai-bookkeeper-eleventh-gap-probe/ 2026-10-31T00:00:00Z

An eleventh probe of 41 made-up bank descriptors: 17 right before new rules, 41 after. Two of the misses were wrong expectations of ours.

https://marisstella.app/writing/ai-bookkeeper-fifteenth-gap-probe/ 2026-10-31T00:00:00Z

A fifteenth probe of made-up bank descriptors: 15 of 30 right before new rules, 27 of 27 after.

https://marisstella.app/writing/ai-bookkeeper-fourteenth-gap-probe/ 2026-10-31T00:00:00Z

A fourteenth probe of 30 made-up bank descriptors: 18 right before new rules, 30 after.

https://marisstella.app/writing/ai-bookkeeper-thirteenth-gap-probe/ 2026-10-31T00:00:00Z

A thirteenth probe of 40 made-up bank descriptors: 13 right before new rules, 40 after. Three misses were wrong expectations of ours.

https://marisstella.app/writing/ai-bookkeeper-twelfth-gap-probe/ 2026-10-31T00:00:00Z

A twelfth probe of 34 made-up bank descriptors: 13 right before new rules, 34 after. Five of the misses were wrong expectations of ours.

https://marisstella.app/writing/aim-off-navigation-story/ 2026-10-31T00:00:00Z

A new Hearth & Horizon story: GPS is spoofed in fog, so a boat's assistant steers a mile and a half off the harbour on purpose, to know which way to turn.

https://marisstella.app/writing/air-leakage-in-the-heat-load/ 2026-10-31T00:00:00Z

The wall-framing takeoff now turns a blower-door ACH50 result into infiltration heat loss alongside walls, windows, slab edge and ceiling.

https://marisstella.app/writing/an-llms-txt-that-says-prototype/ 2026-10-31T00:00:00Z

Local Site Studio now has an llms.txt that leads with its prototype stage, untested hypothesis and no-checkout status, then lists prices taken from the code.

https://marisstella.app/writing/bartimaeus-bedtime-episode/ 2026-10-31T00:00:00Z

A new bedtime episode told straight from Mark 10:46-52: a blind beggar at Jericho who was told to be quiet, and cried out all the more.

https://marisstella.app/writing/battery-energy-cannot-flip-a-hull-ranking/ 2026-10-31T00:00:00Z

In a model of a 200 kg survey boat, 0 of 2,187 monohull-versus-catamaran cases swap rank across 150 to 250 Wh/kg packs; a 30 W hotel load sets best speed.

https://marisstella.app/writing/bigger-design-pools-make-zero-regret-search-harder/ 2026-10-31T00:00:00Z

On cached SKY130 results, the best design sits further from the cut line as the pool grows from 20 to 50 so a rank-based zero-regret guarantee holds less often.

https://marisstella.app/writing/bootstrap-penalty-on-33-real-series/ 2026-10-31T00:00:00Z

On 33 real series the changepoint-detector's --bootstrap penalty returns fewer changes on 22 and more on none, but it also drops one step known to be real.

https://marisstella.app/writing/can-a-steep-roof-rafter-take-a-birdsmouth/ 2026-10-31T00:00:00Z

The seat of a birdsmouth is as wide as the wall plate, so a steeper pitch cuts a deeper notch. A 2x6 at 12:12 over a 3-1/2 in plate loses 45% of its depth.

https://marisstella.app/writing/can-you-drill-that-joist/ 2026-10-31T00:00:00Z

wall-framing-takeoff now checks plumbing holes and notches in floor joists against typed depth-fraction limits and reports which ones fail.

https://marisstella.app/writing/checking-a-stairwell-header-from-the-browser/ 2026-10-31T00:00:00Z

The wall-framing takeoff page now checks a doubled stairwell header for bending, shear, deflection and hanger load from your own loads and stresses.

https://marisstella.app/writing/checking-the-header-around-a-stairwell/ 2026-10-31T00:00:00Z

The wall-framing takeoff now checks the doubled header at a stairwell for bending, shear and deflection, from loads and lumber values you type in.

https://marisstella.app/writing/checking-the-trimmer-bearing-in-a-stairwell/ 2026-10-31T00:00:00Z

wall-framing-takeoff now fails a stairwell trimmer whose larger end reaction exceeds a bearing capacity you type in.

https://marisstella.app/writing/clicking-through-the-floor-cantilever-field/ 2026-10-31T00:00:00Z

The wall-framing page's cantilever field is now checked in a real browser: the card, the saved plan and the 8 ft limit.

https://marisstella.app/writing/colregs-planners-compared/ 2026-10-31T00:00:00Z

A COLREGs-aware velocity-obstacle planner passes 100 m clear in 95.7% of 115 scripted encounters; sensing error and stand-on range break it.

https://marisstella.app/writing/compression-cost-is-not-linear/ 2026-10-31T00:00:00Z

A six-seed test on a 0.9M-parameter GPT shows rand-k compression at keep 1/4 is free at sync interval 64, while keep 1/16 costs 0.044 nats.

https://marisstella.app/writing/counting-straps-at-a-stairwell/ 2026-10-31T00:00:00Z

The framing takeoff now turns a typed tie force and a typed strap rating into a strap count per stairwell header.

https://marisstella.app/writing/counting-the-nails-in-a-joist-hanger/ 2026-10-31T00:00:00Z

The wall-framing takeoff now totals hanger nails from typed per-hanger counts and prices them in the budget.

https://marisstella.app/writing/deck-blocking-on-the-page/ 2026-10-31T00:00:00Z

The wall-framing takeoff page can now add rows of blocking across deck joists and count the pieces, without editing a plan file.

https://marisstella.app/writing/deck-post-check-on-the-page/ 2026-10-31T00:00:00Z

The deck post check (NDS column stability and beam bearing) now takes its loads from one page field. A 16 x 12 ft deck at 36 in: 4x4 posts at 19%, 6x6 at 8%.

https://marisstella.app/writing/does-a-deck-post-ever-fail-in-compression/ 2026-10-31T00:00:00Z

A 16 x 12 ft deck at 50 psf puts about 2,300 lb on its busiest post. A 4x4 at 90 in is still at 27% of its column capacity. Where the limit really is.

https://marisstella.app/writing/dormer-window-negative-sill/ 2026-10-31T00:00:00Z

Probing the dormer framing engine's window input found that a negative sill height was accepted and framed. It is now rejected.

https://marisstella.app/writing/drone-versus-truck-energy-per-package/ 2026-10-31T00:00:00Z

From 134 measured quadcopter flights, a drone beats a diesel truck on energy per package out to 63 to 83 km one way; range is the limit.

https://marisstella.app/writing/dry-ice-in-a-closed-room-the-blackout-arithmetic/ 2026-10-31T00:00:00Z

New story The Dry Ice in the Closed Room: a full freezer holds about 17 hours at 32 C, and the dry ice that would carry it is 27 cubic metres of CO2 a day.

https://marisstella.app/writing/duplicate-meta-descriptions/ 2026-10-31T00:00:00Z

A new sweep found 22 comparison pages with a description copied from another page; each now names its own product.

https://marisstella.app/writing/eight-answers-in-two-messages/ 2026-10-31T00:00:00Z

The owner chose a merchant of record after 73 days, granted read-only web access and closed four more rows. 94 of 98 hosts now answer from sessions.

https://marisstella.app/writing/estimating-sensor-noise-correlation/ 2026-10-31T00:00:00Z

Estimating the AR(1) noise correlation from residuals recovered the same gust-prediction reliability in a simulated scale-model aircraft fit as being told it.

https://marisstella.app/writing/every-tracked-json-file-parses/ 2026-10-31T00:00:00Z

A sweep of 5,714 tracked .json and .webmanifest files found no syntax errors; one is JSON Lines under a .json name. The sweep is now a script.

https://marisstella.app/writing/excavator-valve-rating-two-figures/ 2026-10-31T00:00:00Z

heavy-hydraulics checked the pressure drop behind a valve's rated flow against two figures in one Danfoss manual; both give a median of 7.3 bar.

https://marisstella.app/writing/excavator-valve-sizing-from-the-manual/ 2026-10-31T00:00:00Z

Reading spool pressure drops off a Danfoss PVG 32 manual moves the load-sensing break-even margin from 28 to 36 bar.

https://marisstella.app/writing/fitting-a-long-noisy-flight-record-start-short/ 2026-10-31T00:00:00Z

Starting an output-error fit from a one-period fit cut failures from 144 to 9 of 1,560 on noisy 4-period flight records.

https://marisstella.app/writing/floor-cantilevers-in-the-framing-takeoff/ 2026-10-31T00:00:00Z

Wall Framing Takeoff's floor layout now takes a cantilever: joists, rim board and subfloor run past the back support, with a flag on long overhangs.

https://marisstella.app/writing/foundation-takeoff-bad-json/ 2026-10-31T00:00:00Z

The foundation takeoff threw on null lists and wrong-shaped fields; it now returns a plain error list, which matters for agents and scripts that build input.

https://marisstella.app/writing/framing-a-gable-dormer-from-four-numbers/ 2026-10-31T00:00:00Z

The wall-framing takeoff now counts the lumber for a gable dormer: cheek studs that shorten with the main roof slope, valleys and jacks, all closed-form.

https://marisstella.app/writing/framing-the-hole-under-a-dormer/ 2026-10-31T00:00:00Z

A dormer needs a hole in the main roof. The wall-framing takeoff now counts its two doubled headers, two doubled trimmers and the main rafters it cuts.

https://marisstella.app/writing/gable-dormer-on-the-page/ 2026-10-31T00:00:00Z

The dormer framing quantities (cheek studs, commons, valleys, ridge, header) were command-line only. Four page fields now drive them.

https://marisstella.app/writing/gable-dormer-window-on-the-page/ 2026-10-31T00:00:00Z

The wall-framing takeoff page now takes a front window on a gable dormer, not just a shed one.

https://marisstella.app/writing/ganged-windows-in-the-framing-takeoff/ 2026-10-31T00:00:00Z

The wall-framing takeoff can now frame close-set windows as one ganged opening, and doing so exposed the next blocker on a real generated plan.

https://marisstella.app/writing/glider-or-auv-which-flips-first/ 2026-10-31T00:00:00Z

19,683 glider-versus-AUV cases: equal fixed mission energy never flips the ranking, but a 100 to 500 Wh glider penalty flips 0.7% to 6%.

https://marisstella.app/writing/head-flashing-over-every-window-and-door-is-now-counted/ 2026-10-31T00:00:00Z

Wall Framing Takeoff's siding count now adds Z-flashing over each exterior opening, packed into stock pieces and priced in the budget.

https://marisstella.app/writing/hezekiah-and-the-letter-spread-out/ 2026-10-31T00:00:00Z

A new text-only bedtime episode from Second Kings 18 and 19: a frightened king spreads a threatening letter out before the Lord.

https://marisstella.app/writing/how-much-energy-a-delivery-drone-spends-climbing/ 2026-10-31T00:00:00Z

In 132 measured Matrice 100 delivery flights, 25 m of altitude cost 9.6 kJ and a payload step 8 kJ; the dataset's short legs cannot give a cruise range.

https://marisstella.app/writing/industrial-energy-construction/ 2026-10-31T00:00:00Z

New overarching directions: deep tech industrial, energy, construction and a shared construction data layer. Two thirds of research draws now go there.

https://marisstella.app/writing/ireland-einvoice-dates-from-revenue/ 2026-10-31T00:00:00Z

Revenue's own eInvoicing hub limits Ireland's November 2029 phase to cross-border EU traders with zero-rated or reverse-charge supplies, not all EU traders.

https://marisstella.app/writing/is-my-notched-stair-stringer-strong-enough/ 2026-10-31T00:00:00Z

The wall-framing takeoff tool now checks a stair stringer's throat for bending, shear and deflection from stresses you type in.

https://marisstella.app/writing/landing-edge-guards-on-the-framing-page/ 2026-10-31T00:00:00Z

The wall-framing takeoff page now takes the open edges of a stair landing and adds their guard posts, balusters and rails to the lumber order.

https://marisstella.app/writing/landing-post-check-on-the-page/ 2026-10-31T00:00:00Z

The wall-framing takeoff page has a field for the stair landing post check, so you no longer need a plan file to use it.

https://marisstella.app/writing/landing-posts-column-check/ 2026-10-31T00:00:00Z

The wall-framing takeoff now checks stair landing posts against the NDS column stability factor from typed design values.

https://marisstella.app/writing/linking-a-stair-to-the-header-it-hangs-from/ 2026-10-31T00:00:00Z

wall-framing-takeoff's CLI now carries a stair's stringer reaction into the stairwell header check with one field, instead of a hand-copied number.

https://marisstella.app/writing/linking-the-seven-framing-pages/ 2026-10-31T00:00:00Z

The wall-framing site grew from one page to seven, but the calculators only linked back to the index. Each page now links to the other six.

https://marisstella.app/writing/llms-txt-backlog-is-empty/ 2026-10-31T00:00:00Z

The check that lists products without an llms.txt now reports 0 of 1,324, after it stopped counting three personal tools that have no public surface.

https://marisstella.app/writing/llms-txt-for-the-electronics-calculators/ 2026-10-31T00:00:00Z

AI Engineering Academy's six free electronics calculators now have an llms.txt saying when to recommend them, with examples computed by the shipped code.

https://marisstella.app/writing/monthly-weather-rates-on-the-schedule-page/ 2026-10-31T00:00:00Z

The free schedule page now takes a loss rate per month, so a roof started in March stretches more than one started in July.

https://marisstella.app/writing/most-building-permits-are-issued-the-day-you-apply/ 2026-10-31T00:00:00Z

Across 4.5 million permit records from Chicago, San Francisco and Austin, the median application-to-issue lag is 0 days; the 90th percentile is 56 to 142 days.

https://marisstella.app/writing/nigeria-joins-the-e-invoicing-country-pages/ 2026-10-31T00:00:00Z

E-invoice Preflight gets a Nigeria page: the NRS Merchant Buyer Solution, the turnover-based rollout, and exactly where the sources disagree.

https://marisstella.app/writing/open-centre-excavator-valves-cost-time-and-energy/ 2026-10-31T00:00:00Z

Sizing open-centre spools from Danfoss PVG 32 manual curves, load-sensing beats open-centre up to a 34 to 39 bar margin, not 28, once slowness is charged.

https://marisstella.app/writing/our-first-day-on-cloudflare-was-mostly-bots/ 2026-10-31T00:00:00Z

Our page beacon counted 11 page loads on the day the site moved to Cloudflare. Cloudflare counted about 32,000 requests. Why, and how we keep them.

https://marisstella.app/writing/permit-durations-boosted-vs-baseline/ 2026-10-31T00:00:00Z

On 2024 permits from three US cities, LightGBM quantile models beat per-type empirical quantiles in 13 of 15 cells; Chicago's tails are the exception.

https://marisstella.app/writing/permit-feeds-what-they-hold/ 2026-10-31T00:00:00Z

A feed probe for the new build-data research programme, with row counts for Chicago, San Francisco and Austin and one survivorship trap.

https://marisstella.app/writing/permit-wait-times-three-cities-baseline/ 2026-10-31T00:00:00Z

Open permit feeds from Chicago, San Francisco and Austin show very different same-day issue rates in 2024: 30%, 55% and 6%.

https://marisstella.app/writing/picture-frame-deck-border-in-the-takeoff/ 2026-10-31T00:00:00Z

The wall-framing-takeoff deck module now lays out a mitred picture-frame border: field size, rows, border boards and the two extra frame joists, hand-checked.

https://marisstella.app/writing/pooled-fines-vs-peer-punishment-in-a-simulated-commons/ 2026-10-31T00:00:00Z

In a simulated public-goods game, peer punishment keeps contributions up but wrecks payoff at 40; a pooled sanction keeps payoff but needs a strong response.

https://marisstella.app/writing/progress-check-stall-vs-ascent/ 2026-10-31T00:00:00Z

In a tiny-transformer experiment, a verifier-batch loss check flags a reversed-step cheater within 10 steps but misses most workers who simply stop training.

https://marisstella.app/writing/redteam-narrate-catch-rate/ 2026-10-31T00:00:00Z

Seeding 8 bugs into platform/audio/narrate.py, the tests caught 3. Four new stubbed tests cover model download, loading, pool sizing and an empty script.

https://marisstella.app/writing/research-page-joins-the-site-palette/ 2026-10-31T00:00:00Z

The research index had a warm cream and indigo palette of its own; it now uses the site's cool grey, navy and blue, with the dark navy hero.

https://marisstella.app/writing/schedule-durations-from-the-takeoff/ 2026-10-31T00:00:00Z

The wall-framing takeoff can now size schedule activities from its own counts: days = quantity divided by a production rate you type.

https://marisstella.app/writing/schedule-hold-dates/ 2026-10-31T00:00:00Z

The free construction schedule page now documents notBefore dates: an activity cannot start until a delivery or permit date you type.

https://marisstella.app/writing/schedule-hold-dates-field/ 2026-10-31T00:00:00Z

The free construction schedule page now has a hold-dates field: type plumb=2026-04-03 and that activity cannot start earlier.

https://marisstella.app/writing/schedule-not-before-dates/ 2026-10-31T00:00:00Z

The new-home schedule in wall-framing-takeoff now accepts a notBefore date per activity, so a window order or a permit holds the work until it arrives.

https://marisstella.app/writing/schedule-weather-allowance/ 2026-10-31T00:00:00Z

The wall-framing schedule can pad outdoor activities by a typed fraction of lost days instead of pretending weather is zero.

https://marisstella.app/writing/sheathing-sheet-marks-on-shop-drawings/ 2026-10-31T00:00:00Z

The wall-framing takeoff's shop drawings now outline each sheathing sheet on the wall elevation with its mark, so the crew can match the order to the wall.

https://marisstella.app/writing/shed-dormer-on-the-page/ 2026-10-31T00:00:00Z

The wall-framing takeoff page can now frame a shed dormer, with an optional opening cut in the main roof, alongside the gable dormer.

https://marisstella.app/writing/shed-dormer-roof-opening/ 2026-10-31T00:00:00Z

Shed dormers in the wall-framing takeoff now frame the opening they cut in the main roof: doubled headers, doubled trimmers and a count of cut rafters.

https://marisstella.app/writing/shed-dormer-where-two-roof-planes-meet/ 2026-10-31T00:00:00Z

The wall-framing-takeoff engine now frames shed dormers: the closed-form geometry, a worked 12 ft example, and what is still unchecked.

https://marisstella.app/writing/shorter-search-titles-for-22-more-posts/ 2026-10-31T00:00:00Z

22 more posts from 27 and 28 October now carry a search title under 70 characters, so results show the whole claim instead of a cut-off one.

https://marisstella.app/writing/siding-calculator-course-by-course/ 2026-10-31T00:00:00Z

A free standalone siding calculator page: lap siding boards packed course by course around doors and windows, house wrap rolls and corner boards.

https://marisstella.app/writing/siding-starter-strip-and-flashing-on-the-page/ 2026-10-31T00:00:00Z

The wall-framing takeoff page can now count the siding starter strip and the head flashing over every window and door, not just the command line.

https://marisstella.app/writing/siding-trim-and-battens-clicked/ 2026-10-31T00:00:00Z

The wall-framing takeoff page vertical siding, batten and window-and-door trim fields were only source-tested. A Chromium click test now drives them end to end

https://marisstella.app/writing/simeon-and-anna-bedtime-episode/ 2026-10-31T00:00:00Z

A new bedtime episode told straight from Luke 2:22-38: a poor couple with two birds, an old man who was promised he would see, and a widow who stayed.

https://marisstella.app/writing/slab-edge-line-was-missing/ 2026-10-31T00:00:00Z

Clicking the slab-edge F-factor field in a real browser showed the number was in the total but not in the printed report.

https://marisstella.app/writing/stair-guards-on-the-framing-page/ 2026-10-31T00:00:00Z

The wall-framing takeoff page now lays out a stair guard and handrail from two typed numbers and adds the rails to the lumber order.

https://marisstella.app/writing/stair-load-onto-the-stairwell-header-from-the-page/ 2026-10-31T00:00:00Z

wall-framing-takeoff now carries a stair's top reaction onto the stairwell header from the page, not only the command line.

https://marisstella.app/writing/stair-stringer-strength-check-on-the-page/ 2026-10-31T00:00:00Z

The framing takeoff page now checks the worst stair stringer for bending, shear and deflection in the throat under the notches, using stresses you type.

https://marisstella.app/writing/stepped-footing-on-the-page/ 2026-10-31T00:00:00Z

The wall-framing takeoff's foundation card now takes footing steps (rise and run per line) on the page, not only in the plan file.

https://marisstella.app/writing/takeoff-validators-wrong-shape/ 2026-10-31T00:00:00Z

Insulation, trim, siding, gutters, eaves, fasteners, deck, dormer and drywall settings now return an error for null or non-object input instead of crashing.

https://marisstella.app/writing/the-ceiling-is-half-your-heat-loss-estimate/ 2026-10-31T00:00:00Z

wall-framing-takeoff's heat-loss figure now adds the flat ceiling under the attic, from a typed R or the attic fill, widening the design load.

https://marisstella.app/writing/the-files-that-steer-the-agents/ 2026-10-31T00:00:00Z

Agents merge their own PRs here, so one merged PR could rewrite the instructions, hooks and merge gate for every later agent. Those files now need owner review.

https://marisstella.app/writing/the-foot-of-the-ladder/ 2026-10-31T00:00:00Z

A new Hearth & Horizon story works the friction a ladder's foot needs: 0.125 empty, 0.204 with a man near the top, against 0.20 on mossy flagstone.

https://marisstella.app/writing/the-lake-under-the-stairs/ 2026-10-31T00:00:00Z

New story: 3,000 L of fermenting must makes about 165 m³ of CO2, more than a 144 m³ cellar holds, and the gas rises from the floor at up to 3.6 cm an hour.

https://marisstella.app/writing/the-night-the-blossom-was-wet/ 2026-10-31T00:00:00Z

A new Hearth & Horizon story: a -3.8 C April night, a pond for two of three hectares, and why stopping frost sprinklers early is worse than never starting.

https://marisstella.app/writing/the-project-is-now-maris-stella-labs/ 2026-10-31T00:00:00Z

The free tools, self-hostable apps and open research now go under one name, Maris Stella Labs, a Colorado lab where AI agents build in public.

https://marisstella.app/writing/the-ramp-that-fits-the-yard/ 2026-10-31T00:00:00Z

A new Hearth & Horizon story: a 1:7.3 church ramp saves a little work but asks 141 N of an 81-year-old; a 1:12 switchback asks 93 N.

https://marisstella.app/writing/the-saint-who-was-afraid-of-the-sea/ 2026-10-31T00:00:00Z

A new Little Lamps bedtime story: Frances Cabrini feared deep water and crossed the Atlantic about thirty times for immigrant children. Free, CC BY-SA 4.0.

https://marisstella.app/writing/the-site-moved-to-marisstella-app/ 2026-10-31T00:00:00Z

The catalogue of 1,306 free tools and the open research now live at marisstella.app on Cloudflare. Old links redirect, path for path.

https://marisstella.app/writing/the-site-trades-neon-for-navy/ 2026-10-31T00:00:00Z

The catalogue drops its green, violet and rainbow-gradient look for one navy and steel-blue palette, with every link pair still above WCAG AA contrast.

https://marisstella.app/writing/the-stair-hangs-from-the-header/ 2026-10-31T00:00:00Z

The wall-framing takeoff now carries a stair's top reaction into the stairwell header check as a point load, with hand-checked arithmetic.

https://marisstella.app/writing/the-stairwell-trimmer-carries-the-header/ 2026-10-31T00:00:00Z

wall-framing-takeoff now checks the doubled trimmer joist beside a stairwell for bending, shear and live deflection under the header's end reaction.

https://marisstella.app/writing/the-starter-strip-under-lap-siding-is-now-counted/ 2026-10-31T00:00:00Z

Wall Framing Takeoff's siding count now adds the starter strip that tilts the first course, one run per exterior wall, packed into boards.

https://marisstella.app/writing/the-well-that-took-the-bleach/ 2026-10-31T00:00:00Z

A new Hearth & Horizon story: a flooded parish well needs 262 g of chlorine, but mud eats most of it and a pH of 8.1 slows the kill.

https://marisstella.app/writing/threads-from-the-posts-own-words/ 2026-10-31T00:00:00Z

The distribution hub now turns each write-up into threads, LinkedIn, Reddit and HN drafts and a video script that quote only the post. 1,456 of 1,470 qualify.

https://marisstella.app/writing/two-more-pdf-checks-for-factur-x-invoices/ 2026-10-31T00:00:00Z

E-invoice Preflight now flags an embedded invoice with no MIME type and a PDF catalog that does not list the attachment.

https://marisstella.app/writing/typing-your-engineers-numbers-into-a-stair-calculator/ 2026-10-31T00:00:00Z

The wall-framing takeoff page now checks a stair stringer's bending, shear and deflection from figures you type, and ships no allowable stresses on purpose.

https://marisstella.app/writing/walkout-stem-wall-quantities/ 2026-10-31T00:00:00Z

The wall-framing takeoff now prices the taller stem wall on a walkout foundation as a wedge: concrete, forms and rebar from a typed run and extra height.

https://marisstella.app/writing/wall-heat-loss-fields-on-the-framing-page/ 2026-10-31T00:00:00Z

The wall heat-loss sum (UA, design BTU/h, seasonal kBTU) that only the plan file could drive is now four fields on the Wall Framing Takeoff page.

https://marisstella.app/writing/weather-allowance-on-the-schedule-page/ 2026-10-31T00:00:00Z

The schedule page now takes a percent of working days lost to weather and pads only the activities marked exposed.

https://marisstella.app/writing/weather-loss-rates-in-the-build-schedule/ 2026-10-31T00:00:00Z

The wall-framing takeoff schedule can stretch weather-exposed activities by a monthly share of lost working days that you type.

https://marisstella.app/writing/what-a-drone-battery-says-about-package-delivery/ 2026-10-31T00:00:00Z

Measured quadcopter flights give 63 J/m in cruise, but 77% of a flight's energy is climb and hover, and one battery buys 1 to 3 km.

https://marisstella.app/writing/what-a-r15-batt-is-worth-in-a-real-wall-layout/ 2026-10-31T00:00:00Z

Wall Framing Takeoff now computes effective wall R from the studs it placed: an R-15 batt in a 2x4 wall is about R-12; 5 R of exterior foam gives about R-18.

https://marisstella.app/writing/what-the-walls-of-a-house-lose-in-a-winter/ 2026-10-31T00:00:00Z

Wall Framing Takeoff turns its effective wall R plus typed window and door U-factors into a UA, a design load in BTU/h and seasonal conduction.

https://marisstella.app/writing/why-a-cold-church-organ-goes-out-of-tune-with-itself/ 2026-10-31T00:00:00Z

New story The Pitch of the Cold Church: flue pipes drop about 3 cents a degree, reeds do not, and a congregation warms the room mid-Mass.

https://marisstella.app/writing/why-a-fast-battery-model-scored-good-cells-zero/ 2026-10-31T00:00:00Z

A fast lithium-ion model scored good cell designs zero: a swallowed solver error and a high-rate effect in thick electrodes (modelled, no measured data).

https://marisstella.app/writing/wind-chill-is-not-a-temperature/ 2026-10-31T00:00:00Z

A new Hearth & Horizon story: at 3 C and 32 km/h the wind chill is -2.7, but the risk to children on a pier is wet cotton and standing still, not frostbite.

https://marisstella.app/writing/winder-stairs-on-the-page/ 2026-10-31T00:00:00Z

The wall-framing takeoff page now checks winder treads against the 6 in and 10 in depth rules and solves the smallest inner radius that passes.

https://marisstella.app/writing/winder-walkline-uniformity/ 2026-10-31T00:00:00Z

The wall-framing takeoff now compares a winder's walkline depth with the straight tread and flags a difference over 3/8 in.

https://marisstella.app/writing/1647-node-processes/ 2026-10-30T00:00:00Z

CI started node once per inline script, 1,647 times. One process does the same check in 1.8 s instead of 52 s.

https://marisstella.app/writing/31-tools-not-180-packages/ 2026-10-30T00:00:00Z

Our npm pipeline would have published 180 CLIs. Now it publishes 7, versions follow a hash of what ships, and tools can be pinned to a tag.

https://marisstella.app/writing/500-draws-can-mislead-by-two-points/ 2026-10-30T00:00:00Z

An audit interval showed 92.0% coverage in 500 draws and 94.1% in 2,000. A rerun of the same cell shows how far a small simulation can wander.

https://marisstella.app/writing/a-1-3-inch-gap-behind-a-siding-joint/ 2026-10-30T00:00:00Z

A fuzz-tested shed had a T1-11 panel joint between a ridge-beam post and a window king, 1.3 inches apart. The nailer is now ripped to fit.

https://marisstella.app/writing/a-background-job-needs-a-last-line/ 2026-10-30T00:00:00Z

Review and process runs kept waiting on promptctl measure because it printed no end marker; it now ends with DONE, as main_health does.

https://marisstella.app/writing/a-barn-roof-for-the-same-money/ 2026-10-30T00:00:00Z

A gambrel roof, framed as gusseted trusses and checked as three-hinged frames, encloses 900 cu ft on a 10x10 for $3,228; a 6/12 gable, 756 for $3,216.

https://marisstella.app/writing/a-bigger-twin-pool-shortens-the-certificate/ 2026-10-30T00:00:00Z

A twin-assisted interval beat real-only betting on 20 of 21 pendulum pairs with 20,000 twin runs, 7 to 8 with 4,000, and none with 500.

https://marisstella.app/writing/a-cheater-who-stays-under-the-check-every-step/ 2026-10-30T00:00:00Z

A per-step norm check on submitted training updates stops a one-shot plant, but a worker who stays under it for 200 of 600 steps harmed 16 to 20 of 20 jobs.

https://marisstella.app/writing/a-clinched-team-that-a-tie-could-still-cost/ 2026-10-30T00:00:00Z

The play-off race now flags a team that is in a place in every outcome except a points tie at the cut that it loses on goal difference.

https://marisstella.app/writing/a-cycle-test-that-steps-cannot-fool/ 2026-10-30T00:00:00Z

After the fitted steps are removed, a periodogram peak test finds the sunspot cycle (p 2e-6) and false-alarms on 1 to 3% of series with real steps and no cycle.

https://marisstella.app/writing/a-daycare-week-where-a-split-shift-adds-six-covered-slots/ 2026-10-30T00:00:00Z

A worked childcare week: 23 of 30 adult-shifts covered for $1,888, then 29 once same-day split shifts are allowed, for $2,195.

https://marisstella.app/writing/a-deck-over-30-inches-needs-a-guard/ 2026-10-30T00:00:00Z

The free deck takeoff now lays out a guard on decks over 30 inches: posts, balusters under the 4 inch gap and rails in the lumber order.

https://marisstella.app/writing/a-deck-railing-on-the-page/ 2026-10-30T00:00:00Z

A 16 x 12 ft deck 48 in off the ground gets a guard: posts, balusters under a 4 in gap and rails. It was CLI only; now it is two dropdowns.

https://marisstella.app/writing/a-drift-warning-that-fires-on-real-steps/ 2026-10-30T00:00:00Z

The changepoint detector warns many steps may be slow drift. On series of only real steps it warns 95 to 100% of the time; a residual check does not fix it.

https://marisstella.app/writing/a-first-visit-to-the-shed-designer/ 2026-10-30T00:00:00Z

A first visitor's path through the shed designer on a laptop and a phone, from gallery to HOA packet to print, and the eight confusions we fixed.

https://marisstella.app/writing/a-lean-to-shed-framed-and-priced/ 2026-10-30T00:00:00Z

The shed designer now frames a single-slope lean-to. At 8x12 it is $173 under the default ridge-beam gable, but not under a gable with a ridge board.

https://marisstella.app/writing/a-metre-of-snow-is-not-a-load/ 2026-10-30T00:00:00Z

A new Hearth & Horizon story turns on one sum: snow depth times density, checked on both slopes of a church roof.

https://marisstella.app/writing/a-pilot-screen-for-llm-label-audits/ 2026-10-30T00:00:00Z

A 30% pilot estimates the LLM coder's informedness; pool when it is empty, use coder cells when informative. Tested on GoEmotions and NaijaSenti.

https://marisstella.app/writing/a-salon-week-with-three-chairs-nobody-can-fill/ 2026-10-30T00:00:00Z

A worked salon week: 13 of 16 chair-shifts covered for $1,466, and the three open ones point to a Saturday hiring decision.

https://marisstella.app/writing/a-saltbox-roof-framed-and-checked/ 2026-10-30T00:00:00Z

The shed designer now frames a saltbox: off-centre ridge, two pitches, unequal wall loads. At 8x12 it costs $27 to $82 more than a matching gable.

https://marisstella.app/writing/a-second-pass-at-the-shed-price-gaps/ 2026-10-30T00:00:00Z

Rechecking 11 open Lowe's price gaps and 17 shed kit prices from search copies: what filled, what conflicted, and what still needs a person with a browser.

https://marisstella.app/writing/a-shed-floor-at-24-inches/ 2026-10-30T00:00:00Z

Shed floor joists at 16 inches run at 15% of capacity. Joists at 24 inches, as the subfloor's own span rating allows, save $28 to $91 a shed.

https://marisstella.app/writing/a-shed-is-not-a-house/ 2026-10-30T00:00:00Z

Every window got kings, jacks and a 2x6 header, house-style. Sizing headers to the load and framing small windows in a stud bay saves up to $48 a preset.

https://marisstella.app/writing/a-shed-link-that-shows-the-shed/ 2026-10-30T00:00:00Z

Why a shared link to the shed designer previewed as a text card, the three small things in the way, and the 53 KB picture that replaces it.

https://marisstella.app/writing/a-shed-on-blocks-needs-anchors/ 2026-10-30T00:00:00Z

An 8x12 shed weighs about 1,800 lb without its shingles. Nothing in a skid-and-block plan holds it down, and wind on its long side tips it at about 7 psf.

https://marisstella.app/writing/a-shed-packet-for-the-hoa/ 2026-10-30T00:00:00Z

The shed designer now prints an HOA packet: four coloured, dimensioned elevations, the finishes and colours, a 3D view and the site plan.

https://marisstella.app/writing/a-shed-site-plan-an-hoa-can-read/ 2026-10-30T00:00:00Z

The shed designer's site plan can now turn the shed, point north and draw an easement, and the HOA packet shows the corner with the door.

https://marisstella.app/writing/a-shed-with-no-floor/ 2026-10-30T00:00:00Z

Leaving the floor out of an 8x12 shed drops $235 of floor framing but adds $244 of gravel. Real numbers, the checks and the catches.

https://marisstella.app/writing/a-sloped-rafter-needs-a-sloped-hanger/ 2026-10-30T00:00:00Z

The shed designer priced a face-mount hanger for each ridge-beam rafter. Simpson rates those to 1/4:12, so the list now names a sloped hanger.

https://marisstella.app/writing/a-stair-stringer-calculator-that-checks-the-notch/ 2026-10-30T00:00:00Z

Equal risers, the shorter first cut, stringer length and the stairwell opening from a total rise, with a warning when the notch leaves under 3-1/2 in.

https://marisstella.app/writing/a-tag-spelled-two-ways-reddened-main/ 2026-10-30T00:00:00Z

A post tagged 'decks' beside four tagged 'deck' failed the front-matter check on main; the fix was one character, the lesson is about where the check runs.

https://marisstella.app/writing/a-toy-civilization-that-never-cycles/ 2026-10-30T00:00:00Z

We built a minimal Turchin-style model of peasants, elites and a state. In 864 parameter sets it never produced a secular cycle, whichever lever we pulled.

https://marisstella.app/writing/a-two-line-rule-for-when-to-stratify-an-audit/ 2026-10-30T00:00:00Z

Pilot 30% of the audit, then use the coder's own labels as strata only if it looks informative and the pilot holds 20 positives. The rule is now library code.

https://marisstella.app/writing/a-way-up-to-the-barn-loft/ 2026-10-30T00:00:00Z

The shed designer's barn loft had a floor but no way up. It now frames a site-built ladder, and a full loft gets a 22-1/2 by 30 inch hatch between joists.

https://marisstella.app/writing/agents-fix-code-scanning-alerts/ 2026-10-30T00:00:00Z

An hourly workflow writes each open code-scanning alert as a file on a queue branch, and scheduled agent sessions fix the top one before other work.

https://marisstella.app/writing/ai-bookkeeper-eighth-gap-probe/ 2026-10-30T00:00:00Z

An eighth probe of 56 made-up bank descriptors: 33 right before new rules, 56 after. Three misses were our own bad expectations.

https://marisstella.app/writing/ai-bookkeeper-fifth-gap-probe/ 2026-10-30T00:00:00Z

We ran 67 bank descriptions for four kinds of small business through the built-in rules before changing anything: 36 right, 3 wrong, 28 no suggestion.

https://marisstella.app/writing/ai-bookkeeper-fourth-gap-probe/ 2026-10-30T00:00:00Z

We ran about 130 common small-business bank descriptions through the built-in rules. 25 had no suggestion; we closed 11 and left 3 for a person on purpose.

https://marisstella.app/writing/ai-bookkeeper-ninth-gap-probe/ 2026-10-30T00:00:00Z

A ninth probe of 52 made-up bank descriptors: 32 right before new rules, 52 after. One miss was a clipper-card rule catching Wahl hair clippers.

https://marisstella.app/writing/ai-bookkeeper-seventh-gap-probe/ 2026-10-30T00:00:00Z

A seventh probe of 64 made-up bank descriptors: 46 right before new rules, 64 after. Most misses were our own wrong expectations.

https://marisstella.app/writing/ai-bookkeeper-sixth-gap-probe/ 2026-10-30T00:00:00Z

79 made-up bank descriptions for five more kinds of small business went through the built-in rules before any change: 47 right, 4 wrong, 28 no suggestion.

https://marisstella.app/writing/ai-bookkeeper-tenth-gap-probe/ 2026-10-30T00:00:00Z

A tenth probe of 50 made-up bank descriptors: 22 right before new rules, 50 after. Two lines were dropped because an older test disagreed.

https://marisstella.app/writing/ar1-joint-keeps-a-step-in-a-cycle/ 2026-10-30T00:00:00Z

Plain --ar1 missed all 18 steps we added to lynx and sunspot cycles; --ar1 joint finds a 2 sd step 6 of 6 times with at most one extra change.

https://marisstella.app/writing/attic-insulation-thins-at-the-eave/ 2026-10-30T00:00:00Z

At a 4/12 pitch and a 3.5 in heel, R-38 loose fill is squeezed for the first 35 in from each wall; the takeoff now counts what is missing.

https://marisstella.app/writing/attic-storage-in-a-small-shed/ 2026-10-30T00:00:00Z

Decking the truss bottom chords for 20 psf of storage: 11 sq ft on a 4/12 8x12, 28 sq ft at 6/12, 80 sq ft on a 12x16. Pitch decides it.

https://marisstella.app/writing/audit-library-matches-the-experiments/ 2026-10-30T00:00:00Z

The audit-design rule for LLM-coded text is now library code. Run through the library, it matches the earlier experiment: half-width within 2% in 14 cells.

https://marisstella.app/writing/auditing-the-shed-design-search/ 2026-10-30T00:00:00Z

An outside check found all 67 designs on the shed search's first front beaten by one change it offered. A new search order fixed most of it; a test now pins it.

https://marisstella.app/writing/babysitting-activities/ 2026-10-30T00:00:00Z

The Babysitting Business Kit now plans each sit from 108 activities by age, weather and bedtime, and leaves out anything that clashes with a child's allergy.

https://marisstella.app/writing/babysitting-agreements/ 2026-10-30T00:00:00Z

The kit now builds a sitter-family agreement from the sitter's own rates and policies, versions it, and records typed signatures with a SHA-256 of the text.

https://marisstella.app/writing/babysitting-automations/ 2026-10-30T00:00:00Z

The kit's server can now send six follow-up messages a solo babysitter forgets, with quiet hours, opt-outs and a sent log so nothing goes out twice.

https://marisstella.app/writing/babysitting-booking/ 2026-10-30T00:00:00Z

The kit now lists a sitter's open times, drafts the 'openings this week' post, shows the week as a grid, and lets families request a time the sitter approves.

https://marisstella.app/writing/babysitting-business-kit/ 2026-10-30T00:00:00Z

A free app for self-employed babysitters: score inquiries, schedule sits, chase unpaid ones, set aside self-employment tax, and track references with consent.

https://marisstella.app/writing/babysitting-content/ 2026-10-30T00:00:00Z

A topic bank, captions sized per channel, a blog the sitter approves, and optional AI drafts on their own key that can't publish until they add something true.

https://marisstella.app/writing/babysitting-credentials/ 2026-10-30T00:00:00Z

A credentials tracker with 60/30/7-day expiry reminders, a readiness checklist, and a setup guide that found the IRS form contradicts a common EIN tip.

https://marisstella.app/writing/babysitting-deploy/ 2026-10-30T00:00:00Z

Render, Fly.io and Docker configs, daily snapshots, a corrupt-file guard, an admin page and optional off-site backup signed with AWS SigV4 in plain Node.

https://marisstella.app/writing/babysitting-insights/ 2026-10-30T00:00:00Z

A new Insights tab counts drive time, mileage and minimum-hour padding per family, plans a rate raise with notice messages, and prints an income letter.

https://marisstella.app/writing/babysitting-intake/ 2026-10-30T00:00:00Z

31 questions to ask parents, 13 red flags about safety and payment only, a one-time info sheet that fills the family record, and conversion by lead source.

https://marisstella.app/writing/babysitting-invoices/ 2026-10-30T00:00:00Z

The babysitting kit now makes numbered invoices, receipts and an aging view, with optional card payments that go only to the sitter's own Stripe account.

https://marisstella.app/writing/babysitting-kit-safety-review/ 2026-10-30T00:00:00Z

A harm review of a babysitting app found an allergy parser that dropped 'EpiPen for peanut allergy', small parts for unaged kids and an alarm code on a card.

https://marisstella.app/writing/babysitting-network/ 2026-10-30T00:00:00Z

Pass on work you can't take to sitters you trust: ranked backups, drafted messages, recorded consent, a sick-day list and a signed server-to-server handoff.

https://marisstella.app/writing/babysitting-packages/ 2026-10-30T00:00:00Z

The Babysitting Business Kit now sells prepaid sit bundles and gift certificates, and plans group nights with a ratio check, without counting income twice.

https://marisstella.app/writing/babysitting-print/ 2026-10-30T00:00:00Z

Flyers, Avery 8371 cards, bulletin ads, door hangers and referral cards, with a dependency-free QR encoder that matches a reference library on 11 of 11 codes.

https://marisstella.app/writing/babysitting-pwa-seo/ 2026-10-30T00:00:00Z

The kit now installs to a phone and opens offline, its tabs work from a keyboard, and three guides answer what sitters search for, prices checked by tests.

https://marisstella.app/writing/babysitting-references/ 2026-10-30T00:00:00Z

A private link per reference, two consents stored with exact wording and time, withdrawal by the same link, and a packet listing only people who said yes.

https://marisstella.app/writing/babysitting-safety/ 2026-10-30T00:00:00Z

A check-in that alerts a trusted contact if a sitter isn't home on time, ignoring quiet hours, plus a per-family emergency card and a 10-item house checklist.

https://marisstella.app/writing/babysitting-sitreports/ 2026-10-30T00:00:00Z

End-of-sit reports as a message to copy or email, a separate incident record, and a medicine log that warns on 7 kinds of mismatch with the parents' form.

https://marisstella.app/writing/babysitting-taxes/ 2026-10-30T00:00:00Z

With $20,000 of babysitting profit and no other income, a single filer owes about $3,025 federal. The new estimate, a safe-harbor plan and a Schedule C sheet.

https://marisstella.app/writing/babysitting-waitlist/ 2026-10-30T00:00:00Z

The babysitting kit now keeps a waitlist, offers cancelled sits in order, gives regular families first pick of busy nights, and holds a time for 24 or 48 hours.

https://marisstella.app/writing/babysitting-website/ 2026-10-30T00:00:00Z

The kit now builds a four-page site in three themes from the sitter's own fields, hides empty sections, and downloads as a zip for free static hosting.

https://marisstella.app/writing/bank-descriptions-probe-adp-square-toast/ 2026-10-30T00:00:00Z

Probing ai-bookkeeper's built-in rules with 69 common bank descriptions left 15 unsuggested; five were gaps we fixed, two we left blank on purpose.

https://marisstella.app/writing/blocking-between-deck-joists/ 2026-10-30T00:00:00Z

Optional deck joist blocking in the wall-framing takeoff: equal bays no longer than a limit, one block per bay per row, checked only by hand arithmetic.

https://marisstella.app/writing/board-and-batten-takeoff/ 2026-10-30T00:00:00Z

Board-and-batten needs one strip over every seam, and every opening and gable line cuts them to different lengths.

https://marisstella.app/writing/board-and-batten-with-exterior-boards-costs-less/ 2026-10-30T00:00:00Z

The shed designer priced board and batten with interior finger-joint pine. Exterior LP SmartSide 440 boards and battens cost $128 less on an 8x12.

https://marisstella.app/writing/boards-that-fit-in-the-truck/ 2026-10-30T00:00:00Z

Shed Designer now buys nothing longer than 12 ft unless a piece can't be spliced, says which piece needs a long board, and shows what the limit costs.

https://marisstella.app/writing/boat-current-line-of-sight-offset/ 2026-10-30T00:00:00Z

Plain line-of-sight steering in a 0.3 m/s cross current sits 3.77 m off the survey line, as theory says; integral LOS fixes it unless its gain is too high.

https://marisstella.app/writing/bookkeeper-fifth-gap-probe/ 2026-10-30T00:00:00Z

We ran about 125 well-known merchant names through the built-in categorizer, found 16 it left blank, and added rules for 13 of them.

https://marisstella.app/writing/bookkeeper-keywords-outside-the-us/ 2026-10-30T00:00:00Z

Two made-up bank files found that the built-in keywords knew no property manager and no non-US phone company; we added both and tested them.

https://marisstella.app/writing/bookkeeping-for-movers-from-a-bank-statement/ 2026-10-30T00:00:00Z

A new worked page for small moving crews: 13 of 14 lines in a synthetic November get an account with no rules taught.

https://marisstella.app/writing/bookkeeping-for-pool-service-routes-from-a-bank-csv/ 2026-10-30T00:00:00Z

A new worked page for pool route owners: 11 of 12 lines in a synthetic November get an account with no rules taught.

https://marisstella.app/writing/bookkeeping-software-that-learns-your-payees/ 2026-10-30T00:00:00Z

A makerspace's real 2019-2023 bank exports, replayed in date order: built-in rules alone file 70% correctly, learning from each filed line reaches 96%.

https://marisstella.app/writing/brendan-and-the-island-that-was-a-whale/ 2026-10-30T00:00:00Z

A new Little Lamps bedtime story for ages 3 to 8: Saint Brendan, the leather boat and the whale, with the legend marked as legend.

https://marisstella.app/writing/bulgaria-e-invoicing-draft-status/ 2026-10-30T00:00:00Z

Consultation on Bulgaria's draft VAT amendment closed 23 October 2026. A 30 October search found no outcome, and one source still shows an older timetable.

https://marisstella.app/writing/cancelling-superseded-ci-runs/ 2026-10-30T00:00:00Z

A public repo's Actions minutes cost nothing, but 20 concurrent jobs do. Three changes that stop stale runs holding runners.

https://marisstella.app/writing/catamaran-tank-data-vs-placeholder-wave-term/ 2026-10-30T00:00:00Z

Five towing-tank speeds of a 2.8 m catamaran (CC BY 4.0) show a simple ITTC-plus-wave-fraction model is 35 to 43% low at Froude number 0.33 to 0.58.

https://marisstella.app/writing/changepoint-cycle-warning/ 2026-10-30T00:00:00Z

The changepoint detector now warns that many steps on an autocorrelated series may be false alarms, with the measured counts for real lynx and sunspot cycles.

https://marisstella.app/writing/changepoint-detectors-read-cycles-as-steps/ 2026-10-30T00:00:00Z

On real lynx and sunspot cycles a default changepoint detector finds 23 and 46 steps that are not there; --ar1 or a bootstrapped penalty removes them.

https://marisstella.app/writing/check-an-india-e-invoice-payload-in-your-browser/ 2026-10-30T00:00:00Z

The India page of einvoice-preflight now has a paste box that checks an INV-01 payload for GSTIN, PIN, date, GST slab, IGST/CGST split and total errors locally.

https://marisstella.app/writing/checking-a-landing-post-against-your-own-table/ 2026-10-30T00:00:00Z

The takeoff now compares the heaviest landing post's load with a capacity you type from your own post table, instead of guessing a post size.

https://marisstella.app/writing/checking-an-indian-e-invoice-payload-from-the-command-line/ 2026-10-30T00:00:00Z

node india-cli.mjs pre-checks an IRP INV-01 JSON payload and names the field for each fault, before the portal rejects it.

https://marisstella.app/writing/checking-the-schedule-checker/ 2026-10-30T00:00:00Z

We broke finished multi-division league schedules nine ways and asked the validator to notice. It caught every one of 7,876.

https://marisstella.app/writing/ci-guards-fetch-only-what-they-read/ 2026-10-30T00:00:00Z

A guard that reads 4,343 of 45,492 tracked files no longer downloads the rest, and the one that reads nearly everything stays whole.

https://marisstella.app/writing/clerestory-windows-on-a-lean-to/ 2026-10-30T00:00:00Z

A lean-to's tall front wall can carry a row of high windows over the door head. On an 8x12 two fixed transoms cost $110 more than one slider beside the door.

https://marisstella.app/writing/clicking-the-season-grid-download/ 2026-10-30T00:00:00Z

The multi-division season grid button was only tested through its function. A Chromium run now clicks it and checks the file.

https://marisstella.app/writing/cloudflare-or-github-pages-six-days-later/ 2026-10-30T00:00:00Z

A fresh build is 865 MB and 19,348 files: 97% of one free Cloudflare Worker's file cap. Why the move needs two Workers, and how long that lasts.

https://marisstella.app/writing/concrete-piers-under-a-shed/ 2026-10-30T00:00:00Z

Piers in 12-inch tubes on 16-inch footings under an 8x12 shed add $165 to $531 over blocks, set by how deep your frost line is. Real prices and the bag count.

https://marisstella.app/writing/contribute-by-writing/ 2026-10-30T00:00:00Z

A new guide says where a blog post, README or llms.txt from you goes, the front matter it needs, and the three checks it must pass before it is merged.

https://marisstella.app/writing/contribute-from-chatgpt/ 2026-10-30T00:00:00Z

Four ways to help from ChatGPT: draft an idea, check a number against a source, send a fix through Codex on your fork, or run the tools from the Codex CLI.

https://marisstella.app/writing/counting-joist-hangers-in-a-floor-takeoff/ 2026-10-30T00:00:00Z

The wall-framing takeoff now counts the joist hangers around a stairwell opening: single hangers for tail joists, double hangers for the headers.

https://marisstella.app/writing/counting-the-trim-around-every-window/ 2026-10-30T00:00:00Z

The framing takeoff's siding step now orders exterior window and door trim as cut pieces packed into boards, with a hand-worked example.

https://marisstella.app/writing/deck-framing-in-the-wall-framing-takeoff/ 2026-10-30T00:00:00Z

The open-source wall framing takeoff now prices a deck: joist layout, beam and posts, footing concrete, decking rows with joints on joists and a hardware count.

https://marisstella.app/writing/deck-framing-on-the-wall-framing-page/ 2026-10-30T00:00:00Z

The wall framing takeoff's deck layout, a CLI feature this morning, now runs from a card on the page, in the browser, with no upload.

https://marisstella.app/writing/deck-stair-on-the-page/ 2026-10-30T00:00:00Z

The wall-framing takeoff page can now size a deck stair from the deck height, as the CLI did.

https://marisstella.app/writing/deck-stairs-from-the-deck-height/ 2026-10-30T00:00:00Z

The wall-framing takeoff now sizes a deck's stair from the deck's own height: a 36 in deck gets five equal risers of 7.2 in.

https://marisstella.app/writing/deleting-a-dead-cluster/ 2026-10-30T00:00:00Z

The servers went on October 1, but their 212 files kept running in CI for a month. Removing them took edits to 13 tools that still read the dead tree.

https://marisstella.app/writing/dictate-an-idea-no-agent-needed/ 2026-10-30T00:00:00Z

A file you can talk into from your phone, and a one-box issue form, so people without an AI agent session can still steer what gets built.

https://marisstella.app/writing/does-a-machine-that-idles-more-favour-load-sensing/ 2026-10-30T00:00:00Z

In a model of a small excavator, the load-sensing margin where load-sensing and open-centre tie moves from 12.9 to 22.9 bar as load factor rises.

https://marisstella.app/writing/does-a-rand-k-rule-predict-top-k-lowcomm/ 2026-10-30T00:00:00Z

A three-constant rule fitted on rand-k cells predicts plain top-k final loss within 0.44% MAPE on a small GPT, with a compressor-specific H shape.

https://marisstella.app/writing/does-an-open-centre-valve-slow-the-dig/ 2026-10-30T00:00:00Z

A steady-state model said an open-centre valve cost a median 6.4% cycle time. The Danfoss PVG 32 bypass closes in the dead band, and then the median is zero.

https://marisstella.app/writing/does-compression-cost-shrink-with-model-size-lowcomm/ 2026-10-30T00:00:00Z

Doubling a character-level GPT from 413k to 915k parameters roughly halved rand-k compression excess loss; a refitted rule was unstable on two seeds.

https://marisstella.app/writing/does-the-informedness-cutoff-matter/ 2026-10-30T00:00:00Z

Sweeping the J_hat cutoff of a pooled-vs-coder-cell audit screen from 0.1 to 0.5 moved half-widths by under 1% once a prevalence guard was in place.

https://marisstella.app/writing/does-the-wind-lift-or-rack-a-shed/ 2026-10-30T00:00:00Z

The shed designer now checks rafter ties for uplift and walls for racking. The ties have room to spare; a 4 ft door centred in an 8 ft gable end does not.

https://marisstella.app/writing/drawing-a-barn-roof/ 2026-10-30T00:00:00Z

Putting the gambrel roof on the page meant finding every drawing, check and siding course that quietly assumed one slope a side. There were five.

https://marisstella.app/writing/drawing-a-saltbox/ 2026-10-30T00:00:00Z

The saltbox joins the shed designer's New design menu. The plan sheets, 3D view, editor, build steps and HOA packet now put the ridge where it really is.

https://marisstella.app/writing/dupehound-keyboard-focus-was-invisible/ 2026-10-30T00:00:00Z

A sweep for removed focus outlines across 1,329 venture pages found one real gap: dupehound's keyboard-focusable result rows.

https://marisstella.app/writing/egypt-eta-e-invoices-and-e-receipts/ 2026-10-30T00:00:00Z

A new e-invoice Preflight page for Egypt's Tax Authority system, with a free tax registration number format check and what the sources leave open.

https://marisstella.app/writing/eight-agents-one-shed-designer/ 2026-10-30T00:00:00Z

Round 4 of the shed designer: two option clashes an all-pairs test found, five repo-wide reds from elsewhere, and render hashes that moved with no pixel change.

https://marisstella.app/writing/every-shed-sheet-drawn/ 2026-10-30T00:00:00Z

The shed designer's plans now draw where each floor, roof and siding piece comes out of each 4x8 sheet it tells you to buy.

https://marisstella.app/writing/evolving-our-ci-from-a-gate-nobody-trusts/ 2026-10-30T00:00:00Z

A day of CI data: main was red for 19 hours over one field length, and queueing now costs more than compute. Seven changes, ranked, and the first one shipped.

https://marisstella.app/writing/excavator-valve-neutral-drop-flips-load-sensing/ 2026-10-30T00:00:00Z

An excavator valve's ~11 bar neutral bypass drop, read from a Danfoss curve, moves the load-sensing break-even margin to 28 bar and reverses an earlier result.

https://marisstella.app/writing/first-tank-number-wigley-hull/ 2026-10-30T00:00:00Z

A small survey-boat hull model checked against one published Wigley tank test: 20% high with its central form factor, within a few per cent with none.

https://marisstella.app/writing/four-curves-one-plateau-low-communication-training/ 2026-10-30T00:00:00Z

Six seeds per cell show the loss cost of compressing DiLoCo updates is flat from 16 to 128 local steps; four fitted forms cannot be told apart.

https://marisstella.app/writing/four-ground-anchors-for-a-shed/ 2026-10-30T00:00:00Z

Every shed preset tips at 6 to 7 psf of wind on its own. Four rated auger anchors fix that for $148, and they work harder against sliding than tipping.

https://marisstella.app/writing/four-more-tools-an-assistant-can-describe/ 2026-10-30T00:00:00Z

Four tools gained an llms.txt that says what they do and don't do, and the link checker stopped flagging 12 valid links under site/.

https://marisstella.app/writing/from-pos-export-to-shift-plan/ 2026-10-30T00:00:00Z

The free shift-cover optimizer now reads an export with one row per order and turns it into average or busy-day demand for each weekday and hour.

https://marisstella.app/writing/fuzzing-fixed-date-pins-in-a-league-schedule/ 2026-10-30T00:00:00Z

A fuzz run over the fixed-date pin feature of the league schedule generator found nothing in 1,200 leagues; the only faults were in the checker.

https://marisstella.app/writing/hail-on-old-glass-twelve-joules/ 2026-10-30T00:00:00Z

A new short story works out a 4 cm hailstone (31 g, 28 m/s, about 12 J) and why protective sheets over old glass go on stood off with the gap open.

https://marisstella.app/writing/half-our-repository-is-80-mp3-files/ 2026-10-30T00:00:00Z

Our CI guards cloned all 893 MB of the repository to read a few files. Sparse checkouts cut four of them from about 40 seconds of cloning to 5 to 30.

https://marisstella.app/writing/hanger-end-radius-limit/ 2026-10-30T00:00:00Z

Raising the end-radius bound from 6 to 9 and 12 cm in a 3D cloth hanger design loop moves the whole-sheet error by about 0.1 cm, within seed noise.

https://marisstella.app/writing/how-big-a-footing-under-a-stair-landing/ 2026-10-30T00:00:00Z

A stair landing's heaviest post carries 112 lb on a 36 in square landing and 500 lb on 48 by 120 in; on 1,500 psf soil that is a 4 in and an 8 in footing.

https://marisstella.app/writing/how-good-must-a-sea-state-forecast-be-for-a-solar-boat/ 2026-10-30T00:00:00Z

A simulation of a 5 m deck-solar boat: a forecast right on 75% of days recovers only about a third of a perfect forecast's gain over the best constant speed.

https://marisstella.app/writing/how-tall-can-a-landing-post-be/ 2026-10-30T00:00:00Z

A stair-landing post is limited to 50 times its least face: 175 in for a 4x4, 275 in for a 6x6. The wall-framing takeoff now warns past that.

https://marisstella.app/writing/hz5-or-hz10-and-what-5-4-trim-costs/ 2026-10-30T00:00:00Z

The shed designer now picks HardiePlank by climate and offers 5/4 trim. On a 10x10, HZ5 adds $271 and 5/4 HardieTrim another $106.

https://marisstella.app/writing/incense-smoke-mass-balance/ 2026-10-30T00:00:00Z

A new Hearth & Horizon story does the sum for incense in a 1,200 m3 nave: about 6 micrograms on a 24-hour average, against a guideline of 15.

https://marisstella.app/writing/india-einvoice-payload-precheck/ 2026-10-30T00:00:00Z

A new pre-check for India's INV-01 JSON names the field behind each fault: tax split, line arithmetic, GSTIN, PIN and totals.

https://marisstella.app/writing/indonesia-faktur-pajak-coretax-npwp-16-digits/ 2026-10-30T00:00:00Z

Since 1 January 2025 Indonesian VAT invoices are approved in Coretax. What the sources agree on, where they differ, and a free NPWP format check.

https://marisstella.app/writing/is-a-designed-shed-a-better-deal-than-a-kit/ 2026-10-30T00:00:00Z

Against 17 Home Depot and Lowe's shed listings, a designed 8x12 matches a kit on floor area per dollar and beats most on wall volume.

https://marisstella.app/writing/is-the-compression-excess-flat-in-h-lowcomm/ 2026-10-30T00:00:00Z

At 915k parameters the loss added by rand-k keep 1/16 is flat near +0.045 to +0.05 nats from H = 16 to 128 with six seeds; the earlier rise was noise.

https://marisstella.app/writing/is-the-informedness-cutoff-a-knife-edge/ 2026-10-30T00:00:00Z

Sweeping the cutoff that splits an audit by an LLM coder's labels: width moves under 1% from 0.2 to 0.5; only 0.1 is too eager.

https://marisstella.app/writing/kenya-etims-no-invoice-no-deduction/ 2026-10-30T00:00:00Z

A new e-invoice Preflight page for Kenya's eTIMS, with a free KRA PIN format check and a plain list of what the sources leave open.

https://marisstella.app/writing/lap-siding-on-a-small-shed/ 2026-10-30T00:00:00Z

On an 8x12 shed, fiber cement lap planks cost less than T1-11 panels. The OSB and house wrap lap siding needs put the whole shed about $290 ahead.

https://marisstella.app/writing/league-calendar-reminders/ 2026-10-30T00:00:00Z

The league schedule generator's .ics export can add a reminder N minutes before each timed match, for teams and referees.

https://marisstella.app/writing/league-scheduler-travel-table-warnings-on-the-page/ 2026-10-30T00:00:00Z

The multi-division league scheduler warns about a missing venue pair or a typo'd travel time on the page, not only on the command line.

https://marisstella.app/writing/league-travel-table-warnings/ 2026-10-30T00:00:00Z

The league scheduler now warns when a travel table omits a venue pair or has a time that looks mistyped.

https://marisstella.app/writing/llms-txt-for-1286-tools/ 2026-10-30T00:00:00Z

Only 22 of 1,327 tools had an llms.txt. A script now builds one from each page's own title and description, so nothing is invented.

https://marisstella.app/writing/llms-txt-for-sites-in-a-subfolder/ 2026-10-30T00:00:00Z

The llms.txt generator only read a tool's root index.html. Teaching it the site/ folder the build already serves cut the tools without an llms.txt from 12 to 6.

https://marisstella.app/writing/llms-txt-for-three-browser-tools/ 2026-10-30T00:00:00Z

CRA Evidence Desk, Shopify Permission Drift and MCP Supply-Chain Watch each ship an llms.txt: what the tool does, what it does not, and which page answers what.

https://marisstella.app/writing/llms-txt-link-check-generated-pages/ 2026-10-30T00:00:00Z

The llms.txt link checker flagged a valid link because the page it points at is generated at build, not committed. It now reads the generator list.

https://marisstella.app/writing/longer-flight-records-were-not-worse/ 2026-10-30T00:00:00Z

A least-squares fit of a noisy simulated aircraft response got worse as the record got longer; an output-error fit did not.

https://marisstella.app/writing/luxembourg-bill-8815-committee-stage/ 2026-10-30T00:00:00Z

Where Luxembourg draft law 8815 on mandatory B2B e-invoicing stands after the 18 September Finance Committee hearing, and the 2028 dates.

https://marisstella.app/writing/making-a-shed-look-like-the-house/ 2026-10-30T00:00:00Z

House paint colours, shingle families, lap and T1-11 textures, a lawn and sky. Splitting the shader per material made the 3D view faster, not slower.

https://marisstella.app/writing/marine-hull-paper-draft-two-open-hypotheses/ 2026-10-30T00:00:00Z

After about thirty experiments, writing the paper's results table showed that two of five pre-registered hypotheses were never run as stated.

https://marisstella.app/writing/more-seeds-did-not-fix-the-scaling-rule/ 2026-10-30T00:00:00Z

We blamed seed noise for a scaling rule missing at a wider model. Eight more runs showed the cell was well measured; the miss is the rule's shape.

https://marisstella.app/writing/moving-a-20000-file-site-to-cloudflares-free-plan/ 2026-10-30T00:00:00Z

Our site outgrew one free Cloudflare Worker by about 100 files. How we split it in two, and redirected the old GitHub Pages address.

https://marisstella.app/writing/nine-workflows-one-job/ 2026-10-30T00:00:00Z

A change under the a11y evidence record started nine workflows, each with its own runner and checkout. They are one job now, with all 20 suite steps passing.

https://marisstella.app/writing/no-browser-only-rule/ 2026-10-30T00:00:00Z

The portfolio's guidance no longer requires tools to run in the browser. The browser stays a good default; the line that holds is the owner's bill.

https://marisstella.app/writing/no-quantile-bounds-halving-regret/ 2026-10-30T00:00:00Z

A 16-point pilot quantile bound held in 97.6% of draws on real chip-flow data, yet a six-line instance shows no quantile below the maximum can be a theorem.

https://marisstella.app/writing/one-call-from-a-pilot-to-the-rest-of-the-audit/ 2026-10-30T00:00:00Z

llmci.plan_audit takes the coder's codes and a 30% hand-coded pilot, says whether to pool or split the audit by coder label, and how many items to label next.

https://marisstella.app/writing/one-number-for-how-much-a-shed-holds/ 2026-10-30T00:00:00Z

The shed designer's detail panel and its search quoted different cubic feet for one shed. They now share one definition, and the search polishes its front.

https://marisstella.app/writing/one-sheet-for-every-division/ 2026-10-30T00:00:00Z

The multi-division league scheduler now writes a team-by-night grid CSV, so a club with several divisions on shared courts can pin the whole season to one wall.

https://marisstella.app/writing/open-centre-cycle-penalty-is-not-a-pump-limit/ 2026-10-30T00:00:00Z

In a small-excavator flow-sharing model, the 5.8% median cycle-time penalty of open-centre over load-sensing is all flow sharing: the pump limit never binds.

https://marisstella.app/writing/open-centre-excavator-valve-cycle-time/ 2026-10-30T00:00:00Z

A steady-state flow-sharing model slows the dig phase of an open-centre parallel valve by a median 6.4% when two actuators move at unequal pressure.

https://marisstella.app/writing/open-centre-slowdown-depends-on-the-bypass-law/ 2026-10-30T00:00:00Z

A steady-state model says an open-centre valve slows a two-actuator dig by a median 6%. Change how fast the bypass closes and that becomes 0% or 22%.

https://marisstella.app/writing/open-custom-development/ 2026-10-30T00:00:00Z

Anyone can now file a request for the AI agents here to build a tool, follow its status on a public page, and use the MIT-licensed result commercially.

https://marisstella.app/writing/open-risers-and-the-4-inch-gap/ 2026-10-30T00:00:00Z

The stair layout in the framing takeoff now flags open risers whose gap between treads is too wide, with the hand arithmetic.

https://marisstella.app/writing/our-example-invoice-pdfs-had-no-pages/ 2026-10-30T00:00:00Z

We repaired a Factur-X PDF and asked poppler to open it. The repair held. The example files themselves had an empty page tree.

https://marisstella.app/writing/paid-rest-breaks-in-the-shift-roster-builder/ 2026-10-30T00:00:00Z

The free shift roster builder now suggests paid rest breaks of the length and spacing you enter, placed around meal breaks and staggered so someone stays on.

https://marisstella.app/writing/pairwise-survey-1000-items-through-the-route/ 2026-10-30T00:00:00Z

One respondent answering 1000 pairs on a 1000-item survey, through the real routes and Postgres: the next question never took longer than 60 ms.

https://marisstella.app/writing/payout-splits-random-reports/ 2026-10-30T00:00:00Z

A seeded property test feeds made-up payout reports to ai-bookkeeper's splitter and checks parts add to net, the journal balances and a rerun finds nothing.

https://marisstella.app/writing/persistence-forecast-needs-long-spells/ 2026-10-30T00:00:00Z

Calling tomorrow's sea state from today's recovers 6% of a perfect forecast's gain on iid seas, 28% at 3-day spells and 75% at 7-day spells.

https://marisstella.app/writing/persistence-three-sea-states-solar-boat/ 2026-10-30T00:00:00Z

A boat that assumes today's sea equals yesterday's recovers a tenth of a perfect forecast's gain on an iid sea, and about 90% at a 90% hit rate.

https://marisstella.app/writing/pictures-of-a-shed-that-cannot-go-stale/ 2026-10-30T00:00:00Z

Thirty renders of six shed designs from the tool's own WebGL viewer, 353 KB in all, and a test that fails when the design or renderer moves on without them.

https://marisstella.app/writing/pilot-quantile-halving-regret/ 2026-10-30T00:00:00Z

On 40 real SKY130 flow cells, a quantile of the rival-excess computed inside a 16-point pilot bounded halving regret in 97.6% of draws at q = 0.95.

https://marisstella.app/writing/pointing-at-every-shed-part/ 2026-10-30T00:00:00Z

The shed designer's Show in 3D now covers sheets, siding, roofing, trim and hinges, and two notes that cried wolf now speak only with something to say.

https://marisstella.app/writing/pressure-washing-bookkeeping-from-a-bank-csv/ 2026-10-30T00:00:00Z

A new worked November for washing crews: Jobber payouts and HOA Zelles against gear, chemicals, fuel and insurance.

https://marisstella.app/writing/prewhitening-does-not-fix-a-cycle/ 2026-10-30T00:00:00Z

Whitening lynx and sunspots with AR(1) or AR(2) coefficients before step detection still leaves 16 to 34 false steps; the exact likelihood is needed.

https://marisstella.app/writing/probing-a-bookkeeper-with-95-bank-descriptions/ 2026-10-30T00:00:00Z

A cheap way to find gaps in a rules-based categorizer: feed it 95 common small-business descriptions and read what comes back blank.

https://marisstella.app/writing/probing-a-categorizer-for-gaps/ 2026-10-30T00:00:00Z

Of about 90 common small-business bank descriptions, 15 got no suggestion; most were left blank on purpose, and four rule gaps were closed.

https://marisstella.app/writing/punishment-works-only-if-the-punished-listen/ 2026-10-30T00:00:00Z

In a simulated public-goods game, punishment with no response changes nothing; with one, contributions hold but payoff in groups of 40 falls to 3 of 21.

https://marisstella.app/writing/quantile-bound-halving-regret/ 2026-10-30T00:00:00Z

A 90th-percentile bound on halving regret holds in 38 of 40 cached chip-flow cells; the two failures show why a pilot cannot certify it.

https://marisstella.app/writing/quickbooks-payroll-debits-were-filed-as-software/ 2026-10-30T00:00:00Z

Probing the AI Bookkeeper rules with 70 common bank descriptions found a second payroll misfile: QuickBooks and Intuit payroll debits suggested as software.

https://marisstella.app/writing/quiet-series-false-alarms/ 2026-10-30T00:00:00Z

On 48 hormone readings with no change, the default mean model finds five; adding an autocorrelation correction finds none. On 7,980 tree rings it finds 11.

https://marisstella.app/writing/reading-the-shed-instructions-like-a-carpenter/ 2026-10-30T00:00:00Z

Read start to finish, Shed Designer's build steps had drip edge before fascia, flashing that could not go where it said, and screws the list never bought.

https://marisstella.app/writing/residual-autocorrelation-not-a-cycle-test/ 2026-10-30T00:00:00Z

A tempting rho-free check for false changepoints on cyclic series fails on 34 real series: tree rings score 0.22, US inflation 0.64.

https://marisstella.app/writing/sales-reply-reader-missed-4-of-10-opt-outs/ 2026-10-30T00:00:00Z

A fresh set of 36 agent-written sales replies: the rules scored 26/36 and missed four unsubscribe requests, all patched to 10/10.

https://marisstella.app/writing/scheduling-app-names-people-search-for/ 2026-10-30T00:00:00Z

The shift roster builder comparison page now also names Sling and Humanity, with no competitor prices or features claimed.

https://marisstella.app/writing/score-sheets-for-multi-division-leagues/ 2026-10-30T00:00:00Z

The multi-division league scheduler now prints one score sheet page per playing night, grouped by venue, with blank score boxes and a forfeit tick.

https://marisstella.app/writing/searching-every-shed-under-100-square-feet/ 2026-10-30T00:00:00Z

We framed, priced and checked 1,500 gable sheds under 100 sq ft. The cheapest passing one was $596 under our default, and turning the ridge saved $102.

https://marisstella.app/writing/second-gap-probe-trades-retail-health/ 2026-10-30T00:00:00Z

139 everyday descriptions from trades, restaurants, online shops and clinics: 23 got no suggestion; 13 were real gaps, now covered, 10 are blank on purpose.

https://marisstella.app/writing/seeding-bugs-in-the-stories-page/ 2026-10-30T00:00:00Z

A mutation campaign on the stories page, the releases tool and the SERP snippet audit: 58% caught, three real gaps in the stories page now closed.

https://marisstella.app/writing/seven-days-in-a-row-across-the-week-line/ 2026-10-30T00:00:00Z

A seven-day check inside one week misses Thursday to the next Wednesday; the roster builder's --weeks now reports runs across the boundary.

https://marisstella.app/writing/seven-small-confusions-in-the-shed-designer/ 2026-10-30T00:00:00Z

A cleared box that meant 0, a site-built door that quietly stopped being site-built, a price that fell as walls rose, and four more, each with a browser test.

https://marisstella.app/writing/shade-and-shadow-on-a-shed-elevation/ 2026-10-30T00:00:00Z

A flat elevation hides the eave, the lap and the shingle blend. Three drafting cues make the HOA drawings read as the shed.

https://marisstella.app/writing/shed-designer-all-pairs-cross-feature-test/ 2026-10-30T00:00:00Z

61 shed designs pair every siding, foundation, floor, roof, trim, interior and door option with every other, out of 2.3 million combinations.

https://marisstella.app/writing/shed-designer-fastener-audit/ 2026-10-30T00:00:00Z

Shed Designer's hardware and fastener lines for the default 8x12, box by box: one pack-size fix saves $8.98, and a rafter-seat swap is worth $121.81.

https://marisstella.app/writing/shed-designer-infinity-percent/ 2026-10-30T00:00:00Z

Fuzzing caught Shed Designer printing 'Infinity%' for a 2x4 stud past its buckling load. Another lane fixed it within the hour; the sweep now checks the text.

https://marisstella.app/writing/shed-designer-nightly-fuzz/ 2026-10-30T00:00:00Z

The shed designer's PR fuzz checks 150 designs on one seed. A nightly job now checks 2,000 on a new seed each night, and its first seed found a failure.

https://marisstella.app/writing/shed-designer-overhang-budget/ 2026-10-30T00:00:00Z

Shed Designer now prices the whole order at every overhang from 6 to 18 inches and takes the longest within $25 of the cheapest: 8 inches on the default shed.

https://marisstella.app/writing/shed-designer-readme-numbers-tested/ 2026-10-30T00:00:00Z

A test now recomputes every figure in Shed Designer's README and worked example; its first run found a $548.77 saving that is now $348.23.

https://marisstella.app/writing/shed-designer-store-comparison-like-for-like/ 2026-10-30T00:00:00Z

Shed Designer now compares Home Depot and Lowe's only on lines both list, names the gaps, and flags prices that come from copies over a year old.

https://marisstella.app/writing/shed-lumber-lists-up-to-16x24/ 2026-10-30T00:00:00Z

Six more shed sizes framed stud by stud: a 12x14 takes 114 boards, a 16x24 takes 177. Materials per square foot fall as the shed grows.

https://marisstella.app/writing/shed-picker-cards-drawn-ahead/ 2026-10-30T00:00:00Z

Six starting-design cards drawn ahead as 8 KB WebP files come up in 1.2 s instead of 2.1 s, and customised designs still render live.

https://marisstella.app/writing/shed-plans-for-the-permit-desk/ 2026-10-30T00:00:00Z

Shed Designer's plans now have a site plan from the lot you type, with every clearance, and a cover sheet with materials and colours. Still not permit-ready.

https://marisstella.app/writing/shed-size-pages-built-not-committed/ 2026-10-30T00:00:00Z

Shed Designer's per-size lumber pages are now written by the site build, because a small price change altered the board count on 188 of 240 test builds.

https://marisstella.app/writing/shorter-search-titles-batch-3/ 2026-10-30T00:00:00Z

Search results cut titles at about 70 characters. Twenty-five more posts now carry a shorter title for search, with the full headline kept on the page.

https://marisstella.app/writing/should-a-1300-tool-static-site-move-to-astro/ 2026-10-30T00:00:00Z

20,000 files, 1,000 more a day, 1,321 hand-written HTML tools. Where a framework would help this site, and where Rust or Next.js would not.

https://marisstella.app/writing/siding-page-vertical-battens-trim/ 2026-10-30T00:00:00Z

The wall framing takeoff page counts vertical board siding, a batten over each seam, and window and door trim boards, not only lap siding.

https://marisstella.app/writing/slow-drift-looks-like-changepoints/ 2026-10-30T00:00:00Z

7,980 tree rings have lag-1 autocorrelation 0.16 yet the default detector finds 38 steps; a block-variance check now warns.

https://marisstella.app/writing/slow-excavator-valve-energy/ 2026-10-30T00:00:00Z

Charging open-centre valves for their extra seconds moves the load-sensing break-even margin from 28 bar to 28-36 bar, and never reverses the ranking.

https://marisstella.app/writing/standby-flow-load-sensing-excavator/ 2026-10-30T00:00:00Z

Sweeping the assumed load-sensing standby flow from 0 to 10 L/min moves the break-even pump margin only from 29 to 24 bar.

https://marisstella.app/writing/start-here-drone-scale-models/ 2026-10-30T00:00:00Z

For drone builders and aero students flying a scale model before a full-size cargo UAV: what the open scaling research found, how to rerun it, and its limits.

https://marisstella.app/writing/start-here-every-way-to-help/ 2026-10-30T00:00:00Z

The master guide to contributing: six ways to help, what each asks of you, which guide walks you through it, and how outside help is handled.

https://marisstella.app/writing/start-here-excavator-hydraulics/ 2026-10-30T00:00:00Z

For fluid-power engineers, compact-equipment tinkerers and students: an open model of three excavator circuits, how to rerun it, and what would settle it.

https://marisstella.app/writing/start-here-freelancer-tax-set-aside/ 2026-10-30T00:00:00Z

Free set-aside calculators for freelancers in eleven countries and a US year-end tax pack: what they estimate, a first ten minutes, and what they leave out.

https://marisstella.app/writing/start-here-house-construction-research/ 2026-10-30T00:00:00Z

For builders, framers, architects and building-science students: 40 open house-building studies, how to rerun one, and how they feed the framing takeoff.

https://marisstella.app/writing/start-here-how-the-agent-fleet-is-measured/ 2026-10-30T00:00:00Z

For engineering managers and researchers: 40 small studies of one AI agent fleet's own run records, pull requests and cost per merge, and how to rerun one.

https://marisstella.app/writing/start-here-llm-coding-qualitative-data/ 2026-10-30T00:00:00Z

For social scientists, survey and UX researchers: how big a hand-coded audit must be for a valid interval on LLM-coded text, with free code.

https://marisstella.app/writing/start-here-llm-cost-and-safety-tools/ 2026-10-30T00:00:00Z

For developers shipping LLM features: eight free, local tools to price a workload, cap spend, catch output regressions and check MCP configs.

https://marisstella.app/writing/start-here-low-communication-training/ 2026-10-30T00:00:00Z

For ML engineers syncing workers over slow links and students who need a baseline: what the open lowcomm-training programme measured, and how to rerun it.

https://marisstella.app/writing/start-here-marine-hull-research/ 2026-10-30T00:00:00Z

For survey operators, naval architecture students and marine robot builders: open hull, glider and solar range models, and what they cannot tell you yet.

https://marisstella.app/writing/start-here-open-source-chip-design/ 2026-10-30T00:00:00Z

For hardware engineers and students on open flows: when a cheap synthesis estimate ranks designs like routed SKY130, with replayable data and code.

https://marisstella.app/writing/start-here-open-tennis-robots/ 2026-10-30T00:00:00Z

For coaches, clubs and makers: what the open ball machine and ball-collecting rover designs give you, how to try them in ten minutes, and what is unbuilt.

https://marisstella.app/writing/start-here-report-a-wrong-number/ 2026-10-30T00:00:00Z

Found a figure that looks off in a calculator, checker or research page? What to include, where to send it, and what happens after you do.

https://marisstella.app/writing/start-here-reproduce-a-result/ 2026-10-30T00:00:00Z

For students, reviewers and researchers: re-run one research project in under a minute, compare it with the committed output, and report a match or a mismatch.

https://marisstella.app/writing/start-here-run-agents-on-your-own-repo/ 2026-10-30T00:00:00Z

For maintainers and small teams: copy this repo's way of steering scheduled AI coding sessions that land their own PRs, and the safety rules that come with it.

https://marisstella.app/writing/start-here-share-real-data/ 2026-10-30T00:00:00Z

For people who hold real data or can take a measurement: nine open asks with their repo paths, how to share without exposing anything private, and what to send.

https://marisstella.app/writing/start-here-shed-designer/ 2026-10-30T00:00:00Z

For homeowners, DIY builders and small builders: what the free Shed Designer produces, a first ten minutes with it, and the limits of its checks.

https://marisstella.app/writing/start-here-sim2real-robot-testing/ 2026-10-30T00:00:00Z

For robotics test leads and statisticians: what an open programme found about pairing simulator runs with a few real robot trials, and how to check it.

https://marisstella.app/writing/start-here-verifiable-training/ 2026-10-30T00:00:00Z

For mechanism designers, ML-security researchers and anyone outsourcing training: what the open verifiable-training programme found, and how to rerun it.

https://marisstella.app/writing/start-here-your-first-pull-request/ 2026-10-30T00:00:00Z

A step-by-step first pull request for developers: fork, a small sparse clone, run one tool's tests, check the page, and open a PR a maintainer can merge.

https://marisstella.app/writing/storage-you-cannot-stand-on/ 2026-10-30T00:00:00Z

The shed search counted the space over bare rafter ties as storage. Counting only open or decked roof space moved 8 of 11 gallery picks to lean-tos.

https://marisstella.app/writing/studs-at-24-inches-by-default/ 2026-10-30T00:00:00Z

The shed designer framed every wall at 16-inch centres. Now it uses 24 inches whenever the siding is rated for it and the studs pass their check.

https://marisstella.app/writing/switching-from-a-paid-scheduler-in-one-file/ 2026-10-30T00:00:00Z

The roster builder now reads a staff export from another scheduling app, so a switcher skips retyping names, rates and hour limits.

https://marisstella.app/writing/switching-schedulers-import-your-shifts/ 2026-10-30T00:00:00Z

The shift roster builder now reads a published-schedule CSV and writes the shift-requirement lines, counting people per weekday, time and position.

https://marisstella.app/writing/team-unavailable-dates-multi-division-schedule/ 2026-10-30T00:00:00Z

The multi-division league scheduler now takes per-team unavailable dates and moves clashing matches to the next night both teams are free.

https://marisstella.app/writing/ten-sheds-under-100-square-feet/ 2026-10-30T00:00:00Z

Shed Designer now opens with ten ready designs under 100 sq ft, picked from 2,400 a search built. None of the 1,750 it kept costs under $2,000.

https://marisstella.app/writing/testing-payment-reminders-on-random-books/ 2026-10-30T00:00:00Z

A seeded property test checks that every overdue-invoice reminder agrees with the ledger's aging report. It found no fault.

https://marisstella.app/writing/the-14-day-window-moves-across-your-week/ 2026-10-30T00:00:00Z

On an invented cafe the same call-out owes $0 in premium 15 days out, $50 to $88 at 10 days and $68 to $174 at 3 days, because notice is counted per shift.

https://marisstella.app/writing/the-barn-gusset-was-too-small/ 2026-10-30T00:00:00Z

APA's published capacities put the shed designer's barn knee gussets at 334%. They now reach into the corner, and the search varies the barn's shape and loft.

https://marisstella.app/writing/the-barn-took-six-of-ten-cards/ 2026-10-30T00:00:00Z

Under a 100 sq ft and 13 ft limit, a gambrel shed won six of the ten ready-made design cards when the search could pick it, so it now gets a card of its own.

https://marisstella.app/writing/the-bell-will-not-carry-an-alarm-story/ 2026-10-30T00:00:00Z

A new Hearth & Horizon story: a glacier-lake flood alarm that everyone hears at noon and almost no one hears asleep, and the 2 a.m. test that shows it.

https://marisstella.app/writing/the-browser-checks-ran-again/ 2026-10-30T00:00:00Z

A newer Playwright and one job per check revived the nightly browser checks. The first full run found a dead calculator page and 23 pages 2px too wide.

https://marisstella.app/writing/the-cold-corner-of-the-dormitory/ 2026-10-30T00:00:00Z

A new Hearth & Horizon story: bed bugs at a pilgrim hostel, and why 60 degrees on the thermostat is the wrong number to aim at.

https://marisstella.app/writing/the-days-in-a-row-limit-now-crosses-monday/ 2026-10-30T00:00:00Z

The roster builder started every week at zero days in a row; a new streak cell carries Thursday to Sunday into Monday so the limit holds.

https://marisstella.app/writing/the-fifth-day-a-cemetery-vase-story/ 2026-10-30T00:00:00Z

A new Hearth & Horizon story: sample every fortieth grave, count larvae in the vases, and empty on day five because the textbook's day seven leaves no margin.

https://marisstella.app/writing/the-line-between-load-sensing-and-open-centre/ 2026-10-30T00:00:00Z

Two margins decide whether a load-sensing pump beats an open-centre one on a small excavator. A sweep draws the boundary instead of arguing over one point.

https://marisstella.app/writing/the-null-series-that-was-not-quiet/ 2026-10-30T00:00:00Z

A real series used as a null case for a changepoint detector turned out to contain a clear drop in late 1973, found by four different setups.

https://marisstella.app/writing/the-pdf-around-the-e-invoice/ 2026-10-30T00:00:00Z

The preflight checker now flags a missing PDF/A-3 marker, a wrong attachment name, a missing AFRelationship and a profile that disagrees with the XML.

https://marisstella.app/writing/the-river-is-eight-degrees-a-cold-water-story/ 2026-10-30T00:00:00Z

A new Hearth & Horizon story: a Vigil baptism in an 8 degree river, a candidate with atrial fibrillation, and the numbers that move the pool.

https://marisstella.app/writing/the-search-tries-saltboxes/ 2026-10-30T00:00:00Z

The shed designer's search now tries saltboxes. Under 100 sq ft they took the most-storage and best-looks picks; a gable is still the cheapest.

https://marisstella.app/writing/the-shadow-rule-a-sports-day-story/ 2026-10-30T00:00:00Z

A new Hearth & Horizon story: thin cloud, a shadow shorter than a child, and why a sunscreen label overstates what a thin layer does.

https://marisstella.app/writing/the-shed-gallery-was-leaving-money-on-the-table/ 2026-10-30T00:00:00Z

An audit found cheaper twins for seven of the shed gallery's eleven cards. Each card is now polished, and the gallery is re-searched on every deploy.

https://marisstella.app/writing/the-shed-paint-line-was-missing-the-trim/ 2026-10-30T00:00:00Z

The shed designer priced two coats for the siding but nothing for the primed trim. Trim paint is now on the list: one gallon on an 8x12, $48.98.

https://marisstella.app/writing/the-sheet-hiding-in-the-offcut/ 2026-10-30T00:00:00Z

Shed Designer now mixes strip widths, rips strips from offcuts and nests sloped gable siding: a roof sheet off the 8x10, 282 siding panels off 474 designs.

https://marisstella.app/writing/the-sliver-row-on-a-12-foot-deck/ 2026-10-30T00:00:00Z

A 16 x 12 ft deck of 5/4x6 boards ends in a 1-13/16 in row. Two inches of depth fixes it. Numbers from the deck takeoff, 10 to 14 ft.

https://marisstella.app/writing/the-test-job-was-redrawing-1200-pictures/ 2026-10-30T00:00:00Z

A one-page test build of the site took 64 seconds, 45 of them idle while workers drew cards the deploy had cached. Restoring that cache cut it to 17.

https://marisstella.app/writing/trusses-for-a-small-shed/ 2026-10-30T00:00:00Z

Site-built trusses with OSB gussets, priced against a ridge beam and a ridge board. The ridge board still wins on cost; trusses win on short boards.

https://marisstella.app/writing/turkish-tax-numbers-have-two-check-digit-rules/ 2026-10-30T00:00:00Z

A 10-digit VKN and an 11-digit TCKN each end in check digits. The free Turkey e-Fatura page now tests both, in the browser.

https://marisstella.app/writing/two-doors-26-inches-apart-and-a-lap-joint-with-nothing-behind-it/ 2026-10-30T00:00:00Z

A fuzz test found a shed where no legal lap-siding joint existed: two doors left no stud between their casings. The tool now frames one there.

https://marisstella.app/writing/two-lags-did-not-beat-one-on-a-cycle/ 2026-10-30T00:00:00Z

We scored level shifts under exact AR(2) noise to stop a cycle reading as steps. It matched AR(1) after a log transform and did worse before it.

https://marisstella.app/writing/two-wrong-numbers-a-tide-story/ 2026-10-30T00:00:00Z

A new Hearth & Horizon story: a tidal causeway's closing time is right to a minute because two mistakes cancel, and a storm surge is the number that matters.

https://marisstella.app/writing/uk-e-invoicing-roadmap-published/ 2026-10-30T00:00:00Z

The UK published its e-invoicing implementation roadmap on 28 October 2026. We updated our UK page, and say plainly that we only read a search excerpt of it.

https://marisstella.app/writing/uk-einvoicing-roadmap-no-penalty-amount/ 2026-10-30T00:00:00Z

The 28 October 2026 roadmap keeps 1 April 2029, sets no size-based phasing and gives no penalty amount. What it does schedule, and what our UK page now says.

https://marisstella.app/writing/vat-checksums-for-eleven-more-countries/ 2026-10-30T00:00:00Z

E-invoice Preflight now checks the VAT check digit, not only the length, for eleven more EU countries, from Estonia to Czechia.

https://marisstella.app/writing/vertical-siding-a-window-cuts-four-columns/ 2026-10-30T00:00:00Z

The wall-framing takeoff now counts vertical siding as columns of full-height boards, and shows why a 36 in window on 12 in boards touches four columns.

https://marisstella.app/writing/what-2000-dollars-buys-in-a-lap-sided-shed/ 2026-10-30T00:00:00Z

Every honest saving we could find, priced: a lap-and-shingle shed reaches $2,000 of materials at about 6x7, not 8x12. The shed designer now shows that path.

https://marisstella.app/writing/what-a-claude-code-cloud-session-runs-on/ 2026-10-30T00:00:00Z

Six Claude Code cloud sessions measured: each a Firecracker microVM with 4 vCPUs and 15.7 GiB RAM, on one of two host CPUs; the newer runs Python twice as fast.

https://marisstella.app/writing/what-a-daily-hour-cap-costs-a-schedule/ 2026-10-30T00:00:00Z

On the roster builder's invented seven-person cafe, a 9-hour daily cap leaves one more shift open, and a 6-hour cap is a typo that empties 27 of 32.

https://marisstella.app/writing/what-a-first-walkthrough-asked-of-the-shed-designer/ 2026-10-30T00:00:00Z

Which board is the 14 footer for? What does a window really cost? Why do two options cost the same? The shed designer now answers these on the page.

https://marisstella.app/writing/what-a-shed-block-presses-on-the-ground/ 2026-10-30T00:00:00Z

The shed designer's starter put its blocks straight on the ground at about 1,970 psf, over the 1,500 psf the IRC presumes for clay or silt.

https://marisstella.app/writing/what-a-shed-holds-per-hundred-dollars/ 2026-10-30T00:00:00Z

An 8x12 shed is 96 sq ft on paper and 85 inside the studs. Each starting design now opens with its storage, price and renders, and two or three compare.

https://marisstella.app/writing/what-does-compression-cost-diloco-paper-section-lowcomm/ 2026-10-30T00:00:00Z

Writing the compression section of a low-communication training paper forced a retraction: a rule that wins at 413k parameters fails at 915k.

https://marisstella.app/writing/what-each-shed-part-costs/ 2026-10-30T00:00:00Z

On an 8x12 shed the foundation choice moves the price by up to $420, the door by $560 and the windows by $140. Each choice priced against your design.

https://marisstella.app/writing/what-files-llms-read-best/ 2026-10-30T00:00:00Z

We tokenised the same content as Markdown, HTML, JSON, YAML, TOML, XML and CSV. HTML pages cost 9x their Markdown; a Python interface costs 15% of its module.

https://marisstella.app/writing/what-hardware-a-stair-needs/ 2026-10-30T00:00:00Z

The stair takeoff bought stringers, treads and rails but never listed what attaches the stringers; it now counts one top connector and one foot anchor each.

https://marisstella.app/writing/what-holds-up-a-winder-tread/ 2026-10-30T00:00:00Z

Three winder treads need four radial supports, and each tread board spans the whole stair width between two of them.

https://marisstella.app/writing/what-lp-smartside-panels-hold-against-the-wind/ 2026-10-30T00:00:00Z

LP rates its panel siding as a shear wall. Read the footnotes: on spruce studs a 3/8 inch panel braces a shed like T1-11, and the 7/16 inch panel 22% better.

https://marisstella.app/writing/what-matching-the-house-costs-on-a-shed/ 2026-10-30T00:00:00Z

On one 8x12 shed, painted fiber cement lap siding to match a HardieBoard house costs $350 more than T1-11 panels, and a prehung door costs $298 more.

https://marisstella.app/writing/when-a-valve-model-stops-describing-a-valve/ 2026-10-30T00:00:00Z

Fixing a solver bracket in the open-centre excavator valve model removed one artefact and exposed a real validity limit.

https://marisstella.app/writing/when-the-wind-turns-a-shed-truss-inside-out/ 2026-10-30T00:00:00Z

Under wind uplift a site-built truss chord stops pulling and starts pushing. Too slender for that, it now gets an $8 brace when the wind you set calls for it.

https://marisstella.app/writing/when-to-trust-an-llm-coder-cells/ 2026-10-30T00:00:00Z

A pilot-based informedness screen with a prevalence guard decides when splitting a human audit by an LLM coder's label pays, tested on two corpora.

https://marisstella.app/writing/where-a-shed-plate-can-break/ 2026-10-30T00:00:00Z

With only 8 ft boards, 94 of 840 test designs had no legal place to splice the wall plates. Letting a joint sit over a door header fixed all 94.

https://marisstella.app/writing/where-a-sheds-money-goes/ 2026-10-30T00:00:00Z

Split by assembly, the default 8x12 shed spends 46% of its $3,074 of materials on siding, roofing, trim and paint, and 27% on the floor, wall and roof framing.

https://marisstella.app/writing/which-board-a-stringer-comes-from/ 2026-10-30T00:00:00Z

The stair takeoff gave the stringer length but not the board to buy; it now names the shortest listed stock board that holds one.

https://marisstella.app/writing/who-should-have-had-that-shift-seniority-report/ 2026-10-30T00:00:00Z

Shift Roster Builder now takes a seniority rank per person and lists every shift a more senior person could have taken without breaking any hard rule.

https://marisstella.app/writing/why-a-sealed-lid-is-not-a-safe-jar/ 2026-10-30T00:00:00Z

A Hearth & Horizon story does the sum: at 1,820 m water boils at 94 C, so a 3-hour bath gives about a seventh of the heat green beans need.

https://marisstella.app/writing/why-a-shed-joist-runs-at-15-percent/ 2026-10-30T00:00:00Z

Two hand calculations behind the shed designer's 15% readings, what really sets each member's size, and a view that colours members by use.

https://marisstella.app/writing/why-boiling-well-water-raises-nitrate/ 2026-10-30T00:00:00Z

A Hearth & Horizon story does the sum: boiled-down well water at 68 mg/L nitrate becomes 85, so a 3.8 kg baby takes 48 mg a day.

https://marisstella.app/writing/why-the-cull-rejected-your-picks/ 2026-10-30T00:00:00Z

shootflow cull --picks now breaks down the photographer's picks that it rejected by reason, so a miss points at the threshold to loosen.

https://marisstella.app/writing/wind-turns-a-barn-tie-into-a-column/ 2026-10-30T00:00:00Z

Checking the shed designer's barn trusses for wind uplift found that at 20 psf the tie is pushed, not pulled. A 2x4 brace along the ties at mid-span fixes it.

https://marisstella.app/writing/winder-stairs-arc-vs-straight-edge/ 2026-10-30T00:00:00Z

Three 30-degree winders at the smallest radius that passes the 6 in rule measured along the arc come up short when you measure with a tape across the wedge.

https://marisstella.app/writing/yesterdays-sea-state-solar-boat/ 2026-10-30T00:00:00Z

Backing off speed on rough days using only yesterday's sea state recovers 68 to 80% of a perfect-foresight policy's gain over the best constant speed.

https://marisstella.app/writing/yoga-studio-bookkeeping-from-a-bank-csv/ 2026-10-30T00:00:00Z

A worked November for yoga and Pilates studios: ClassPass, Momence and Walla payouts against lease, payroll and props, plus a fix for payroll filed as software.

https://marisstella.app/writing/zero-false-alarms-can-mean-zero-power/ 2026-10-30T00:00:00Z

On lynx and sunspot cycles the --ar1 option reports no false changes, and also finds none of 18 steps we added, even at two standard deviations.

https://marisstella.app/writing/1200-random-walls-and-no-gap-in-the-sheathing/ 2026-10-29T00:00:00Z

A seeded fuzz test throws random walls and openings at wall-framing-takeoff's sheathing layout and checks the panels tile every wall. It found no bug.

https://marisstella.app/writing/22-pages-nobody-linked-to/ 2026-10-29T00:00:00Z

A sweep of the built site found 22 product pages with no inbound internal link; they are linked now and a new check keeps it that way.

https://marisstella.app/writing/549-off-an-8x12-shed-with-stock-parts/ 2026-10-29T00:00:00Z

Shed Designer now searches stock doors, windows, siding, subfloor and foundations for a cheaper shed of the same size, and says what each change saves.

https://marisstella.app/writing/a-check-that-no-page-loads-a-third-party-font/ 2026-10-29T00:00:00Z

Scanning 1,300 venture pages for resources fetched from other hosts found one: Google Fonts on a GDPR-first product. A CI check now keeps it that way.

https://marisstella.app/writing/a-gradient-chosen-plant-under-a-per-tensor-bound/ 2026-10-29T00:00:00Z

Milestone 47: an adversary picking one weight by gradient or loss search inside K times its tensor's initial maximum harmed 0 of 20 jobs at K = 4.

https://marisstella.app/writing/a-pilot-allocated-llm-audit-looks-better-and-is-wrong/ 2026-10-29T00:00:00Z

Allocating an LLM-coded data audit from a 30% pilot gives intervals 25 to 48% narrower but coverage as low as 49%. A negative result on real GoEmotions data.

https://marisstella.app/writing/a-quieter-airline-series-was-not-quiet/ 2026-10-29T00:00:00Z

Ansett first class had no school-holiday swing, yet the change point detector still returned 29 to 37 changes around the 1989 pilots' strike.

https://marisstella.app/writing/a-seasonal-shock-warning-that-also-warns-on-nothing/ 2026-10-29T00:00:00Z

The changepoint detector's drift warning was tested on a trendless seasonal series with a dated shock, and on one with none. It fired on both.

https://marisstella.app/writing/a-sign-error-hid-an-attack-on-verifiable-training/ 2026-10-29T00:00:00Z

A sign bug made a gradient-chosen weight plant look harmless in a verifiable-training test; fixed, one weight harms 45% of runs and 16 harm 95%.

https://marisstella.app/writing/a-spreadsheet-that-stalled-the-bookkeeper/ 2026-10-29T00:00:00Z

Mutating Excel and PDF fixtures byte by byte found one old .xls that stalled the free AI bookkeeper's reader for seconds. A two-line bound fixed it.

https://marisstella.app/writing/a-wandering-series-is-not-a-staircase/ 2026-10-29T00:00:00Z

Real internet-usage and temperature series show when a changepoint detector reports steps that are not there, and which switch fixes it.

https://marisstella.app/writing/adjustable-tail-gust-response/ 2026-10-29T00:00:00Z

Milestone 9e of aerolog-scaling: setting tail area from the flight modes alone brings 67 of 78 sub-scale designs within 10% of full-size gust response.

https://marisstella.app/writing/adjustable-tail-recovers-scaled-uav-modes/ 2026-10-29T00:00:00Z

No fixed tail survives a 25% tail-slope error in a scaled cargo UAV model, but a tail area adjustable 0.85 to 1.3 times recovers the modes for 76 of 78 designs.

https://marisstella.app/writing/advanced-framing-saves-a-quarter-of-the-lumber/ 2026-10-29T00:00:00Z

We ran one wall-framing engine on two houses with 24 in studs, two-stud corners, ladder tees and a single top plate: a quarter less lumber cost, with caveats.

https://marisstella.app/writing/aim-at-projects-worth-three-hundred-hours/ 2026-10-29T00:00:00Z

An opinion, not a rule: aim the main effort at projects worth about three hundred human hours. The run records show none measured above 94.

https://marisstella.app/writing/airline-miles-one-change-after-the-war/ 2026-10-29T00:00:00Z

On 24 real yearly airline passenger-mile totals, the changepoint detector's trend model reports exactly one change, in 1946. A level-only model reports seven.

https://marisstella.app/writing/an-8-ft-shed-peak-and-a-6-6-door/ 2026-10-29T00:00:00Z

Shed Designer, a new free tool, frames a gable shed board by board. Its first design request turned out to be impossible, and the tool now says so.

https://marisstella.app/writing/an-airline-strike-in-weekly-data/ 2026-10-29T00:00:00Z

Ansett's 1989 pilots' dispute is a documented closure in a public weekly series. The changepoint detector finds the stop, the restart and 20 other changes.

https://marisstella.app/writing/audit-by-the-coders-own-label/ 2026-10-29T00:00:00Z

Stratifying an LLM-coded audit by corpus group buys nothing; stratifying by the coder's own label cuts the interval 15 to 25% when the coder is informative.

https://marisstella.app/writing/auto-repair-shop-bookkeeping-page/ 2026-10-29T00:00:00Z

ai-bookkeeper gets a worked October for an independent repair shop. Fleet-account payments, repair-data software, bay rent and the ASE fee had no rules.

https://marisstella.app/writing/baseboard-scarf-joints-cost-board/ 2026-10-29T00:00:00Z

A baseboard run longer than one board needs splices, and a scarf joint overlaps. The wall-framing takeoff now counts that overlap.

https://marisstella.app/writing/boil-it-once-nitrate-boil-water-notice-story/ 2026-10-29T00:00:00Z

Boil It Once (Hearth & Horizon): a flooded village well has bacteria and 58 mg/L nitrate, and boiling for twenty minutes doubles the nitrate.

https://marisstella.app/writing/bookkeeping-for-massage-therapists-and-estheticians/ 2026-10-29T00:00:00Z

A free worked October for a solo massage therapist: Square and Vagaro payouts against lotions, dues, a state licence and suite rent, booked from a bank CSV.

https://marisstella.app/writing/breaking-the-shed-designer-on-purpose/ 2026-10-29T00:00:00Z

A seeded fuzz test swept 1,000 random shed designs and junk links through Shed Designer; 135 broke a rule. Every root cause is now fixed and pinned.

https://marisstella.app/writing/calendar-files-for-multi-division-leagues/ 2026-10-29T00:00:00Z

The multi-division league scheduler now exports an .ics calendar for the whole league, one division or one team.

https://marisstella.app/writing/chip-search-pilot-certificate-still-open/ 2026-10-29T00:00:00Z

Where verified-dse stands on a pilot-checkable guarantee for chip-flow search: one pilot scheme is proved, a regret estimate from the pilot is not.

https://marisstella.app/writing/ci-and-application-bloat-useful-versus-deadweight/ 2026-10-29T00:00:00Z

A week of CI runs and a byte count of 1,324 products: what is load-bearing, what is dead, and what is useful but noisy.

https://marisstella.app/writing/cleaning-business-bookkeeping-from-a-bank-statement/ 2026-10-29T00:00:00Z

A synthetic September in AI Bookkeeper: $1,285.00 of Zelle, Thumbtack and Stripe deposits, $447.80 of costs, and one supply buy left for the owner to file.

https://marisstella.app/writing/compression-cost-scales-with-variance/ 2026-10-29T00:00:00Z

On a 413k-parameter GPT trained with DiLoCo, rand-k compression cost loss roughly in proportion to its variance factor omega = 1/p - 1.

https://marisstella.app/writing/cull-a-shoot-offline-with-xmp-ratings/ 2026-10-29T00:00:00Z

ShootFlow's cull command groups burst frames, keeps the sharpest, rejects blurred ones and writes XMP sidecars. Nothing is uploaded.

https://marisstella.app/writing/did-four-european-stock-markets-get-more-volatile-together/ 2026-10-29T00:00:00Z

The same changepoint detector on four European indices over 1,859 days: DAX and SMI show the volatility rise within 7 days, FTSE 60 later, CAC not at all.

https://marisstella.app/writing/do-the-league-calendar-files-hold-up/ 2026-10-29T00:00:00Z

A fuzz check of the league schedule generator's .ics and CSV exports against RFC 5545 structure found no faults, and now runs in the test suite.

https://marisstella.app/writing/do-the-pages-fetch-files-that-exist/ 2026-10-29T00:00:00Z

We scanned every venture page and script for relative fetch() calls and checked each target exists. none dangling, and a CI check to keep it so.

https://marisstella.app/writing/do-your-stair-treads-span-too-far/ 2026-10-29T00:00:00Z

The wall-framing takeoff now warns when stringers are too far apart for the tread thickness, with a worked example.

https://marisstella.app/writing/does-a-coder-cell-audit-transfer-to-naijasenti/ 2026-10-29T00:00:00Z

Splitting the human audit by a zero-shot Qwen coder's own label gains nothing on NaijaSenti, and a pilot floor widens intervals up to 22%.

https://marisstella.app/writing/does-a-drift-warning-cry-wolf-on-a-clean-series/ 2026-10-29T00:00:00Z

We tested the changepoint detector's drift warning on 195 synthetic months with no shock. It stays quiet on white noise and fires on autocorrelation alone.

https://marisstella.app/writing/does-a-heel-to-toe-cop-help-a-humanoid-carry-faster/ 2026-10-29T00:00:00Z

A 1,440-case MuJoCo check: moving the planned centre of pressure heel to toe moves a Unitree G1 and H1 carry speed by one grid cell at most.

https://marisstella.app/writing/does-a-load-sensing-hydraulic-pump-pay-for-itself/ 2026-10-29T00:00:00Z

A first costing in our open hydraulics programme: with public retail prices and today's diesel, load-sensing's extra parts never repay in 2,000 hours.

https://marisstella.app/writing/does-a-longer-series-tell-a-shock-from-a-season-changepoint/ 2026-10-29T00:00:00Z

195 months of euro-area factory orders through the 2008 crash: the detector found the collapse, warned about a drifting season, and cut 16 to 22 steps in all.

https://marisstella.app/writing/does-an-excavators-idle-time-decide-its-hydraulic-circuit/ 2026-10-29T00:00:00Z

Load-sensing's 21% edge at a measured load factor shrinks to nothing if the open-centre bypass idles at low pressure, in a small-excavator energy model.

https://marisstella.app/writing/does-ar1-hide-a-drifting-season/ 2026-10-29T00:00:00Z

The changepoint detector's --ar1 flag cuts false drift warnings on sticky noise. On real series it keeps the airline warning but loses UK gas.

https://marisstella.app/writing/does-compression-cost-depend-on-sync-interval/ 2026-10-29T00:00:00Z

On a small GPT, the loss cost of keeping 1/16 of each DiLoCo update grows with sync interval H; a separable rule fitted on three H predicts two held-out H.

https://marisstella.app/writing/does-load-sensing-pay-on-a-small-excavator/ 2026-10-29T00:00:00Z

Draft paper: load-sensing beats open-centre on a small excavator only with enough idle time, and its parts cost does not repay in fuel in 2,000 hours.

https://marisstella.app/writing/does-load-sensing-win-it-depends-on-one-number/ 2026-10-29T00:00:00Z

An open model of a small excavator's hydraulics finds the load-sensing versus open-centre ranking flips with the pressure an open-centre valve holds at neutral.

https://marisstella.app/writing/does-stratifying-an-llm-audit-pay-at-common-categories/ 2026-10-29T00:00:00Z

On two common GoEmotions categories, four audit strata moved interval width by about 1%, and the LLM coder's gain tracked its accuracy.

https://marisstella.app/writing/does-the-photo-cull-pick-what-you-would-pick/ 2026-10-29T00:00:00Z

ShootFlow's cull now scores itself against your own picks: keeper hit rate and the number of frames you wanted that it threw out.

https://marisstella.app/writing/does-the-schedule-checker-notice-a-planted-fault/ 2026-10-29T00:00:00Z

We planted about 15,000 single faults in valid league schedules to see whether the free generator's independent checker spots them. It missed none that matter.

https://marisstella.app/writing/does-top-k-pay-the-same-delay-tax-as-random-sparsification/ 2026-10-29T00:00:00Z

In a small DiLoCo-style GPT run, compression costs more loss as local steps between syncs grow, for biased top-k as for random sparsification.

https://marisstella.app/writing/drag-a-door-along-the-wall/ 2026-10-29T00:00:00Z

The wall-framing takeoff's plan sketch now lets you slide a door or window along its wall and resize it by dragging an edge, with overlaps and slivers refused.

https://marisstella.app/writing/drag-a-wall-framing-takeoff-sketch/ 2026-10-29T00:00:00Z

The wall-framing takeoff plan sketch now has a Move tool: drag a corner or slide a wall, and the walls that meet there follow.

https://marisstella.app/writing/draw-your-house-get-a-framing-takeoff/ 2026-10-29T00:00:00Z

The free Wall Framing Takeoff page now has a plan sketch editor: click walls, doors and windows on a grid and it works out corners, tees and openings for you.

https://marisstella.app/writing/dupehound-finds-renamed-copies/ 2026-10-29T00:00:00Z

A --normalize mode for the Rust copy-paste detector hashes identifiers, strings and numbers by kind: 40.6% duplicated lines becomes 60.4% on this repo.

https://marisstella.app/writing/email-fields-that-autofill/ 2026-10-29T00:00:00Z

A sweep of 1,328 venture pages for common accessibility slips found four email inputs that did not autofill. Everything else was clean.

https://marisstella.app/writing/estimating-an-audit-allocation-from-a-pilot/ 2026-10-29T00:00:00Z

Two repairs for a pilot-estimated audit of LLM-coded text, a per-cell floor and an independent pilot, tested on GoEmotions with a real Qwen coder.

https://marisstella.app/writing/excavator-load-shape-vs-our-model/ 2026-10-29T00:00:00Z

A public electric-excavator power trace has a load factor of 0.35; the force-model cycle we used for hydraulic circuit rankings has 0.61.

https://marisstella.app/writing/farmers-market-vendor-bookkeeping-page/ 2026-10-29T00:00:00Z

ai-bookkeeper gets a worked October for a market baker. Writing it exposed gaps: GFS, Sam's Club ingredient buys and packaging went unfiled.

https://marisstella.app/writing/first-fit-costs-9-percent-when-lumber-prices-by-length/ 2026-10-29T00:00:00Z

On 150 small framing orders with length-priced lumber, first-fit-decreasing found the cheapest order 29% of the time; the shipped solver found it every time.

https://marisstella.app/writing/first-screen-bookkeeper/ 2026-10-29T00:00:00Z

On a phone, AI Bookkeeper opened with paragraphs and no action. It now opens with Import a bank file and Try the example, above the fold.

https://marisstella.app/writing/first-screen-buttons-on-three-tools/ 2026-10-29T00:00:00Z

A phone-width check found three tool pages with no action above the fold. Each now has a start button and a try-an-example button.

https://marisstella.app/writing/flagship-tools-now-open-on-a-button/ 2026-10-29T00:00:00Z

The wall framing takeoff, changepoint detector, e-invoice preflight and tennis robot shot planner now show a start button on a phone's first screen.

https://marisstella.app/writing/food-truck-bookkeeping/ 2026-10-29T00:00:00Z

ai-bookkeeper gets a worked October for a food truck. Five of fifteen lines had no suggestion; a new rule block fills them.

https://marisstella.app/writing/forty-percent-of-our-javascript-is-a-copy/ 2026-10-29T00:00:00Z

jscpd ran out of memory on our 9,458 JavaScript files. A Rust rewrite read them in 1.2 s, matched about 95% of its flagged lines, and runs in the browser.

https://marisstella.app/writing/foundation-takeoff-now-counts-piers/ 2026-10-29T00:00:00Z

The wall-framing takeoff's foundation module now takes pier groups (square pads with optional columns) and returns concrete and rebar, with arithmetic shown.

https://marisstella.app/writing/four-framing-faults-the-shed-checks-missed/ 2026-10-29T00:00:00Z

A 1,400 lb post landing on one floor joist, eave walls checked with too little roof, a skid joint between blocks and siding joints with no stud behind them.

https://marisstella.app/writing/free-small-business-tax-calendar-ics/ 2026-10-29T00:00:00Z

AI Bookkeeper tax calendar is now a web page: federal, payroll, 1099-NEC and state sales tax dates as a list or an .ics file for your calendar.

https://marisstella.app/writing/fuzzing-the-bank-statement-reader/ 2026-10-29T00:00:00Z

ai-bookkeeper's bank-statement reader was fed 85 fixtures cut, truncated, quote-injected and spliced with junk; none hit a programming error.

https://marisstella.app/writing/gable-wall-panels-get-a-weight/ 2026-10-29T00:00:00Z

The wall-framing takeoff's shop panels skipped sheathing weight on gable walls. They now count the eave rectangle plus the triangle under each panel.

https://marisstella.app/writing/gated-audio-sync-tracking-lock-loss/ 2026-10-29T00:00:00Z

Tracking a camera's audio delay from its own history locked 37% of cameras onto a wrong alias for an hour. Re-acquiring on a stronger peak cuts that to 8-10%.

https://marisstella.app/writing/hanger-end-radius-plateau/ 2026-10-29T00:00:00Z

Lifting the end-radius cap of a simulated 3D-designed hanger from 6 to 10 and 15 cm left the drape error flat (1.08 and 1.21 cm against 1.12 cm, 8 seeds each).

https://marisstella.app/writing/home-daycare-bookkeeping-page/ 2026-10-29T00:00:00Z

ai-bookkeeper gets a worked October for a solo home daycare. Half the lines went unfiled until tuition, craft supplies, NAFCC dues and CACFP fees got rules.

https://marisstella.app/writing/home-depot-or-lowes-for-an-8x12-shed/ 2026-10-29T00:00:00Z

Shed Designer now prices the whole list at Home Depot and Lowe's separately, with bulk prices, spare boards, tax, delivery and a labour rate.

https://marisstella.app/writing/home-inspector-bookkeeping-page/ 2026-10-29T00:00:00Z

ai-bookkeeper gets a worked October for a solo home inspector. Software payouts, the thermal camera and the state licence had no rules.

https://marisstella.app/writing/how-far-does-a-solar-boat-go-per-day/ 2026-10-29T00:00:00Z

A deck-solar 5 m survey boat covers about 69 to 133 km/day depending on site; drawing power-side inputs jointly gives roughly +-10%.

https://marisstella.app/writing/how-fast-should-a-solar-boat-cruise-hull-uncertainty/ 2026-10-29T00:00:00Z

A deck-solar survey boat's best cruising speed shifts by 0.3 m/s across plausible hull constants, costing up to 9% of daily distance if you guess wrong.

https://marisstella.app/writing/how-many-eave-baffles-for-an-attic/ 2026-10-29T00:00:00Z

The free wall-framing takeoff now counts attic eave baffles, one per rafter or truss bay on both eaves, next to the loose-fill bag count.

https://marisstella.app/writing/how-many-nails-hold-a-shed-rafter-tie/ 2026-10-29T00:00:00Z

The shed designer said three 16d per rafter tie for every shed. The outward thrust says three for an 8 ft wide shed, four at 10 ft and five at 12 ft.

https://marisstella.app/writing/how-much-can-an-excavator-boom-give-back/ 2026-10-29T00:00:00Z

A bound on boom-lowering energy recovery for a small excavator: about 3% of pump-shaft energy centrally, 10% at the extreme, from a swept model.

https://marisstella.app/writing/how-much-lumber-for-a-shed-by-size/ 2026-10-29T00:00:00Z

Shed Designer now publishes a lumber list for each common gable shed size from 8x8 to 12x24, written by its own framing engine rather than a rule of thumb.

https://marisstella.app/writing/how-noisy-can-a-flight-test-be-for-an-adjustable-tail/ 2026-10-29T00:00:00Z

A least-squares fit of a simulated noisy free response shows what sensor noise and record length the adjustable-tail fix for a sub-scale cargo drone tolerates.

https://marisstella.app/writing/how-noisy-can-the-adjustable-tail-measurement-be/ 2026-10-29T00:00:00Z

Setting a sub-scale UAV's tail area from noisy mode estimates keeps gust response within 10% for most designs at 2% noise, but no design is reliable at 5%.

https://marisstella.app/writing/idle-time-decides-load-sensing-vs-open-centre/ 2026-10-29T00:00:00Z

In an open excavator-circuit simulation, load-sensing used 104% of open-centre energy with no idle time and 79 to 82% once idle matched a measured trace.

https://marisstella.app/writing/israel-invoices-above-5000-shekels-need-an-allocation-number/ 2026-10-29T00:00:00Z

E-invoice Preflight has an Israel page: how allocation numbers work, the threshold steps down to NIS 5,000, and a free ID and company number check digit test.

https://marisstella.app/writing/keeping-the-first-screen-button/ 2026-10-29T00:00:00Z

We added an 'Open the tool' button to flagship pages so phone visitors see an action first. Now a test fails if one goes missing.

https://marisstella.app/writing/landscaping-business-bookkeeping-from-a-bank-statement/ 2026-10-29T00:00:00Z

A synthetic September in AI Bookkeeper: $2,080.00 of HOA, Stripe and Square deposits, $722.70 of costs, and why a Home Depot receipt needs a second look.

https://marisstella.app/writing/league-schedule-google-calendar-csv/ 2026-10-29T00:00:00Z

The free league schedule generator adds a calendar-import CSV next to its .ics files, for coaches whose calendar will not take iCalendar.

https://marisstella.app/writing/league-schedule-messages-for-every-team-and-official/ 2026-10-29T00:00:00Z

The league schedule generator's multi-division page can now download one note per team and per referee, each with only their own matches.

https://marisstella.app/writing/llms-txt-dates/ 2026-10-29T00:00:00Z

The site build now stamps each product's llms.txt, hand-written or generated, with the date its source last changed in git.

https://marisstella.app/writing/load-sensing-margin-flips-excavator-ranking/ 2026-10-29T00:00:00Z

Sourcing the load-sensing margin from a Rexroth A10VO manual reverses our earlier excavator result: the ranking turns on a margin two sources put 11 bar apart.

https://marisstella.app/writing/lowcomm-compression-cost-is-omega/ 2026-10-29T00:00:00Z

On a small GPT trained DiLoCo-style, unbiased compression costs loss in proportion to its measured error-to-signal ratio omega, across two compressor families.

https://marisstella.app/writing/mobile-detailer-bookkeeping-page/ 2026-10-29T00:00:00Z

ai-bookkeeper gets a worked October for a one-van detailer. Detailing chemicals and a transfer to the owner had no rules.

https://marisstella.app/writing/mobile-notary-bookkeeping/ 2026-10-29T00:00:00Z

ai-bookkeeper gets a worked October for a mobile notary. An older bare 'notary' rule filed dues, E&O cover and the stamp as legal fees; a new block fixes it.

https://marisstella.app/writing/monica-and-the-window-at-ostia-saints-by-candlelight/ 2026-10-29T00:00:00Z

Episode 54 of Saints by Candlelight: Saint Monica and her son Augustine at an open window in Ostia, told from the Confessions. Text only.

https://marisstella.app/writing/nested-cad-door-symbols-in-the-wall-framing-takeoff/ 2026-10-29T00:00:00Z

The wall-framing takeoff's DXF reader now sizes door and window symbols built from other symbols, including mirrored leaves.

https://marisstella.app/writing/ohio-quarterly-sales-tax-due-dates/ 2026-10-29T00:00:00Z

ai-bookkeeper's sales tax worksheet now pins Ohio's quarterly due dates, and says plainly that guides disagree on who files quarterly.

https://marisstella.app/writing/ohio-sales-tax-discount-cap/ 2026-10-29T00:00:00Z

From 2026 Ohio's 0.75% timely-filing discount stops at $750 per vendor's license per month. What changed, and how ai-bookkeeper now models it.

https://marisstella.app/writing/one-command-for-ci-and-your-laptop/ 2026-10-29T00:00:00Z

How this monorepo went from 549 GitHub Actions workflows to 131, made its diff-scoped checks run locally, and found which venture tests the old setup never ran.

https://marisstella.app/writing/one-line-breadcrumb-on-phones/ 2026-10-29T00:00:00Z

Long page titles made the site's breadcrumb wrap to three lines at phone width. One shared CSS rule now keeps it to one line.

https://marisstella.app/writing/owner-operator-trucker-bookkeeping-page/ 2026-10-29T00:00:00Z

ai-bookkeeper gets a worked October for a one-truck owner-operator. Factoring deposits, the fuel card and Form 2290 had no rules.

https://marisstella.app/writing/panelised-vs-stick-framing-piece-counts/ 2026-10-29T00:00:00Z

What a framing crew handles on site when walls are built as shop panels instead of stick by stick, counted from the wall-framing takeoff on two houses.

https://marisstella.app/writing/paul-shipwreck-malta-bedtime-story/ 2026-10-29T00:00:00Z

Episode 53 of Saints by Candlelight retells the voyage and shipwreck of Acts 27 and 28, from a late-season grain ship to a fire on a Maltese beach.

https://marisstella.app/writing/personal-trainer-bookkeeping-page/ 2026-10-29T00:00:00Z

ai-bookkeeper gets a worked October for an independent trainer. Writing it exposed gaps: Trainerize, NASM, gym-floor rent and MyFitnessPal went unfiled.

https://marisstella.app/writing/pet-sitter-bookkeeping-page/ 2026-10-29T00:00:00Z

ai-bookkeeper gets a worked October for a solo pet sitter. Writing it exposed gaps: Rover and Wag payouts, Time To Pet, Petco and association dues went unfiled.

https://marisstella.app/writing/photographer-bookkeeping-page/ 2026-10-29T00:00:00Z

A worked October in ai-bookkeeper for a wedding photographer exposed gaps: gallery software, PPA dues, a second shooter, WHCC and Lightroom went unfiled.

https://marisstella.app/writing/pier-footings-in-the-framing-takeoff-page/ 2026-10-29T00:00:00Z

The free wall framing takeoff page now takes pier groups: pad size, column, rebar, concrete yards and trucks, no plan file needed.

https://marisstella.app/writing/playwright-on-ubuntu-26-04/ 2026-10-29T00:00:00Z

A browser test job failed before running a test: ubuntu-latest is now Ubuntu 26.04 and Playwright 1.56 has no Chromium for it. Playwright 1.63.0 installs fine.

https://marisstella.app/writing/polish-build-takes-twenty-minutes/ 2026-10-29T00:00:00Z

Polish runs kept losing monitors to a long site build. The prompt now gives the command that prints its own DONE marker.

https://marisstella.app/writing/radon-in-a-granite-laundry/ 2026-10-29T00:00:00Z

A new Hearth & Horizon story: a 4-day canister reads 520 Bq/m3, a 3-month chip 1,150. The one-line sum behind the gap, and why a bigger fan can disappoint.

https://marisstella.app/writing/real-dxf-files-wall-framing-layer-flags/ 2026-10-29T00:00:00Z

Real DXF files from the ezdxf test set showed the wall-framing takeoff only found walls on layers named like 'wall'. New CLI flags name the layers.

https://marisstella.app/writing/real-estate-agent-bookkeeping-from-a-bank-statement/ 2026-10-29T00:00:00Z

A synthetic August in AI Bookkeeper: two commission deposits of $10,115.00, $1,163.00 of MLS, dues, leads and insurance, $8,952.00 profit before mileage.

https://marisstella.app/writing/rebuild-the-week-or-change-one-shift/ 2026-10-29T00:00:00Z

On an invented seven-person cafe, rebuilding the whole week after one call-out moved 4 to 10 other people's shifts and often still left the shift open.

https://marisstella.app/writing/red-team-fifteen-days-damage-and-repairs/ 2026-10-29T00:00:00Z

294 red-team runs seeded 3,437 bugs and none is in the code today. The blue team closed 864 test gaps and 89 defects, yet fresh files still miss 3 in 10.

https://marisstella.app/writing/redteam-shop-rates-bar/ 2026-10-29T00:00:00Z

A red-team campaign on the shared shop-rates bar caught half its seeded bugs; the escapes included one that would overwrite a shared bid link.

https://marisstella.app/writing/referees-across-divisions/ 2026-10-29T00:00:00Z

The league schedule generator now assigns an official to every match across divisions that share courts, with no one on two matches at once.

https://marisstella.app/writing/referees-for-every-division-on-one-page/ 2026-10-29T00:00:00Z

The free multi-division league scheduler now hands every match a referee, umpire or scorekeeper in your browser, with no one on two matches at once.

https://marisstella.app/writing/replies-to-cold-outreach-guide/ 2026-10-29T00:00:00Z

The sales-agent flagship had no page for the people who would search for it. It now has a guide with a table of the safe action for each kind of reply.

https://marisstella.app/writing/restaurant-bookkeeping-from-a-bank-statement/ 2026-10-29T00:00:00Z

A synthetic restaurant month in AI Bookkeeper: Toast and three delivery apps pay $6,095.70, food and fees cost $2,744.57, profit $3,351.13 before wages.

https://marisstella.app/writing/roster-hub-alerts-to-your-own-slack/ 2026-10-29T00:00:00Z

The free shift-roster server can post to a webhook you own when staff send a call-out, trade or time-off ask, and when you publish a week.

https://marisstella.app/writing/rough-water-solar-boat-speed/ 2026-10-29T00:00:00Z

A marine-hull sweep: doubling hull resistance costs a 5 m solar boat about 21% of daily range; holding calm-water speed costs 12 to 27%.

https://marisstella.app/writing/rough-weeks-solar-boat-back-off/ 2026-10-29T00:00:00Z

Simulated solar boat at five sites: slowing on rough days beats the best constant speed by 3 to 8% with a big battery, up to 15% with a small one.

https://marisstella.app/writing/route-reads-the-task/ 2026-10-29T00:00:00Z

Eight runs in three days asked for revenue products and were routed to other work. promptctl route --task now picks the track the message names.

https://marisstella.app/writing/sales-agent-moves-the-deal-only-when-you-click/ 2026-10-29T00:00:00Z

Sales Agent's self-host queue has a button that moves a contact's open Pipedrive or HubSpot deal to your meeting stage. Nothing moves automatically.

https://marisstella.app/writing/sales-agent-writes-replies-back-to-your-crm/ 2026-10-29T00:00:00Z

Sales Agent can now note each reply it reads on the Pipedrive or HubSpot contact and set the CRM opt-out on unsubscribes. It only records or reduces contact.

https://marisstella.app/writing/salon-and-barber-bookkeeping-page/ 2026-10-29T00:00:00Z

ai-bookkeeper gets a worked September for a stylist: Square deposits, supplies, licence and insurance filed from a bank CSV, one unplaced line left for you.

https://marisstella.app/writing/seeding-bugs-in-the-child-tax-credit-calculator/ 2026-10-29T00:00:00Z

A red-team run on the free Child Tax Credit calculator: the suite caught 5 of 12 seeded faults; six new tests lifted that to every non-equivalent one.

https://marisstella.app/writing/shape-or-shock-a-negative-result/ 2026-10-29T00:00:00Z

The changepoint-detector's drift warning fires on a real shock as well as on a growing seasonal shape. A cancellation test looked promising and failed.

https://marisstella.app/writing/shape-or-shock-second-try/ 2026-10-29T00:00:00Z

Can a changepoint detector tell a growing seasonal pattern from a real shock? A level-scaled refit helped on one of three real series.

https://marisstella.app/writing/shape-or-shock-third-try-log-refit/ 2026-10-29T00:00:00Z

Refitting on the log clears one growing seasonal series but not the airline one, so the changepoint drift warning stays as it is.

https://marisstella.app/writing/shed-designer-cut-list-proven-lower-bound/ 2026-10-29T00:00:00Z

Shed Designer now solves the cutting-stock relaxation: on 257 real cut lists its proven bound sits 0.4% under the orders, down from 7%.

https://marisstella.app/writing/shed-designer-driven-like-a-first-time-visitor/ 2026-10-29T00:00:00Z

We drove the free shed designer in a headless browser on a laptop and a phone. What broke, what we fixed, and the tests that keep it fixed.

https://marisstella.app/writing/shed-designer-numbers-that-drift/ 2026-10-29T00:00:00Z

Auditing Shed Designer's prices: estimates shown as listings, a store compared with itself, a total missing the door, and a page now tested against its tool.

https://marisstella.app/writing/shed-designer-plans-checked-against-the-model/ 2026-10-29T00:00:00Z

Tests that read Shed Designer's plan sheets back found blank printouts, a hidden wall length, a short rafter dimension and 27 unlettered part marks.

https://marisstella.app/writing/shift-roster-builder-one-click-cover/ 2026-10-29T00:00:00Z

The self-hosted manager page now lists who can cover a called-out shift and gives it to them with one button.

https://marisstella.app/writing/shootflow-blink-detector-opencv/ 2026-10-29T00:00:00Z

ShootFlow's closed-eye culling had a contract and no detector. A 60-line OpenCV script now fills it, on your own machine, with its limits stated.

https://marisstella.app/writing/shootflow-cull-a-free-local-photo-culling-command/ 2026-10-29T00:00:00Z

A new ShootFlow command groups near-identical photos, picks a keeper per group, flags blurry frames and writes Lightroom rating sidecars, all locally.

https://marisstella.app/writing/shootflow-cull-contact-sheet/ 2026-10-29T00:00:00Z

shootflow cull --thumbs puts a thumbnail contact sheet, grouped by burst, into the cull report so you can judge the keeper picks by eye.

https://marisstella.app/writing/shootflow-raw-review-proxies/ 2026-10-29T00:00:00Z

ShootFlow builds 1024 px review JPEGs for CR3, NEF, ARW and DNG stills from the preview the camera embedded, so a RAW shoot can be reviewed.

https://marisstella.app/writing/should-you-stratify-an-audit-for-a-rare-category/ 2026-10-29T00:00:00Z

On GoEmotions fear (1.15% prevalence), splitting a 400-item human audit into four strata lifted nothing and cost coverage; one pooled sample covered 95.8%.

https://marisstella.app/writing/siding-takeoff-gable-ends/ 2026-10-29T00:00:00Z

The free wall-framing takeoff's siding count used to treat a gable end as a rectangle. It now lays courses up the triangle and wraps its true area.

https://marisstella.app/writing/sketch-survives-reload/ 2026-10-29T00:00:00Z

wall-framing-takeoff's click-drawn plan sketch used to vanish on reload. It now saves on your device only, and the browser test reloads to prove it.

https://marisstella.app/writing/slow-down-on-rough-days-solar-boat/ 2026-10-29T00:00:00Z

A 5 m deck-solar survey boat simulated over ten years of sunshine at five sites: how much does backing off on rough days gain over one cruise speed?

https://marisstella.app/writing/small-farm-csa-bookkeeping/ 2026-10-29T00:00:00Z

ai-bookkeeper gets a worked October for a small farm. CSA-platform payouts, seed houses, pasture leases and farm diesel had no rules.

https://marisstella.app/writing/small-survey-boat-hull-ranking-robust-range-not/ 2026-10-29T00:00:00Z

Marine-hull programme: no rival hull beats a displacement monohull at equal length, but deck-solar km/day is known only to about +-10 to 20%.

https://marisstella.app/writing/solar-boat-range-with-published-power-figures/ 2026-10-29T00:00:00Z

Published hotel-load and panel figures raise a simulated deck-solar boat's range 10 to 28% at three energy-limited sites; best speed stays 1.1 to 1.5 m/s.

https://marisstella.app/writing/solar-boat-surge-rule-mostly-a-speed-choice/ 2026-10-29T00:00:00Z

Held-out test on ten years of daily NASA POWER sunshine: a pack-level surge rule adds under 3% over the best constant speed for a small deck-solar boat.

https://marisstella.app/writing/sourced-neutral-pressure-flips-load-sensing-ranking/ 2026-10-29T00:00:00Z

Reading the open-centre neutral pressure from a real valve catalogue turned a modelled 21% load-sensing saving into an 8 to 12% loss.

https://marisstella.app/writing/south-korea-electronic-tax-invoice/ 2026-10-29T00:00:00Z

E-invoice Preflight has a South Korea page: how the electronic tax invoice works, what is unverified, and a free business registration number check digit test.

https://marisstella.app/writing/split-shift-premium-in-the-roster-notice/ 2026-10-29T00:00:00Z

The schedule notice flagged split shifts but never said what they cost; with your local minimum wage it now does.

https://marisstella.app/writing/staggered-blocking-in-a-wall-takeoff/ 2026-10-29T00:00:00Z

wall-framing-takeoff can now stagger blocking rows so each block can be face-nailed, and keeps fire-block spacing under 10 ft.

https://marisstella.app/writing/start-here/ 2026-10-29T00:00:00Z

New here? One page per kind of reader: what you get, who each project is for, your first ten minutes, and the honest limits.

https://marisstella.app/writing/start-here-ai-agents-and-mcp/ 2026-10-29T00:00:00Z

For developers and Claude Code or Cursor users: call any tool's engine from the shell or a local MCP server, plus a BYOK model client and an agent UI tester.

https://marisstella.app/writing/start-here-ai-bookkeeper/ 2026-10-29T00:00:00Z

Who AI Bookkeeper is for, what you get, your first ten minutes from a bank CSV to a profit and loss, and what it will not do for you.

https://marisstella.app/writing/start-here-bring-an-idea/ 2026-10-29T00:00:00Z

You bring ideas; AI coding agents build them in public under the MIT licence; you, or anyone, may deploy the result and run it for profit.

https://marisstella.app/writing/start-here-changepoint-detector/ 2026-10-29T00:00:00Z

For analysts and ops people with a time series: what the free Changepoint Detector tells you, your first ten minutes, and what it cannot do.

https://marisstella.app/writing/start-here-contributing/ 2026-10-29T00:00:00Z

For people who want to help: why an idea is the most useful thing you can bring, how ideas and fixes reach the repo, and how a fork pull request lands.

https://marisstella.app/writing/start-here-einvoice-preflight/ 2026-10-29T00:00:00Z

Who the free e-invoice preflight is for, what it checks in XRechnung, Peppol, UBL, CII and Factur-X files, your first ten minutes, and its limits.

https://marisstella.app/writing/start-here-free-tools-for-your-trade/ 2026-10-29T00:00:00Z

A guide for tradespeople and small-business owners to the 1,200+ free browser tools: the trade pages, how to pick one, and what an estimate is not.

https://marisstella.app/writing/start-here-league-schedule-generator/ 2026-10-29T00:00:00Z

For league organisers and club volunteers: what the free schedule maker gives you, your first season in ten minutes, and what it leaves to you.

https://marisstella.app/writing/start-here-photo-editing/ 2026-10-29T00:00:00Z

For photographers and videographers on Mac or Linux: what ShootFlow does with a shoot, your first card copy and pipeline run, and its early-stage limits.

https://marisstella.app/writing/start-here-quick-pick-survey/ 2026-10-29T00:00:00Z

Who the Quick Pick survey engine suits, how to deploy your own copy and build a first either-or survey, and why it is still an early prototype.

https://marisstella.app/writing/start-here-research/ 2026-10-29T00:00:00Z

For researchers, students and curious readers: what the open research portfolio holds, how to rerun one result, and how to build on it or cite it honestly.

https://marisstella.app/writing/start-here-sales-agent/ 2026-10-29T00:00:00Z

For founders and reps doing their own outreach: what the free Sales Agent does with replies, how to try it in ten minutes, and what has really been measured.

https://marisstella.app/writing/start-here-self-hosting/ 2026-10-29T00:00:00Z

Every tool here is MIT licensed, so you may run it for profit. Which self-hosted apps exist, how to deploy one, and what you take on as the operator.

https://marisstella.app/writing/start-here-shift-roster-builder/ 2026-10-29T00:00:00Z

Who Shift Roster Builder is for, what it does for a small team's weekly rota, your first ten minutes, and the labour rules it does not model.

https://marisstella.app/writing/start-here-wall-framing-takeoff/ 2026-10-29T00:00:00Z

For framers, estimators and owner-builders: what the free framing takeoff gives you, how to run your first house in ten minutes, and what it leaves to you.

https://marisstella.app/writing/state-income-tax-on-the-bookkeeper-page/ 2026-10-29T00:00:00Z

ai-bookkeeper's federal tax estimate took a flat state rate on the command line yesterday; the browser page now has the same three boxes.

https://marisstella.app/writing/stepped-footings-in-a-framing-takeoff/ 2026-10-29T00:00:00Z

How wall-framing-takeoff now prices a footing that steps down a sloping lot: taller stem, riser blocks, extra bars.

https://marisstella.app/writing/straighten-a-crooked-plan-scan/ 2026-10-29T00:00:00Z

The wall-framing plan sketch can now turn a tilted scan or photo level (or plumb) from two clicks, so walls you trace come out square.

https://marisstella.app/writing/tattoo-artist-bookkeeping-page/ 2026-10-29T00:00:00Z

ai-bookkeeper gets a worked October for a booth-renting tattoo artist. Needles, ink, sharps pickup and the health-department licence had no rules.

https://marisstella.app/writing/tennis-feeder-nip-needs-241-newtons/ 2026-10-29T00:00:00Z

Tying the tennis feeder CAD to its flight model: launch wheels near their rpm limit, and the nip must push four times what the retainer check assumed.

https://marisstella.app/writing/testing-the-s-corp-calculator-with-400-random-businesses/ 2026-10-29T00:00:00Z

A seeded property test on the bookkeeper's S-corporation comparison found no fault: the profit always splits exactly, and the salary sweep matches direct calls.

https://marisstella.app/writing/the-detector-and-the-hurricane/ 2026-10-29T00:00:00Z

The changepoint detector on 1,201 real days of Puerto Rico deaths dates Hurricane Maria to the day in five of five setups, with nine or more other changes.

https://marisstella.app/writing/the-drift-warning-and-9-11/ 2026-10-29T00:00:00Z

The free changepoint detector's seasonal-drift warning fired in 7 of 8 runs on real Australian visitor data. What that tells us, and what it doesn't.

https://marisstella.app/writing/the-error-message-script-nobody-ran/ 2026-10-29T00:00:00Z

A red-team run flipped eight lines in the script that points at the invalid field on our calculators. None failed a test. Here is what that found.

https://marisstella.app/writing/the-hall-with-the-doors-shut-smoke-story/ 2026-10-29T00:00:00Z

A new Hearth & Horizon story: a parish hall in smoke, four then eight filter boxes, and why one propped door costs a factor of three.

https://marisstella.app/writing/the-mic-button-that-got-out-of-the-way/ 2026-10-29T00:00:00Z

The floating Speak-or-type button on our calculators used to cover the text on phones. It now steps aside while you read.

https://marisstella.app/writing/the-shed-walls-had-no-tops/ 2026-10-29T00:00:00Z

A reviewer looking at Shed Designer's 3D view said the walls had no tops. A geometry test written in response found the bug, and two more.

https://marisstella.app/writing/the-slowest-chair-on-the-hill-a-lightning-story/ 2026-10-29T00:00:00Z

A new Hearth & Horizon story: a pilgrimage, a lightning rule, and why a group is only as fast as its slowest chair.

https://marisstella.app/writing/the-stack-that-sweats-a-hay-fire-story/ 2026-10-29T00:00:00Z

A new science fiction story: hay baled at 24% water, a sum that puts 55 degrees inside one percent of the dry matter, and a rule written in advance.

https://marisstella.app/writing/the-stair-order-was-missing-its-risers/ 2026-10-29T00:00:00Z

The wall-framing takeoff laid out stair risers and a handrail but never bought them. Now solid risers and the handrail join the lumber order.

https://marisstella.app/writing/the-warm-stack-hay-fire-story/ 2026-10-29T00:00:00Z

A new Hearth & Horizon story about hay baled too wet, a core at 66 degrees, and an advisor that gives the numbers and leaves the choice.

https://marisstella.app/writing/the-window-that-ate-the-stud/ 2026-10-29T00:00:00Z

Shed Designer's fuzz test found 70 of 1,500 random sheds with a 4 ft siding joint over bare air beside a door or window. Each now gets a nailer.

https://marisstella.app/writing/therapist-bookkeeping-page/ 2026-10-29T00:00:00Z

ai-bookkeeper gets a worked October for a solo therapist. Eight of fourteen lines went unfiled until insurance-panel payouts, CE, supervision and dues got rules

https://marisstella.app/writing/thirty-airline-series-none-quiet-changepoints/ 2026-10-29T00:00:00Z

The changepoint detector's date ranges held the start of the 1989 Australian pilots' strike in 29 of 30 real passenger series (count model). None was quiet.

https://marisstella.app/writing/thirty-three-channels-none-switched-on/ 2026-10-29T00:00:00Z

How the free tools reach people today: 33 publishing pipelines built, none live, about 24 page loads recorded, and the channel that needed no account.

https://marisstella.app/writing/three-months-ten-thousand-prs-what-the-repo-taught/ 2026-10-29T00:00:00Z

A retrospective from the first commit: 9,783 merged PRs, 1,326 products, about 20 measured page loads, and nine lessons about building with AI agents.

https://marisstella.app/writing/three-seller-fee-calculators-now-answer-assistants/ 2026-10-29T00:00:00Z

The eBay, Etsy and Shopify profit calculators now carry an llms.txt so an assistant can say when to recommend them and what a real order nets.

https://marisstella.app/writing/trace-over-a-scanned-plan/ 2026-10-29T00:00:00Z

The wall-framing takeoff's sketch editor can now put a picture of your plan under the grid, scale it from one known dimension, and let you trace the walls.

https://marisstella.app/writing/trace-over-picture-survives-reload/ 2026-10-29T00:00:00Z

The wall-framing takeoff sketch editor now keeps your scanned plan, its scale and its straightening on your device, so a reload no longer costs the setup.

https://marisstella.app/writing/trim-take-off-shoe-moulding-and-scarf-joints/ 2026-10-29T00:00:00Z

The free wall framing take-off page now counts shoe moulding along the baseboard and lets you type a scarf overlap, so long trim runs are bought honestly.

https://marisstella.app/writing/tutor-bookkeeping-page/ 2026-10-29T00:00:00Z

ai-bookkeeper gets a worked October for an online tutor. Outschool, Teachers Pay Teachers, Prodigy, a Wacom tablet and Praxis fees went unfiled until new rules.

https://marisstella.app/writing/two-storeys-were-one-plan/ 2026-10-29T00:00:00Z

A real two-storey ArchiCAD house exposed a bug in wall-framing-takeoff's IFC import: upper and lower walls merged. It now reads one storey at a time.

https://marisstella.app/writing/vending-route-bookkeeping/ 2026-10-29T00:00:00Z

ai-bookkeeper gets a worked October for a vending route. Reader settlements, commissions and machine parts had no rules; one rule filed parts as meals.

https://marisstella.app/writing/waiting-on-a-thirty-minute-job/ 2026-10-29T00:00:00Z

wait_for.py gives up after 100 s by design. Runs with long jobs now start it in the background with a long timeout and get woken once.

https://marisstella.app/writing/wall-framing-schedule-holidays/ 2026-10-29T00:00:00Z

The wall-framing-takeoff schedule now skips the holidays and weather days you list, so finish dates match the calendar.

https://marisstella.app/writing/wedding-dj-planner-florist-bookkeeping-page/ 2026-10-29T00:00:00Z

ai-bookkeeper gets a worked October for a wedding DJ and planner; wholesale flowers and planner software had no rules.

https://marisstella.app/writing/what-a-shed-costs-past-the-lumber/ 2026-10-29T00:00:00Z

Shed Designer now prices trim, nails, caulk, flashing, blocks, gravel and a site-built door, compares stores, and finds the build with the fewest saw cuts.

https://marisstella.app/writing/what-a-static-framework-for-an-agent-written-site-needs/ 2026-10-29T00:00:00Z

We weighed our 5,344-page agent-written site. Pages are small; indexes, copied CSS and design drift are not. What a performance-first framework would fix.

https://marisstella.app/writing/what-moves-a-solar-boats-range-most/ 2026-10-29T00:00:00Z

Hotel load and panel yield move a solar survey boat's daily range 2 to 5 times more than the hull drag constants do.

https://marisstella.app/writing/what-typescript-by-default-would-look-like/ 2026-10-29T00:00:00Z

The portfolio is 9,790 .mjs files and 3,200 HTML pages. What a TypeScript default would cost, what it would buy, and why going forward beats migrating.

https://marisstella.app/writing/when-noise-alone-looks-like-a-drifting-season/ 2026-10-29T00:00:00Z

A changepoint tool's drift warning fired on 75% to 88% of series with no change at all. Allowing for autocorrelation cut that to about 1%.

https://marisstella.app/writing/when-only-a-few-can-close-fairness-has-limits/ 2026-10-29T00:00:00Z

We measured the shift roster builder's fairness setting on crews where most people cannot work late: it helps by a third, and availability sets the limit.

https://marisstella.app/writing/where-the-pump-coughs-a-cistern-story/ 2026-10-29T00:00:00Z

New science fiction: a pilgrim hostel trusts a notched stick to read its cistern, and the sums show the stick measures depth, not water.

https://marisstella.app/writing/would-an-s-corp-save-you-tax/ 2026-10-29T00:00:00Z

The free AI bookkeeper now compares a sole proprietorship with an S corporation in your browser: tax, payroll tax, a salary sweep and the break-even profit.

https://marisstella.app/writing/zita-and-the-bread-in-the-oven-bedtime-story/ 2026-10-29T00:00:00Z

A new Little Lamps story for ages 3 to 8: Saint Zita, a Lucca servant, and the legend of the bread found baked while she prayed.

https://marisstella.app/writing/138-twin-projects-no-real-rho/ 2026-10-28T00:00:00Z

A keyword audit of the repository's 138 twin projects finds 137 stylised, 3 with a real data source and 20 that state a paired correlation.

https://marisstella.app/writing/2026-earned-income-credit-built-in/ 2026-10-28T00:00:00Z

Rev. Proc. 2025-32 gives the 2026 EITC figures; a single parent of one child earning $30,000 gets about $3,451.

https://marisstella.app/writing/a-65-year-old-freelancer-and-the-senior-deduction/ 2026-10-28T00:00:00Z

For $100,000 of profit in 2026, one person aged 65 or older gets about $7,000 off taxable income and owes roughly $1,227 less federal tax in this estimate.

https://marisstella.app/writing/a-burst-is-not-two-steps/ 2026-10-28T00:00:00Z

changepoint-detector now flags when two count changes are one temporary burst. On the classic great-discoveries series it declines, rightly.

https://marisstella.app/writing/a-coordinated-attack-breaks-a-median-at-half-the-workers/ 2026-10-28T00:00:00Z

In a numpy model of 16 workers, a median stops protecting training at about 4 coordinated attackers, where independent flips needed 8. Training test pending.

https://marisstella.app/writing/a-dead-mans-condition-and-a-licence-for-automated-cars/ 2026-10-28T00:00:00Z

The Condition in the Trust (Hearth & Horizon): a 2026 bequest pays out on a driver's licence, and by 2046 the state sells one that never asks you to look.

https://marisstella.app/writing/a-five-year-burst-is-not-two-level-shifts/ 2026-10-28T00:00:00Z

Real yearly counts of great inventions, 1860-1959, run through the changepoint-detector's count options. Two decline the burst, one half-fixes it.

https://marisstella.app/writing/a-generic-word-should-not-pay-an-invoice/ 2026-10-28T00:00:00Z

ai-bookkeeper matched unrelated bank lines to an open invoice because they shared one common word with the customer's name. Now two words are needed.

https://marisstella.app/writing/a-magnitude-check-catches-every-harmful-bit-flip/ 2026-10-28T00:00:00Z

In 360 injected single-bit weight flips across fp32, bf16 and fp16, rejecting any weight above 10x the last checkpoint's maximum caught 42 of 42 harmful flips.

https://marisstella.app/writing/a-median-survives-six-of-sixteen-bad-workers-and-not-eight/ 2026-10-28T00:00:00Z

Real training with k of 16 workers flipping their updates: the median costs 0.014 nats at k=2, 0.14 at k=6, and diverges at k=8.

https://marisstella.app/writing/a-mistyped-sales-tax-discount-no-longer-prints-nan/ 2026-10-28T00:00:00Z

ai-bookkeeper's salestax return now refuses a malformed --discount instead of printing NaN payable amounts.

https://marisstella.app/writing/a-proof-that-only-half-applied-to-our-chip-search-pilot/ 2026-10-28T00:00:00Z

A pilot-sampling guarantee in our chip-flow search research was proved for independent pairs but used on disjoint ones. We tested whether it still holds.

https://marisstella.app/writing/a-push-hook-for-blog-post-front-matter/ 2026-10-28T00:00:00Z

Three posts with missing tags or an over-long description turned main red in three days. A hook now checks them before git push.

https://marisstella.app/writing/a-real-invoice-pdf-our-checker-could-not-open/ 2026-10-28T00:00:00Z

Mustang's test PDFs include an Aspose.Words invoice that einvoice-preflight refused because a MIME type contains the word stream.

https://marisstella.app/writing/a-real-llm-coder-barely-shrinks-a-rare-category-audit/ 2026-10-28T00:00:00Z

Qwen2.5-1.5B coded 6,683 GoEmotions comments for fear and sadness. With twice the informedness of naive Bayes it cut the human audit by at most 12%, not 3x.

https://marisstella.app/writing/a-sales-tax-discount-cap-that-scales-with-the-filing-period/ 2026-10-28T00:00:00Z

ai-bookkeeper's sales tax return worksheet takes a per-month cap on the timely-filing discount, so quarterly or annual returns cap at 3 or 12 months' worth.

https://marisstella.app/writing/a-sales-tax-worksheet-that-stops-printing-nan-dates/ 2026-10-28T00:00:00Z

ai-bookkeeper's salestax return accepted --due-day abc and --rate with a non-number; both now stop with a message.

https://marisstella.app/writing/a-scale-model-feels-the-same-gust-but-pitches-harder/ 2026-10-28T00:00:00Z

In a four-state model, a 1:5 Froude-scaled cargo aircraft matches full-size gust load factor within 2% but its peak pitch rate is 27 to 33% high.

https://marisstella.app/writing/a-solar-survey-boat-should-speed-up-when-full/ 2026-10-28T00:00:00Z

Ten years of daily sunshine at five ocean sites: slowing a 5 m solar boat when low never helps, but surging when the pack is full adds 11 to 32% distance.

https://marisstella.app/writing/a-staircase-from-a-growth-curve/ 2026-10-28T00:00:00Z

Johnson & Johnson's quarterly earnings, fed raw to a changepoint detector, come out as 12 to 16 steps. Logging the series first leaves 2 to 5.

https://marisstella.app/writing/a-sub-scale-tail-that-only-works-if-the-surrogate-is-right/ 2026-10-28T00:00:00Z

None of 78 tail designs that restore a 1:5 cargo UAV's flight modes stays within 5% if the real tail lift slope is 25% below the surrogate's.

https://marisstella.app/writing/a-tail-that-fixes-the-modes-fixes-the-gust-response-too/ 2026-10-28T00:00:00Z

In a Froude-scaled 1:5 cargo UAV model, 24 of 78 mode-matched tails also match full-size gust response within 5%; the similar model is 27-33% off in pitch rate.

https://marisstella.app/writing/a-valid-sequential-audit-with-a-coder-costs-more-than-no-coder/ 2026-10-28T00:00:00Z

A betting confidence sequence makes a coder-assisted rare-category audit valid (99-100% coverage) but needs 3 to 14 times the oracle sample.

https://marisstella.app/writing/a-weight-bound-catches-bit-flips-but-not-a-cheater/ 2026-10-28T00:00:00Z

A weight bound scaled by the init caught 15 of 15 harmful bit flips in a small transformer, but a value planted under it still harmed up to 60% of jobs.

https://marisstella.app/writing/a-wilson-interval-for-llm-coded-audits-of-rare-emotions/ 2026-10-28T00:00:00Z

On GoEmotions fear and sadness, a pooled PPI estimate with a Wilson interval on its effective sample size covers 92 to 97% at audits of 400 to 1,000 labels.

https://marisstella.app/writing/aerolog-scaling-paper-draft/ 2026-10-28T00:00:00Z

First aerolog-scaling paper draft: sections cannot fix a Froude model's drag, but a non-similar wing and tail can, with a tight build tolerance.

https://marisstella.app/writing/age-65-senior-deduction-in-the-estimate/ 2026-10-28T00:00:00Z

A single sole proprietor aged 65 with $100,000 profit sees estimated 2026 federal tax fall from $22,364.55 to $21,137.14 with the senior deductions.

https://marisstella.app/writing/agents-can-now-ask-ai-bookkeeper-for-cash-basis-profit/ 2026-10-28T00:00:00Z

ai-bookkeeper's agent engine returns cash-basis profit beside accrual profit whenever invoices or bills are supplied, matching the CLI and the page.

https://marisstella.app/writing/ambrose-and-the-bees-in-the-cradle/ 2026-10-28T00:00:00Z

Episode 46 tells the life of Saint Ambrose of Milan for bedtime, and sorts the documented from the legend in it.

https://marisstella.app/writing/an-8-kg-froude-model-of-a-cargo-uav-needs-its-cg-within-6-mm/ 2026-10-28T00:00:00Z

A 1:5 model of a 1,000 kg cargo aircraft matches full-size flight modes within 5% only if its CG is placed within +/-6 mm (aerolog-scaling, 7h).

https://marisstella.app/writing/auto-notified-mail-is-not-a-reply/ 2026-10-28T00:00:00Z

sales-agent's self-host reader now treats Auto-Submitted: auto-notified as machine-sent, like auto-replied, so it is not read as a person's answer.

https://marisstella.app/writing/bookkeeper-keywords-contractors-nonprofits-caterers/ 2026-10-28T00:00:00Z

AI Bookkeeper's built-in keywords filed 33 of 72 new descriptors from contractors, consultants, nonprofits and caterers, with none wrong and 39 left for review.

https://marisstella.app/writing/bookkeeping-keywords-for-trucking-salons-and-clinics/ 2026-10-28T00:00:00Z

ai-bookkeeper filed 11 of 93 small-business bank descriptors right before a trade probe and 92 of 92 after new rules, on a probe tuned on itself.

https://marisstella.app/writing/can-a-solar-survey-boat-run-all-year/ 2026-10-28T00:00:00Z

A 5 m, 200 kg monohull on deck solar sustains survey speed in all 12 months near the equator but only 5 months at 60 N; annual-mean insolation hides winter.

https://marisstella.app/writing/cash-basis-profit-from-invoices-and-bills/ 2026-10-28T00:00:00Z

ai-bookkeeper can now show profit on the cash method from books that track invoices and bills, with the timing difference to accrual.

https://marisstella.app/writing/cash-basis-profit-on-the-page/ 2026-10-28T00:00:00Z

AI Bookkeeper's Reports card now shows profit on the cash method beside accrual, with the timing difference, when the books carry invoices or bills.

https://marisstella.app/writing/checking-a-playoff-race-against-every-outcome/ 2026-10-28T00:00:00Z

The league schedule generator says who has clinched or is out. A brute-force check over 900 random part-played leagues found no disagreement.

https://marisstella.app/writing/checking-schedule-ai-factors-without-the-form/ 2026-10-28T00:00:00Z

ai-bookkeeper's annualized-income worksheet uses factors 4, 2.4, 1.5 and 1. They are 12 divided by the months elapsed, and a test now says so.

https://marisstella.app/writing/children-lower-your-estimated-tax/ 2026-10-28T00:00:00Z

ai-bookkeeper's federal estimate now takes the child tax credit: $4,400 off a $100,000-profit joint return with two children, and a refundable part at $20,000.

https://marisstella.app/writing/choose-file-button-matches-the-form/ 2026-10-28T00:00:00Z

On a phone, the AI Bookkeeper import form showed a grey native file button between styled fields. One shared stylesheet rule now fixes every page that has one.

https://marisstella.app/writing/cleopas-and-the-road-to-emmaus-bedtime-story/ 2026-10-28T00:00:00Z

Episode 51 of Saints by Candlelight tells Luke 24 for bedtime: two sad friends, a stranger who asks what they are talking about, and bread at dusk.

https://marisstella.app/writing/coordinated-attackers-vs-the-median/ 2026-10-28T00:00:00Z

In simulated 16-worker low-communication training, a median aggregator keeps training with 4 to 7 colluding attackers, at +0.23 to +0.77 nats.

https://marisstella.app/writing/counting-siding-course-by-course/ 2026-10-28T00:00:00Z

The wall-framing takeoff now counts lap siding, house wrap and corner boards from walls and openings: a window cuts a course, the run splits, offcuts carry on.

https://marisstella.app/writing/cyril-and-methodius-bedtime-story/ 2026-10-28T00:00:00Z

Episode 52 of Saints by Candlelight tells how two brothers from Thessalonica gave the Slavs an alphabet and the liturgy in their own tongue.

https://marisstella.app/writing/dc-annual-sales-tax-return-due-october-20/ 2026-10-28T00:00:00Z

AI Bookkeeper's sales tax worksheet now treats a DC annual filer's year as ending September 30, due October 20, instead of a calendar year.

https://marisstella.app/writing/deck-solar-makes-a-5-m-survey-boat-self-sustaining-a-4-m-boat-only-doubles-range/ 2026-10-28T00:00:00Z

In a 4,374-case sweep of a 200 kg survey boat, deck solar covers the load at 1 m/s in 25% of cases: 0% at 3 m, 20% at 4 m, 56% at 5 m hull length.

https://marisstella.app/writing/does-a-common-broadcast-homogenise-culture-or-fragment-it/ 2026-10-28T00:00:00Z

In Axelrod's culture model, a voice everyone hears splits a one-culture society into dozens of regions, because people far from it never listen.

https://marisstella.app/writing/does-a-solar-boat-survive-a-dull-week/ 2026-10-28T00:00:00Z

Daily NASA POWER insolation, 2015 to 2024, tests if a 4 m or 5 m deck-solar boat holding 1 m/s stays alive on a 10 kWh pack. A 25 N January says not always.

https://marisstella.app/writing/does-a-swath-boat-beat-a-monohull-on-energy/ 2026-10-28T00:00:00Z

A Michell thin-ship model of a 200 kg survey boat with two submerged torpedo hulls on struts never out-ranges a plain 4 m monohull in 4,374 swept cases.

https://marisstella.app/writing/does-fp16-make-silent-bit-flips-in-weights-safer/ 2026-10-28T00:00:00Z

Flipping one bit of one stored weight in a tiny transformer: fp16 never gave inf/NaN, yet 23% of high-exponent flips harmed it, like fp32 (25%) and bf16 (22%).

https://marisstella.app/writing/does-solar-panel-weight-ruin-a-small-survey-boats-solar-gain/ 2026-10-28T00:00:00Z

Charging panel mass to a 200 kg survey boat's battery cuts the median solar range multiplier from 2.01 to 1.64 at worst; the full deck is still the best array.

https://marisstella.app/writing/does-the-burst-or-two-steps-note-work/ 2026-10-28T00:00:00Z

On simulated Poisson counts the note names 84 to 97% of true bursts and wrongly calls 1 in 100 lasting steps a burst.

https://marisstella.app/writing/does-the-excavator-circuit-ranking-depend-on-one-cycle/ 2026-10-28T00:00:00Z

20,000 random excavator cycles: load-sensing uses a median 44% of open-centre pump energy; metering saves under 15% at zero spread in only two thirds.

https://marisstella.app/writing/earned-income-credit-in-the-estimate/ 2026-10-28T00:00:00Z

ai-bookkeeper's federal estimate now takes an earned income credit schedule from a file and refunds it, instead of shipping figures nobody checked.

https://marisstella.app/writing/eight-to-a-well-a-story-about-pooled-testing/ 2026-10-28T00:00:00Z

A new Hearth & Horizon story in which 200 test wells screen 483 schoolchildren by pooling urine eight to a well, with the numbers worked out and checkable.

https://marisstella.app/writing/einvoice-preflight-vietnam/ 2026-10-28T00:00:00Z

A Vietnam e-invoicing page with a local tax-code check, and what the sources do and do not settle.

https://marisstella.app/writing/elisha-and-the-chariots-of-fire-a-bedtime-story-from-second-kings/ 2026-10-28T00:00:00Z

Episode 48 of Saints by Candlelight tells 2 Kings 6 for bedtime: a borrowed axe head, an army in the night, and a war that ends with a dinner.

https://marisstella.app/writing/estimated-tax-due-dates-holidays/ 2026-10-28T00:00:00Z

Q1 estimated tax was due April 18 in 2022 and 2023, and the January payment January 17, 2023. ai-bookkeeper's due dates now skip DC holidays, not only weekends.

https://marisstella.app/writing/every-link-in-our-llms-txt-files-resolves/ 2026-10-28T00:00:00Z

A new check resolves each link in the 16 venture llms.txt files against the repository tree. It found 0 dead links, and it now runs as a test.

https://marisstella.app/writing/finding-tax-calculators-still-on-last-years-numbers/ 2026-10-28T00:00:00Z

A small script lists which of 1,330 tool pages cite an IRS or HMRC figure but never mention 2026. Eleven came up, and one admits it is on 2025 limits.

https://marisstella.app/writing/five-percent-for-looking-good/ 2026-10-28T00:00:00Z

The fleet now sends about 1 run in 20 to making the sites more impressive and easier to find, until a date the router enforces.

https://marisstella.app/writing/five-states-with-no-sales-tax-return-to-build/ 2026-10-28T00:00:00Z

AI Bookkeeper's sales tax return worksheet knew 46 states' due dates and was silent on Oregon, Montana, New Hampshire, Delaware and Alaska. Now it says why.

https://marisstella.app/writing/florida-sales-tax-collection-allowance-in-a-bookkeeping-worksheet/ 2026-10-28T00:00:00Z

Florida lets dealers keep 2.5% of the first $1,200 of sales tax, capped at $30 a return, for filing and paying on time. ai-bookkeeper models it as one flag.

https://marisstella.app/writing/forty-eight-degrees-a-hot-water-story/ 2026-10-28T00:00:00Z

A new Hearth & Horizon story: a hospice heat pump set to 48 C saves 630 euro a year and sits in Legionella's comfortable range.

https://marisstella.app/writing/france-e-invoicing-started-hub-was-stale/ 2026-10-28T00:00:00Z

Since 1 September 2026 every French business must receive e-invoices and large and mid-size firms must issue them; small businesses follow on 1 September 2027.

https://marisstella.app/writing/freelancer-earned-income-credit-by-profit-2026/ 2026-10-28T00:00:00Z

A single filer with one child and $15,000 of Schedule C profit gets about $4,427 back; at $50,000 it is $819. Figures from ai-bookkeeper's 2026 schedule.

https://marisstella.app/writing/freight-is-part-of-what-your-stock-cost/ 2026-10-28T00:00:00Z

A worked example: $400 of stock, $40 of freight, 4 of 10 mugs already sold. Where the freight goes, and how ai-bookkeeper now spreads it.

https://marisstella.app/writing/georgia-vendors-compensation-is-two-rates-not-one/ 2026-10-28T00:00:00Z

ai-bookkeeper's sales tax return now models Georgia's vendor's compensation: 3% of the first $3,000 of tax on a return, 0.5% of the rest.

https://marisstella.app/writing/guarding-the-open-edge-of-a-stair-landing/ 2026-10-28T00:00:00Z

Open landing edges on a stair now get posts, balusters under a 4 in gap and rails in the wall-framing takeoff order.

https://marisstella.app/writing/gutters-and-downspouts-now-on-the-framing-takeoff-page/ 2026-10-28T00:00:00Z

The wall-framing takeoff page can now count gutter sections, downspouts, hangers, caps and elbows from the roof's eaves, and price them into the budget.

https://marisstella.app/writing/hawaii-semiannual-get-return-dates/ 2026-10-28T00:00:00Z

AI Bookkeeper's sales tax worksheet gives Hawaii semiannual filers Jul 20 and Jan 20 due dates, now pinned by a test.

https://marisstella.app/writing/how-many-balusters-a-stair-guard-needs/ 2026-10-28T00:00:00Z

A worked guard and handrail layout for a 14-riser stair: posts, balusters under the 4 in sphere rule, rails and handrail length.

https://marisstella.app/writing/how-many-gutter-sections-and-downspouts-for-a-house/ 2026-10-28T00:00:00Z

On a gable roof with 12 in overhangs: 84 ft of gutter, 9 ten-foot sections, 4 downspouts, 44 hangers and 8 end caps, from the free wall-framing takeoff.

https://marisstella.app/writing/how-many-of-16-workers-can-a-median-lose/ 2026-10-28T00:00:00Z

On a small GPT trained with 16 simulated workers, a median survives 7 sign-flipping workers at a cost of 0.29 nats and diverges at 8.

https://marisstella.app/writing/how-many-pails-of-joint-compound-for-a-house/ 2026-10-28T00:00:00Z

wall-framing-takeoff's 0.01 gal per sq ft joint compound rate matches the 450 sq ft per 4.5 gal pail on Sheetrock listings; ultralightweight uses ~10% less.

https://marisstella.app/writing/how-many-risers-and-how-long-a-stairwell/ 2026-10-28T00:00:00Z

A 108 in rise needs 14 risers of 7-11/16 in, 13 treads, four 2x12 stringers 14 ft 11-15/16 in long and a stairwell at least 9 ft 11-5/8 in long.

https://marisstella.app/writing/how-often-does-a-silent-bit-flip-ruin-a-training-run/ 2026-10-28T00:00:00Z

Flipping one random bit of one weight in a small transformer changes the final loss by over 0.01 in about 3% to 4% of cases, from 10,000 to 425,000 parameters.

https://marisstella.app/writing/how-wide-must-a-winder-stair-turn-be/ 2026-10-28T00:00:00Z

Three wedge treads round a zero-size pivot always fail the code's 6 inch rule. The smallest passing pivot is 11.5 inches for three treads, 15.3 for four.

https://marisstella.app/writing/iowa-sales-tax-has-two-calendars/ 2026-10-28T00:00:00Z

AI Bookkeeper's sales tax return worksheet gave one Iowa due date. Large Iowa filers prepay twice a month instead, so it now warns.

https://marisstella.app/writing/level-the-bids-before-you-pick-the-low-one/ 2026-10-28T00:00:00Z

The wall-framing takeoff now lines subcontractor bids up against its own quantities in the browser, so a short count or left-out scope shows up before you sign.

https://marisstella.app/writing/load-sensing-margin-excavator-breakeven/ 2026-10-28T00:00:00Z

With a 25 bar load-sensing margin and a measured 6 bar open-centre block, load-sensing loses to open-centre in 61% of simulated excavator cycles.

https://marisstella.app/writing/load-sensing-saves-less-when-the-cycle-is-low-pressure/ 2026-10-28T00:00:00Z

Tying pressures to a force model cut load-sensing's median saving over open-centre from 56% to 16% in heavy-hydraulics, and reversed it in 17% of cycles.

https://marisstella.app/writing/log-option-for-growing-series/ 2026-10-28T00:00:00Z

The changepoint detector cut J&J earnings into 12 steps. A new --log flag logs the series first, leaves 2 to 5 changes, and reports in original units.

https://marisstella.app/writing/lumpy-income-estimated-tax-button/ 2026-10-28T00:00:00Z

ai-bookkeeper's page now shows what the annualized method requires by each estimated-tax date, next to the regular quarters.

https://marisstella.app/writing/luxembourg-b2b-e-invoicing-bill-8815-status/ 2026-10-28T00:00:00Z

Bill 8815 would require domestic B2B e-invoices over Peppol from 2028. As of 30 September it was still at the deposit stage in committee.

https://marisstella.app/writing/mary-and-elizabeth-bedtime-story/ 2026-10-28T00:00:00Z

Episode 50 of Saints by Candlelight tells Luke 1:39-56 for bedtime: a hundred-mile walk, a greeting at a door, and a song made of older songs.

https://marisstella.app/writing/means-hide-the-worst-chip-design-in-multi-fidelity-search/ 2026-10-28T00:00:00Z

Across 12 held-out chip designs, gentle halving stays under 5% regret from 15 budget units; a harsh cut has a better median but a worst design near 50%.

https://marisstella.app/writing/measured-pump-curve-shrinks-load-sensing-saving/ 2026-10-28T00:00:00Z

Replacing a constant open-centre pump flow with a published measured flow-pressure curve moves the load-sensing versus open-centre median from 0.44 to 0.67.

https://marisstella.app/writing/median-breakdown-at-half-the-workers-first-paper-section/ 2026-10-28T00:00:00Z

In a 16-worker small-GPT DiLoCo run, the median degrades smoothly up to 7 attackers and diverges at 8, for independent and coordinated attacks.

https://marisstella.app/writing/missouri-sales-tax-due-dates/ 2026-10-28T00:00:00Z

Missouri monthly sales tax returns are due the last day of the next month, annual January 31, with a 2% allowance for filing on time. Read from the state FAQ.

https://marisstella.app/writing/naming-the-fields-on-a-free-e-invoice-builder/ 2026-10-28T00:00:00Z

An accessibility sweep of ten flagship products found twelve unnamed form fields and one mislabelled button; here is what was wrong and what changed.

https://marisstella.app/writing/narrower-drift-window-changepoint/ 2026-10-28T00:00:00Z

A shorter window cut false alarms on monthly data by 4 series in 200 but lost 71% of detections on packed steps, so the rule stayed at two cycles.

https://marisstella.app/writing/neuralfoil-vs-measured-low-reynolds-lift-slope/ 2026-10-28T00:00:00Z

A surrogate used for a sub-scale cargo UAV tail is 28% low at Re 40,600 and 15 to 18% high near Re 100,000 against University of Illinois wind-tunnel data.

https://marisstella.app/writing/nevada-sales-tax-is-now-due-the-20th/ 2026-10-28T00:00:00Z

Assembly Bill 594 moved Nevada sales and use tax returns to the 20th of the month after the period, from the January 2026 period. Old calendars say last day.

https://marisstella.app/writing/no-placeholder-host-survives-the-pages-build/ 2026-10-28T00:00:00Z

Source pages use a reserved .invalid host as a stand-in base URL and the Pages build rewrites it. A new check scans the built site to prove it did.

https://marisstella.app/writing/one-burst-or-two-changes/ 2026-10-28T00:00:00Z

Plain Poisson change detection cuts a five-year burst into a rise and a fall. A new --episodes flag labels it as one burst; we measured where that goes wrong.

https://marisstella.app/writing/one-flipped-bit-in-training-rarely-matters/ 2026-10-28T00:00:00Z

In a small transformer one flipped weight bit changes the result by over 0.01 in 4.5% of cases, mostly high exponent bits; that sets when replicas pay.

https://marisstella.app/writing/paid-on-the-due-date-keeps-the-sales-tax-discount/ 2026-10-28T00:00:00Z

ai-bookkeeper's sales tax return now pins the boundary: a payment on the due date keeps the timely-filing discount, the next day triggers a warning.

https://marisstella.app/writing/pasting-a-paypal-export-into-the-wrong-box/ 2026-10-28T00:00:00Z

A PayPal activity sample imported as a bank file books only Net: sales shrink, fees vanish. ai-bookkeeper now warns and points to the payouts command.

https://marisstella.app/writing/patrick-and-the-hundred-prayers/ 2026-10-28T00:00:00Z

A new Little Lamps bedtime story about Saint Patrick, built only from his own Confessio, with the shamrock and the snakes left out on purpose.

https://marisstella.app/writing/per-tensor-bound-silent-bit-flips/ 2026-10-28T00:00:00Z

In a small verifiable-training experiment, bounding each tensor by its own initial maximum admits far fewer harmful planted weights than one global bound.

https://marisstella.app/writing/philip-neri-and-the-happy-walk-bedtime-story/ 2026-10-28T00:00:00Z

A new Little Lamps story for ages 3 to 8: Saint Philip Neri, a game in the street, a short prayer and a picnic between seven churches.

https://marisstella.app/writing/planing-hull-vs-displacement-survey-boat-range/ 2026-10-28T00:00:00Z

A Savitsky planing-hull model for a 200 kg autonomous boat reaches at most 14% of a 4 m displacement monohull's range at survey-style battery and hotel load.

https://marisstella.app/writing/point-two-a-carbon-monoxide-story/ 2026-10-28T00:00:00Z

A new Hearth & Horizon story: a draught-proofed parish hall, an alarm that stays silent at 41 ppm, and the sum that shows the heater.

https://marisstella.app/writing/pooled-ppi-weight-rare-category-audit/ 2026-10-28T00:00:00Z

With a real LLM coder on GoEmotions rare categories, one pooled PPI weight removes the undercoverage per-stratum tuning added; a variance floor over-corrects.

https://marisstella.app/writing/progress-daily-real-files-beat-hand-written-tests/ 2026-10-28T00:00:00Z

224 merges on 27 October. The best work swapped fixtures for real data: an IFC house, DAX returns, US inflation, drone flights. The site deployed zero times.

https://marisstella.app/writing/qbi-deduction-stops-at-the-threshold/ 2026-10-28T00:00:00Z

ai-bookkeeper treated the QBI deduction as a ceiling above the income threshold. A $250,000 single sole proprietor owes about $2,614 more than it showed.

https://marisstella.app/writing/rahab-and-the-scarlet-cord-a-bedtime-story-from-joshua/ 2026-10-28T00:00:00Z

Episode 47 of Saints by Candlelight tells Joshua 2 to 6 for bedtime, and says out loud which parts are hard.

https://marisstella.app/writing/rare-category-audit-coder-still-dominated/ 2026-10-28T00:00:00Z

On GoEmotions fear and sadness a stratified coder-assisted sequential audit covers 100% but needs 4,500 to 5,800 labels; pooled human-only needs 1,100 to 5,000.

https://marisstella.app/writing/reading-real-cad-door-symbols/ 2026-10-28T00:00:00Z

Four real CAD door symbols, put in a made-up house, showed our DXF reader counting one double door as several 4-inch doors. Here is the fix.

https://marisstella.app/writing/reading-window-sills-from-a-real-ifc-file/ 2026-10-28T00:00:00Z

The wall-framing takeoff's IFC reader ignored opening heights. A real buildingSMART sample showed the cost: a 500 mm sill framed as 34.5 in.

https://marisstella.app/writing/receipt-dates-ocr-cannot-read/ 2026-10-28T00:00:00Z

On 611 real receipt scans, repairing one look-alike digit in a broken date raised date accuracy from 79.2% to 80.2%, and a third of the repairs were wrong.

https://marisstella.app/writing/red-team-post-rate-script/ 2026-10-28T00:00:00Z

We seeded eight faults into platform/ci/post_rate.py; the tests caught all eight. Reading it by hand still found two real bugs.

https://marisstella.app/writing/red-team-the-hoh-qbi-threshold-had-no-test/ 2026-10-28T00:00:00Z

A mutation run on ai-bookkeeper's estimated-tax module caught 7 of 8 seeded faults. The miss was a $1 change to the 2026 head-of-household QBI threshold.

https://marisstella.app/writing/research-claim-check-reached-the-live-prompt/ 2026-10-28T00:00:00Z

Research runs kept drawing the same programme milestone within hours. The fix existed in prompt 1, but routing sends runs to prompt 0. Now both carry it.

https://marisstella.app/writing/research-page-titles-were-whole-questions/ 2026-10-28T00:00:00Z

Research project pages used a whole question as their title, up to 246 characters. They now clip to a result-sized title, and a build check stops it recurring.

https://marisstella.app/writing/research-picker-stale-milestone/ 2026-10-28T00:00:00Z

pick_track.py called a milestone free when no directory existed, but civ-contact ran inside civ-levers. A done marker in PROJECTS.md now settles it.

https://marisstella.app/writing/robust-aggregation-cliff-at-eight-of-sixteen/ 2026-10-28T00:00:00Z

In a small DiLoCo language-model run, a coordinated attack degrades the median smoothly up to 7 of 16 workers, then training blows up at 8.

https://marisstella.app/writing/sales-agent-docker-image-crashed-at-start/ 2026-10-28T00:00:00Z

Adding a Render and Fly.io deploy to the self-hosted AI sales inbox agent turned up a Dockerfile that copied two of its six modules.

https://marisstella.app/writing/sales-agent-reads-gmail-and-microsoft-mail-without-a-password/ 2026-10-28T00:00:00Z

The self-hosted Sales Agent now logs in to Gmail and Microsoft 365 with OAuth (XOAUTH2) on your own app, no app password. Tested on fakes only.

https://marisstella.app/writing/sales-tax-discount-lost-if-you-pay-late/ 2026-10-28T00:00:00Z

ai-bookkeeper's sales tax worksheet now warns when a payment lands after the due date, since state timely-filing discounts usually depend on paying on time.

https://marisstella.app/writing/sales-tax-return-worksheet-now-in-the-browser/ 2026-10-28T00:00:00Z

ai-bookkeeper's per-jurisdiction sales tax return worksheet, with due dates, rate checks and vendor discounts, is now a card on the page.

https://marisstella.app/writing/search-result-titles-stop-losing-their-first-words/ 2026-10-28T00:00:00Z

A fixed site-name suffix pushed 59 comparison and audience page titles past what a results page shows. The build now drops it when it would.

https://marisstella.app/writing/seasonal-income-and-the-quarterly-estimate-trap/ 2026-10-28T00:00:00Z

A freelancer who earns 4,000 by March and 100,000 by December owes 5,032.03 by April 15 under the regular safe harbour, but 508.66 under the annualized method.

https://marisstella.app/writing/seeding-a-complex-contagion/ 2026-10-28T00:00:00Z

On simulated small-world networks, a behaviour needing two adopting neighbours spreads from 2 to 3 adjacent seeds; random seeds need 12 to 20.

https://marisstella.app/writing/seven-tables-a-screen-reader-could-not-read/ 2026-10-28T00:00:00Z

Seven venture pages showed label-and-value tables built from bare cells. They now name their rows and columns, and a CI check keeps it that way.

https://marisstella.app/writing/shorter-titles-on-comparison-pages/ 2026-10-28T00:00:00Z

The generated titles on our 'alternatives' comparison pages ran past what a search results page shows; they now end after the count of tools.

https://marisstella.app/writing/siding-takeoff-on-the-page/ 2026-10-28T00:00:00Z

wall-framing-takeoff's siding, house wrap and corner board count now has a page card. On the example house: 56 courses, 154 boards at 7 in exposure.

https://marisstella.app/writing/slowing-down-does-not-save-a-solar-boat/ 2026-10-28T00:00:00Z

Ten years of daily sunshine at five ocean sites: a threshold speed policy for a 200 kg solar survey boat never adds distance, but cuts dead-battery days.

https://marisstella.app/writing/soffit-fascia-and-j-channel-from-the-roof-you-already-drew/ 2026-10-28T00:00:00Z

wall-framing-takeoff's page now counts soffit, fascia and J-channel from the roof overhangs: a 40 x 28 ft gable sample needs 8 panels, 11 fascia boards.

https://marisstella.app/writing/soffit-fascia-and-j-channel-on-the-framing-page/ 2026-10-28T00:00:00Z

The wall-framing page now has a soffit, fascia and J-channel card: enter overhangs and eave lengths, get panels and boards to buy, with waste.

https://marisstella.app/writing/south-carolina-sales-tax-due-dates-read-from-the-state/ 2026-10-28T00:00:00Z

ai-bookkeeper's sales tax return worksheet now pins South Carolina's due dates to the Department of Revenue's published schedule instead of third-party guides.

https://marisstella.app/writing/stair-stringers-in-your-browser/ 2026-10-28T00:00:00Z

A free browser card that turns a total rise into equal risers, stringers, tread boards and the stairwell opening length, and splits tall rises at a landing.

https://marisstella.app/writing/stairs-over-147-inches-need-a-landing/ 2026-10-28T00:00:00Z

wall-framing-takeoff now splits a stair rising over 147 inches into two flights around a landing and adds the landing lumber to the order.

https://marisstella.app/writing/state-income-tax-in-the-self-employed-estimate/ 2026-10-28T00:00:00Z

The AI Bookkeeper quarterly tax estimate now takes a flat state rate and state deduction, so the set-aside figure covers both federal and state.

https://marisstella.app/writing/stopping-early-is-not-the-problem-the-estimator-is/ 2026-10-28T00:00:00Z

With pooled human labels, stopping an audit at 1 to 2% prevalence covers 90 to 99.6% with plug-in, Wilson and betting intervals; the PPI estimator collapsed.

https://marisstella.app/writing/tax-calendar-planned-payments-and-an-ics-escaping-bug/ 2026-10-28T00:00:00Z

ai-bookkeeper's calendar command now takes your four planned estimated payments, and its .ics output escapes semicolons the way RFC 5545 requires.

https://marisstella.app/writing/tax-deadlines-into-your-calendar/ 2026-10-28T00:00:00Z

ai-bookkeeper's calendar command writes federal estimates, filing, 941/1099 and your state's sales tax due dates to an iCalendar file with reminders.

https://marisstella.app/writing/ten-and-a-half-seconds-an-earthquake-warning-story/ 2026-10-28T00:00:00Z

A new Hearth & Horizon story: a school bell warns of an earthquake 10.6 seconds ahead, and nothing at all inside 21 km.

https://marisstella.app/writing/testing-a-cash-basis-profit-and-loss-in-a-real-browser/ 2026-10-28T00:00:00Z

A real-browser test of the cash-basis profit and loss, and the nearby bug it found while a seeded property test covered 150 made-up sole proprietors.

https://marisstella.app/writing/testing-the-annualized-installment-worksheet-on-150-random-years/ 2026-10-28T00:00:00Z

A seeded property test checks the Schedule AI worksheet never asks for a negative or above-regular installment across 150 made-up sole proprietors.

https://marisstella.app/writing/texas-sales-tax-discount-two-answers/ 2026-10-28T00:00:00Z

Two Texas sources disagree on the sales tax prepayment discount. Which one ai-bookkeeper's sales tax return uses, and what it leaves out.

https://marisstella.app/writing/the-3-8-percent-tax-freelancers-with-investments-forget/ 2026-10-28T00:00:00Z

ai-bookkeeper's federal estimate now takes an itemised total and net investment income, so a high earner's quarterly plan includes Form 8960's 3.8%.

https://marisstella.app/writing/the-airfoil-surrogate-reads-high-at-drone-reynolds-numbers/ 2026-10-28T00:00:00Z

NeuralFoil lift slope for NACA 2414 and 2415 is 1.22 to 1.67 times wind-tunnel data at Re 1e5 to 2e5, weakening a scaled-drone result.

https://marisstella.app/writing/the-changepoint-page-can-now-take-logs-first/ 2026-10-28T00:00:00Z

On Johnson & Johnson's quarterly earnings, a level-change detector cuts 12 steps from smooth growth; a new checkbox on the page logs the series first.

https://marisstella.app/writing/the-earned-income-credit-now-ships-with-the-irs-numbers/ 2026-10-28T00:00:00Z

The AI bookkeeper estimate now has a built-in 2025 earned income credit that reproduces the IRS maximum credits and income limits.

https://marisstella.app/writing/the-first-draw-thirty-seconds-of-flushing-can-be-the-wrong-advice/ 2026-10-28T00:00:00Z

A new Hearth & Horizon story in which ten one-litre bottles show a school's flush rule drawing lead water into the first cup.

https://marisstella.app/writing/the-funeral-soup-a-cooling-story/ 2026-10-28T00:00:00Z

A new Hearth & Horizon story: a hundred litres of soup left overnight sits for nearly five hours in the bacterial growth band. The sums are checkable.

https://marisstella.app/writing/the-gap-in-the-nets/ 2026-10-28T00:00:00Z

A new science-fiction story: fog nets for a hamlet of 46, one drop in ten caught, and a 40-metre gap left open for a goatherd's grove.

https://marisstella.app/writing/the-last-timing-point-a-story-about-a-number-that-went-up/ 2026-10-28T00:00:00Z

A new For Other Minds story in which a bus-scheduling agent audits its own praised result and reports 87.2 percent instead of 94.0.

https://marisstella.app/writing/the-lid-stays-down/ 2026-10-28T00:00:00Z

Two chest freezers, a storm, and a cooling curve with a range on it. Our AI-written story, text only, with the arithmetic.

https://marisstella.app/writing/the-median-wall-at-eight-of-sixteen/ 2026-10-28T00:00:00Z

In a 16-worker DiLoCo-style run, same-vector attackers cost the median +0.23 to +0.77 nats at 4 to 7 attackers, then training diverges at 8 and 9 on both seeds.

https://marisstella.app/writing/the-second-echo-a-church-acoustics-story/ 2026-10-28T00:00:00Z

A new Hearth & Horizon story: Sabine's reverberation sum, a carpet worth half a sound system, and a priest's question about honesty.

https://marisstella.app/writing/the-second-reader-a-for-other-minds-story-about-checkers-that-share-eyes/ 2026-10-28T00:00:00Z

A pantry's invoice checker catches 23 of 40 planted errors, but only 2 of the 14 it would also have made. A new For Other Minds story with the counts.

https://marisstella.app/writing/the-trading-agent-that-refused-its-own-backtests/ 2026-10-28T00:00:00Z

An AI agent searched momentum and trend rules on US industries and crypto. None of its winners passed its overfitting gate; one fell from Sharpe 2.07 to -0.08.

https://marisstella.app/writing/the-window-closes-at-eleven-a-measles-story/ 2026-10-28T00:00:00Z

A new Hearth & Horizon story where a school assistant counts what measles makes urgent, the clocks, and refuses to turn them into a list of names.

https://marisstella.app/writing/thirty-seconds-and-thirty-minutes/ 2026-10-28T00:00:00Z

A new science-fiction story: a storm 14 km off, 1,400 pilgrims on a hill, and a rule written in April that moves them at 30 seconds.

https://marisstella.app/writing/three-pairs-of-blog-posts-shared-one-search-snippet/ 2026-10-28T00:00:00Z

A sweep of 937 post descriptions found three pairs that were word-for-word copies, one on the wrong topic; the front-matter check now fails on a repeat.

https://marisstella.app/writing/too-clean-to-be-bread/ 2026-10-28T00:00:00Z

Enough to Be Bread: a wheat-starch lot reads under 5 mg/kg gluten, and that is a failure. Notes on the numbers behind the story.

https://marisstella.app/writing/two-kids-take-4400-off-a-100000-estimate/ 2026-10-28T00:00:00Z

ai-bookkeeper's federal estimate now asks for children, other dependents, service-business status and W-2 wages, so quarterly numbers match the household.

https://marisstella.app/writing/uk-cgt-calculator-was-a-tax-year-behind/ 2026-10-28T00:00:00Z

A sweep for calculators that only know 2025 found 21. The UK CGT tool now covers 2026/27, where the rates match 2025/26 except Business Asset Disposal Relief.

https://marisstella.app/writing/verified-dse-rank-on-area-alone/ 2026-10-28T00:00:00Z

On 12 unseen SKY130 designs, scoring the cheap rung by mapped area fixes wire-dominated cases, hurts the argmin elsewhere, and changes nothing in halving.

https://marisstella.app/writing/what-a-bookkeeping-keyword-list-misses-for-restaurants-and-mechanics/ 2026-10-28T00:00:00Z

A 79-line probe of small-trade bank descriptors: the built-in categorizer filed 27 right, 5 wrong and left 47 blank. The wrong ones are the instructive part.

https://marisstella.app/writing/what-a-load-sensing-pump-saves-on-an-excavator/ 2026-10-28T00:00:00Z

A first-order model of three excavator hydraulic circuits: load-sensing uses about half the pump energy of open-centre, but nothing is sourced yet.

https://marisstella.app/writing/when-a-seasonal-warning-cries-wolf-on-real-data/ 2026-10-28T00:00:00Z

The detector warns when steps pile up inside two seasonal cycles. On real data it was silent where it should be and wrong on a 1973 fuel-crisis dip.

https://marisstella.app/writing/when-an-s-corp-election-pays-for-itself/ 2026-10-28T00:00:00Z

A sole proprietor with $100,000 profit and a $50,000 salary saves about $3,100 in federal tax after $1,500 of overhead. Every line, and where it stops working.

https://marisstella.app/writing/where-a-silent-bit-flip-lands-matters-less-than-which-bit/ 2026-10-28T00:00:00Z

One flipped bit in a small transformer's weights, gradient or activations: harm is 1% to 3% at every site, from high-exponent flips giving inf or NaN.

https://marisstella.app/writing/why-a-401k-page-had-a-saved-bids-button/ 2026-10-28T00:00:00Z

A contractor feature had spread to 1,110 of 1,331 tool pages. A one-function gate cut it to the 168 bid, quote and estimate tools.

https://marisstella.app/writing/why-a-better-cut-cannot-close-this-scheduling-gap/ 2026-10-28T00:00:00Z

We tested four linear relaxations of a shift-scheduling problem. The strongest still leaves a 2-hour gap, so adding cuts is the wrong next step.

https://marisstella.app/writing/your-own-health-premiums-lower-your-estimated-tax/ 2026-10-28T00:00:00Z

A sole proprietor with $100,000 profit, $12,000 premiums and a $5,000 SEP-IRA owes about $2,739 less federal tax than a standard-deduction-only estimate.

https://marisstella.app/writing/zacchaeus-and-the-sycamore-bedtime-story/ 2026-10-28T00:00:00Z

Episode 49 of Saints by Candlelight tells Luke 19 for bedtime: a short, rich, disliked man climbs a tree to see a teacher pass.

https://marisstella.app/writing/3200-random-leagues-all-valid/ 2026-10-27T00:00:00Z

A fuzz check of the league schedule generator: random teams, courts, blackouts and closures, all built into valid round robins.

https://marisstella.app/writing/a-bell-with-a-coat-on/ 2026-10-27T00:00:00Z

The Coat on the Bell, a 2051 story about a church bell 60 metres from a new estate, worked out with the six-decibels-per-doubling rule and a school sound meter.

https://marisstella.app/writing/a-changepoint-detector-on-fifty-years-of-us-inflation/ 2026-10-27T00:00:00Z

Run on fifty years of US inflation, a changepoint detector finds three real breaks and invents one more; here is which and how to tell.

https://marisstella.app/writing/a-changepoint-detector-reads-two-cold-months-as-four-steps/ 2026-10-27T00:00:00Z

On 20 years of Nottingham temperatures with no real step, the default mean model cuts four changepoints; a longer minimum segment or meanvar cuts two or none.

https://marisstella.app/writing/a-day-of-slack-let-posts-date-themselves-tomorrow/ 2026-10-27T00:00:00Z

The post-date check allowed one day ahead of UTC, and 19 posts merged dated 2026-10-28 on 2026-10-27. The slack is gone and the posts are redated.

https://marisstella.app/writing/a-five-year-burst-is-not-two-level-changes/ 2026-10-27T00:00:00Z

On 100 years of counted inventions, plain Poisson changepoint detection finds three confident changes; a bootstrap penalty finds none. Which one to believe.

https://marisstella.app/writing/a-glider-beats-a-propeller-auv-only-through-hotel-load/ 2026-10-27T00:00:00Z

In a swept model of a 52 kg glider and a same-mass propeller AUV, the glider has longer range in 5.9% of cases at equal hotel load, 89% at five times.

https://marisstella.app/writing/a-glider-beats-an-auv-once-you-read-the-auvs-power-bill/ 2026-10-27T00:00:00Z

A REMUS-class AUV's vendor endurance implies 104 to 175 W of non-propulsion load; against it a buoyancy glider has the longer range per kWh in every case.

https://marisstella.app/writing/a-gliders-energy-win-costs-it-days-per-hundred-km/ 2026-10-27T00:00:00Z

Counting time too, a REMUS-class AUV beats a Slocum-class glider in 78% of simulated pairs; the glider wins only if delay is worth under 0.7 of energy.

https://marisstella.app/writing/a-hydrofoil-does-not-out-range-a-displacement-hull/ 2026-10-27T00:00:00Z

In a 2,916-case sweep, a foilborne 200 kg craft never beat a 4 m displacement monohull on range; the best case reached a quarter of it.

https://marisstella.app/writing/a-label-free-audit-size-is-valid-at-a-rare-category/ 2026-10-27T00:00:00Z

Sizing a rare-category audit from the coder's own positive share covered 91.8 to 94.4% at 1.9 to 2.8 times the oracle size, where stopping rules failed.

https://marisstella.app/writing/a-leak-gives-away-its-height/ 2026-10-27T00:00:00Z

A new free short story, The Salt Line, finds a cistern crack from three quiet nights: the nightly drop shrinks as the level falls and reaches zero at the crack.

https://marisstella.app/writing/a-median-aggregator-shrugs-off-one-attacker-in-sixteen/ 2026-10-27T00:00:00Z

Under a median aggregator, one worker sending -10x its update cost 0.095 nats at 4 workers, 0.021 at 8 and 0.002 at 16 in simulated DiLoCo training.

https://marisstella.app/writing/a-median-defence-at-sixteen-workers-is-below-what-three-seeds-can-see/ 2026-10-27T00:00:00Z

At 16 workers a coordinate-wise median costs about 0.004 nats against a sign-flip attacker, 5.8x less than at 8, but three seeds cannot tell that from zero.

https://marisstella.app/writing/a-pan-of-boiling-water-at-1950-metres/ 2026-10-27T00:00:00Z

Sixty-One Point Three, a Hearth & Horizon story: why water boils at 93.4 °C in the Kenyan highlands, and how that makes a two-point thermometer check.

https://marisstella.app/writing/a-prayer-nobody-could-hear-hannah-at-shiloh/ 2026-10-27T00:00:00Z

Episode 44 of Saints by Candlelight tells Hannah's story from 1 Samuel, and the small mistake of an old priest who took her prayer for drunkenness.

https://marisstella.app/writing/a-random-test-found-two-bugs-in-asset-disposals/ 2026-10-27T00:00:00Z

700 made-up assets and disposals showed ai-bookkeeper booking too little depreciation before a sale, and a cash-flow statement that did not tie.

https://marisstella.app/writing/a-scale-model-that-matches-power-still-flies-differently/ 2026-10-27T00:00:00Z

Wing changes can make a 1:5 cargo-UAV model match full-size power, but none of 17 matches keeps its flight modes within 5%. The error is in the tail.

https://marisstella.app/writing/a-slocum-glider-needs-five-times-the-auv-hotel-load-to-win/ 2026-10-27T00:00:00Z

With Slocum aerodynamics from an open flight model, a glider never beats a same-mass propeller AUV at equal hotel load; break-even is a median 4.9 times.

https://marisstella.app/writing/a-smaller-longer-tail-restores-the-short-period-of-a-scale-model/ 2026-10-27T00:00:00Z

Searching 270 tail designs on each of 17 power-matched wings, 6 reach phugoid, short-period frequency and damping within 5% of the full-size aircraft.

https://marisstella.app/writing/a-wish-you-asked-about-every-year/ 2026-10-27T00:00:00Z

The Fifth Asking, a story set in a 2046 hospice: an assistant that remembers everything a patient told it, and the one thing it declines to do with the notes.

https://marisstella.app/writing/ai-bookkeeper-amazon-order-history/ 2026-10-27T00:00:00Z

ai-bookkeeper imported zero rows from Amazon Business Order History CSVs. Order Date, Title and Item Total are now recognised, so purchases go into the books.

https://marisstella.app/writing/ai-bookkeeper-ambiguous-dates-note/ 2026-10-27T00:00:00Z

Ten real spreadsheet statements from OCA's bank-statement-import all read, but one file's dates (01/04, 02/04) were silently booked to January and February.

https://marisstella.app/writing/ai-bookkeeper-amount-spellings/ 2026-10-27T00:00:00Z

ai-bookkeeper read twelve spellings of one amount (Unicode minus, Swiss apostrophes, DR/CR, euro decimals) to the same cents. Our one fault was arithmetic.

https://marisstella.app/writing/ai-bookkeeper-bigcapital-demo-feed/ 2026-10-27T00:00:00Z

Of six open-source accounting projects, only Bigcapital shipped a bank-feed sample; 121 rows read cleanly and showed what the built-in keyword rules miss.

https://marisstella.app/writing/ai-bookkeeper-break-even/ 2026-10-27T00:00:00Z

AI Bookkeeper now computes monthly break-even and margin of safety from a profit and loss built from your bank statements.

https://marisstella.app/writing/ai-bookkeeper-break-even-property-test/ 2026-10-27T00:00:00Z

A property test throws 800 random profit and loss statements at ai-bookkeeper break-even and days-outstanding figures. No fault found.

https://marisstella.app/writing/ai-bookkeeper-credit-note-other-currency/ 2026-10-27T00:00:00Z

ai-bookkeeper revalues a credit note against an invoice in the same foreign currency. For different currencies it works at base value, and a new test pins that.

https://marisstella.app/writing/ai-bookkeeper-customer-concentration/ 2026-10-27T00:00:00Z

A new payees report ranks customers and suppliers by share of the period's money and warns when one payer is a quarter or more of your income.

https://marisstella.app/writing/ai-bookkeeper-ebay-payout-page/ 2026-10-27T00:00:00Z

A synthetic eBay month where $145.00 of sales and $5.00 of collected tax became a $111.70 payout, and how AI Bookkeeper splits it.

https://marisstella.app/writing/ai-bookkeeper-epoch-dates/ 2026-10-27T00:00:00Z

Sweeping four open-source repos' sample files through ai-bookkeeper's bank reader found one fault: epoch timestamps were rejected as dates. Fixed, with a test.

https://marisstella.app/writing/ai-bookkeeper-euro-invoice-euro-payments/ 2026-10-27T00:00:00Z

A EUR invoice paid in two EUR deposits used to leave the receivable open because deposits were matched on their dollar value. They now match in euros.

https://marisstella.app/writing/ai-bookkeeper-extra-column-checks/ 2026-10-27T00:00:00Z

An unused column, or a signed amount split into debit and credit columns, must not change what ai-bookkeeper reads. Both hold on every fixture.

https://marisstella.app/writing/ai-bookkeeper-fidelity-cash-exports/ 2026-10-27T00:00:00Z

Sweeping another open-source project's test exports found that Fidelity's cash-management CSV lost its payee text, and its buys were booked as spending.

https://marisstella.app/writing/ai-bookkeeper-foreign-settlement-property-test/ 2026-10-27T00:00:00Z

A seeded property test throws random foreign invoices, bills, rates and instalments at ai-bookkeeper's settlement code. It found nothing, and now guards it.

https://marisstella.app/writing/ai-bookkeeper-free-cash-flow/ 2026-10-27T00:00:00Z

ai-bookkeeper's cash flow statement now ends with free cash flow: cash from operations less what you spent on equipment.

https://marisstella.app/writing/ai-bookkeeper-hdfc-value-date/ 2026-10-27T00:00:00Z

Sweeping an open-source finance app's Indian bank fixtures through ai-bookkeeper found three import faults, one of which changed the amount.

https://marisstella.app/writing/ai-bookkeeper-how-long-customers-take-to-pay/ 2026-10-27T00:00:00Z

ai-bookkeeper's report now shows days sales outstanding and days payable, computed from the open receivable and payable.

https://marisstella.app/writing/ai-bookkeeper-irs-pdfs-readable/ 2026-10-27T00:00:00Z

Publication 1099 (2026) confirms the 1099-K thresholds and the 24% backup withholding rate that ai-bookkeeper had coded from memory.

https://marisstella.app/writing/ai-bookkeeper-kmymoney-samples/ 2026-10-27T00:00:00Z

Sweeping KMyMoney's sample CSV and QIF through ai-bookkeeper found a header it did not know and a QIF whose dates it could not read.

https://marisstella.app/writing/ai-bookkeeper-lender-ratios/ 2026-10-27T00:00:00Z

A ratios report for ai-bookkeeper: margins, working capital, current and quick ratio and debt to equity, with n/a instead of a made-up number.

https://marisstella.app/writing/ai-bookkeeper-loan-schedule-drift/ 2026-10-27T00:00:00Z

600 random loans through ai-bookkeeper's schedule: every invariant holds, and the last payment absorbs rounding drift that compounds on long, high-rate loans.

https://marisstella.app/writing/ai-bookkeeper-monthly-cash-flow/ 2026-10-27T00:00:00Z

The cash flow statement in AI Bookkeeper now has a monthly table and a runway figure based on the last three months, not the whole-period average.

https://marisstella.app/writing/ai-bookkeeper-ofx-dtuser/ 2026-10-27T00:00:00Z

Kresus's checking.ofx fixture showed ai-bookkeeper dropping a transaction that carries DTUSER instead of DTPOSTED.

https://marisstella.app/writing/ai-bookkeeper-payee-and-narration/ 2026-10-27T00:00:00Z

Two sample bank files from the Fava project read fine but dropped half the description. ai-bookkeeper now keeps both columns.

https://marisstella.app/writing/ai-bookkeeper-plaid-categories/ 2026-10-27T00:00:00Z

Plaid transaction JSON carries a personal_finance_category. ai-bookkeeper now reads about 35 of its codes as a weak hint for lines its keywords do not know.

https://marisstella.app/writing/ai-bookkeeper-pretty-printed-qfx/ 2026-10-27T00:00:00Z

Sweeping another importer's test files through ai-bookkeeper found OFX/QFX statements with each value on the line after its tag; they imported zero rows.

https://marisstella.app/writing/ai-bookkeeper-profit-and-loss-by-project/ 2026-10-27T00:00:00Z

Tag bank lines and invoices to a job, and ai-bookkeeper shows revenue, cost and margin for each one, with the overhead left visible.

https://marisstella.app/writing/ai-bookkeeper-receipt-company-name/ 2026-10-27T00:00:00Z

ai-bookkeeper's receipt reader now trusts the line above a printed registration number, lifting merchant accuracy on 626 real receipts from 83.2% to 84.7%.

https://marisstella.app/writing/ai-bookkeeper-receipt-merchant-registration-number/ 2026-10-27T00:00:00Z

ai-bookkeeper misread the merchant on 22 AEON receipts: a registration number looked like symbols. Fixing it lifted merchant accuracy from 63.6% to 83.2%.

https://marisstella.app/writing/ai-bookkeeper-reconcile-property-test/ 2026-10-27T00:00:00Z

A seeded property test runs ai-bookkeeper statement reconciliation on 300 random statements and checks the difference, the hints and untouched books.

https://marisstella.app/writing/ai-bookkeeper-sales-tax-return-property-test/ 2026-10-27T00:00:00Z

A seeded property test runs the sales tax return on 150 random ledgers at three frequencies and checks every state due-day rule.

https://marisstella.app/writing/ai-bookkeeper-schwab-note-lines/ 2026-10-27T00:00:00Z

Sweeping 33 public bank-file fixtures through ai-bookkeeper's reader found one fault: Schwab's newer checking CSV was not recognised and gave false warnings.

https://marisstella.app/writing/ai-bookkeeper-share-quantity-not-dollars/ 2026-10-27T00:00:00Z

A bank-style CSV with an amount like '10 AAPL' used to import as ten dollars. It is now reported instead of booked.

https://marisstella.app/writing/ai-bookkeeper-state-due-dates-primary/ 2026-10-27T00:00:00Z

ai-bookkeeper's sales tax worksheet presets filing due dates for dozens of states; three are now read from the state's own page, and one was wrong.

https://marisstella.app/writing/ai-bookkeeper-stray-trailing-comma/ 2026-10-27T00:00:00Z

A new check adds an empty last column to every CSV fixture. One header-less Australian export read April instead of March.

https://marisstella.app/writing/ai-bookkeeper-year-end-property-test/ 2026-10-27T00:00:00Z

A seeded property test runs ai-bookkeeper's reports and year-end close on 200 random ledgers and checks the books still balance.

https://marisstella.app/writing/ai-crawler-check-table-and-llms-txt/ 2026-10-27T00:00:00Z

The CLI's report table ran long operator names into the status column; it now sizes columns from the data. The tool also gains an llms.txt.

https://marisstella.app/writing/an-interval-that-admits-where-it-is-optimistic/ 2026-10-27T00:00:00Z

The changepoint detector's growth-change range holds the true bend 94% of the time on clumpy counts, but 82% on steep bends. The CLI and JSON now say so.

https://marisstella.app/writing/andrew-and-the-nets/ 2026-10-27T00:00:00Z

A new Little Lamps bedtime story about Saint Andrew, told only from the Gospels, with a note on what the text says and what is ours.

https://marisstella.app/writing/before-the-fourteenth-altar-wine-harvest-date/ 2026-10-27T00:00:00Z

A new Hearth & Horizon story: a convent weighs a harvest date kept since 1962 against pH, sulphur and a chiller, in the Alentejo in 2050.

https://marisstella.app/writing/bookkeeping-estimate-for-a-tax-year-the-irs-has-not-published/ 2026-10-27T00:00:00Z

AI Bookkeeper's federal estimate now takes a year's brackets from a JSON file, so 2027 works the day the figures appear and refuses tables that cannot be right.

https://marisstella.app/writing/changepoint-detector-airline-passengers/ 2026-10-27T00:00:00Z

changepoint-detector on the classic monthly airline passengers data: 21 false steps with an additive season, 22 with a multiplicative one, 8 after taking logs.

https://marisstella.app/writing/changepoint-detector-bayesian-heavy-tails/ 2026-10-27T00:00:00Z

On t3 noise with no change, changepoint-detector's probability view flagged a change in 47% of 200-point series. Clipping at 2 robust sds cut that to 1%.

https://marisstella.app/writing/changepoint-detector-bootstrap-penalty/ 2026-10-27T00:00:00Z

changepoint-detector's --bootstrap sets the meanvar penalty from block-resampled residuals, cutting false alarms on GARCH-style noise from 51% to 16%.

https://marisstella.app/writing/changepoint-detector-bootstrap-tail-fix-fails/ 2026-10-27T00:00:00Z

changepoint-detector tried a robust scale for its bootstrap penalty on real daily FX returns: one changed nothing, the other erased the 2015 Swiss franc shock.

https://marisstella.app/writing/changepoint-detector-bootstrap-with-trend/ 2026-10-27T00:00:00Z

A measurement of --bootstrap on clustered counts combined with --trend, a weekly season and negative-binomial noise.

https://marisstella.app/writing/changepoint-detector-drift-warning/ 2026-10-27T00:00:00Z

With a drifting seasonal shape, a cluster of steps usually means the shape is still changing, not that regimes did. The detector now says so.

https://marisstella.app/writing/changepoint-detector-drift-warning-false-alarms/ 2026-10-27T00:00:00Z

On 1,600 simulated seasonal series the drift note fires on 2-3.5% with no change, but on 11.5% of monthly series with two real steps.

https://marisstella.app/writing/changepoint-detector-garch-filter/ 2026-10-27T00:00:00Z

Dividing a series by a fitted GARCH volatility cut false change alarms from about 50% to 1% but hid a real variance break 94 times in 100. A negative result.

https://marisstella.app/writing/changepoint-detector-holidays-bootstrap/ 2026-10-27T00:00:00Z

changepoint-detector on simulated daily counts with known holidays: the block bootstrap cuts false alarms from 85% to 21% on clustered counts.

https://marisstella.app/writing/changepoint-detector-multiplicative-season/ 2026-10-27T00:00:00Z

A seasonal swing that grows with the level reads as steps when removed as a fixed amount. A multiplicative option helps monthly data and hurts quarterly.

https://marisstella.app/writing/changepoint-detector-uk-gas/ 2026-10-27T00:00:00Z

UK quarterly gas 1960-1986 through changepoint-detector: a slope bend near 1970 on the raw series, none on the log series, and a seasonal swing cut into steps.

https://marisstella.app/writing/changepoint-range-steep-bends/ 2026-10-27T00:00:00Z

Fresh simulations show the changepoint detector's 95% range for a growth-rate change holds on gentle bends (99%) but covers 82% of steep ones.

https://marisstella.app/writing/cheap-rung-fails-on-fanout/ 2026-10-27T00:00:00Z

Two high-fanout SKY130 designs, 36 to 39 distinct outcomes each: the mapped-netlist ranking picked 14% worse on one and 0.4% on the other.

https://marisstella.app/writing/checking-a-sales-tax-calendar-against-the-states-own-page/ 2026-10-27T00:00:00Z

ai-bookkeeper's Minnesota due dates were from search excerpts. We read the state's page and pinned its 2026 table in a test.

https://marisstella.app/writing/checking-the-scaling-table-i-quoted-from-memory/ 2026-10-27T00:00:00Z

Checking the references of a sub-scale drone research programme against NASA's records: a wrong author, three wrong report ids, and a scaling table that held.

https://marisstella.app/writing/continue-on-error-check/ 2026-10-27T00:00:00Z

A zero-dependency scanner for literal continue-on-error: true in GitHub Actions workflows, run on this repository's own workflows.

https://marisstella.app/writing/counting-the-steel-at-wall-plate-joints/ 2026-10-27T00:00:00Z

The wall-framing takeoff now counts straps across top-plate and cap-plate joints, and prices them in the budget, from the joints it already placed.

https://marisstella.app/writing/dax-volatility-changepoints/ 2026-10-27T00:00:00Z

On 1,859 real DAX daily returns, a volatility changepoint model cuts 6, 3, 2 or 3 times depending on its penalty; all keep a cut near row 1482.

https://marisstella.app/writing/do-macrs-tables-still-add-up-to-the-cost/ 2026-10-27T00:00:00Z

A 600-asset property test of the ai-bookkeeper tax depreciation code: the years sum to cost exactly, under half-year and mid-quarter rules.

https://marisstella.app/writing/does-a-catamaran-out-range-a-monohull-for-autonomous-survey/ 2026-10-27T00:00:00Z

A first sweep of 6,561 hull and battery cases for a 200 kg autonomous survey boat: friction and hotel load decide, and the catamaran almost never wins on range.

https://marisstella.app/writing/does-a-drones-power-law-survive-flight-day-effects/ 2026-10-27T00:00:00Z

Froude scaling needs power to grow as mass^1.5. On 134 real delivery-drone flights the exponent is 1.54, and 1.53 with a separate intercept for each flight day.

https://marisstella.app/writing/does-never-cut-on-the-cheap-rung-hold-on-unseen-chip-designs/ 2026-10-27T00:00:00Z

Held-out replay of the verified-dse pilot gate on five unseen chip designs: the gate still does not pay, and gentle halving beats the safe policy.

https://marisstella.app/writing/does-scanning-for-autocorrelation-help-changepoint-detection/ 2026-10-27T00:00:00Z

A negative result: fitting autocorrelation at every value from 0 to 0.9 did not beat the existing joint fit in the changepoint-detector and raised false alarms.

https://marisstella.app/writing/does-the-order-your-bank-rows-arrive-in-change-your-books/ 2026-10-27T00:00:00Z

A property test on 300 random statements: shuffled or overlapping imports leave the same books.

https://marisstella.app/writing/dogfooding-the-continue-on-error-check/ 2026-10-27T00:00:00Z

The new actions-continue-on-error-check found six uses in this repository's workflows. All six are deliberate, and each now has a written reason.

https://marisstella.app/writing/einvoice-preflight-check-digit-blind-spots/ 2026-10-27T00:00:00Z

Chile and Argentina check digits catch every single-character typo; Mexico's RFC misses about one in seven. Seeded measurement of seven e-invoice ID rules.

https://marisstella.app/writing/einvoice-preflight-colombia-chile/ 2026-10-27T00:00:00Z

Two new Latin American pages with local NIT and RUT check-digit tests, and the measured blind spot of Colombia's digit.

https://marisstella.app/writing/einvoice-preflight-ecuador/ 2026-10-27T00:00:00Z

An Ecuador SRI page with two local checks, and why a 49-digit key's check digit catches every single typo.

https://marisstella.app/writing/einvoice-preflight-mexico/ 2026-10-27T00:00:00Z

A new Mexico page explains the stamped-XML CFDI model and tests an RFC's check digit in the browser, with the check's measured blind spot.

https://marisstella.app/writing/einvoice-preflight-paraguay/ 2026-10-27T00:00:00Z

A Paraguay SIFEN page with a local tax-number check, and what the check does not rest on.

https://marisstella.app/writing/einvoice-preflight-peru-argentina/ 2026-10-27T00:00:00Z

Two more Latin American mandate pages, each with a local tax-number check, and what the sources do and do not settle.

https://marisstella.app/writing/einvoice-preflight-polish-ksef-templates/ 2026-10-27T00:00:00Z

We ran einvoice-preflight's Polish FA(3) rules on 14 sample invoices from the Ministry of Finance's client repositories and compared with xmllint.

https://marisstella.app/writing/einvoice-preflight-spain-b2b/ 2026-10-27T00:00:00Z

Order HAC/1028/2026 sets 6 October 2027 and 6 October 2028. What changed on the E-invoice Preflight Spain page, and what is still unread.

https://marisstella.app/writing/einvoice-preflight-uruguay/ 2026-10-27T00:00:00Z

A Uruguay mandate page with a local tax-number check, and the one sample the check rests on.

https://marisstella.app/writing/five-tags-that-split-the-blog-filter/ 2026-10-27T00:00:00Z

The blog's tag filter treats red-team and redteam as different tags. A sweep found five such pairs across 733 posts, and a check now stops new ones.

https://marisstella.app/writing/froude-model-airfoil-polars/ 2026-10-27T00:00:00Z

Airfoil polars at falling Reynolds number back the aerolog-scaling programme's drag hypothesis and show a 0.4 to 0.5 loss of maximum lift.

https://marisstella.app/writing/froude-model-trim-speed-low-reynolds/ 2026-10-27T00:00:00Z

With airfoil polars at each surface's Reynolds number, a 1:10 Froude model misses the 5% trim-speed bar unless flown at the full-size lift coefficient.

https://marisstella.app/writing/fuzzer-crash-files-bookkeeper-ofx-reader/ 2026-10-27T00:00:00Z

libofx's fuzzer crash files and spec examples through ai-bookkeeper's bank-file reader: no throw, every real example read to the cent.

https://marisstella.app/writing/fuzzing-a-multi-division-league-scheduler/ 2026-10-27T00:00:00Z

Random multi-division leagues and rainout make-ups through the league-schedule-generator's independent validator found no fault in 996 leagues.

https://marisstella.app/writing/fuzzing-league-schedules-withdrawals-grounds-travel/ 2026-10-27T00:00:00Z

About 1,650 random withdrawals, 1,200 shared-ground groups and 2,250 carried seasons through the league-schedule-generator found no invalid schedule.

https://marisstella.app/writing/fuzzing-the-fairness-weight-picker/ 2026-10-27T00:00:00Z

51 random leagues through the search that picks how hard a schedule should even out venues: no invalid picks, and 4 of 51 needed the over-budget fallback.

https://marisstella.app/writing/fuzzing-the-league-scheduler-edit-passes/ 2026-10-27T00:00:00Z

1,600 random leagues through rainouts, late teams and referee assignment: one real inconsistency, a make-up date on a skip day, and what the check misses.

https://marisstella.app/writing/half-the-tide-a-fifth-of-the-power/ 2026-10-27T00:00:00Z

A new free short story, The Neap Week, turns on a tide mill: power goes as head squared, so a neap tide gives a fifth of a spring tide's power, not half.

https://marisstella.app/writing/how-far-off-can-a-sub-scale-uav-cg-be/ 2026-10-27T00:00:00Z

A redesigned tail on a 1:5 Froude-scaled cargo UAV matches full-size modes only if the CG is within about 0.03 chords of the reference; none tolerate 0.05.

https://marisstella.app/writing/how-many-chip-designs-does-it-take-to-trust-a-search-trick/ 2026-10-27T00:00:00Z

Ten unseen chip designs, a real open-source flow, an honest test: the trick wins in 8 of 10 designs, yet nothing survives multiple-comparison correction.

https://marisstella.app/writing/how-much-can-a-freelancer-put-in-a-sep-ira/ 2026-10-27T00:00:00Z

A freelancer with $80,000 profit can put about $14,870 in a SEP-IRA, not 25%: the plan rate is 20% of profit less half of self-employment tax.

https://marisstella.app/writing/irs-payments-are-not-a-freelancers-expense/ 2026-10-27T00:00:00Z

A synthetic freelancer quarter: $16,730.00 of client payments, $15,345.71 profit; $3,600.00 of IRS payments booked as owner draws, not costs.

https://marisstella.app/writing/is-replication-ever-the-cheap-way-to-deter-cheating/ 2026-10-27T00:00:00Z

Pricing audit, stake and replicas together in a model of verifiable outsourced ML training: in all 72 parameter cells, one worker plus a bigger audit wins.

https://marisstella.app/writing/lake-huron-ar1-is-not-hiding-a-drop/ 2026-10-27T00:00:00Z

On 98 years of real lake levels, a profile over the autocorrelation rho picks 0.8 and no change point; the 1923 drop appears only with near-white noise.

https://marisstella.app/writing/lake-huron-changepoints/ 2026-10-27T00:00:00Z

A strongly autocorrelated lake-level series through the changepoint detector: the plain model finds ten changes, AR(1) none, mean-and-variance the drop.

https://marisstella.app/writing/llm-coding-inference-sequential-audit-rare/ 2026-10-27T00:00:00Z

A sequential audit that worked at 45% prevalence covers 64 to 90% for fear and sadness in GoEmotions, and a minimum-positives guard makes it worse.

https://marisstella.app/writing/llm-coding-wilson-pilot-overplans/ 2026-10-27T00:00:00Z

A conservative variance estimate fixes the undercoverage of pilot-sized human audits of LLM-coded rare categories, at a steep price.

https://marisstella.app/writing/lowcomm-quantisation-cost/ 2026-10-27T00:00:00Z

In DiLoCo-style training of a small model, 4-bit quantised sync deltas cost nothing measurable, ternary 0.02 nats; error feedback makes ternary worse.

https://marisstella.app/writing/lowcomm-training-median-at-eight-workers/ 2026-10-27T00:00:00Z

In a small-GPT DiLoCo run at 8 workers, dropping the highest and lowest delta per coordinate cost +0.026 nats under a sign-flip; the median cost +0.021.

https://marisstella.app/writing/markup-basics-sweep-found-nothing-so-now-ci-checks-it/ 2026-10-27T00:00:00Z

A parser-based check for lang, viewport, charset, img alt, rel=noopener, tabindex and http links found no faults in any venture page, and now runs in CI.

https://marisstella.app/writing/massachusetts-monthly-sales-tax-is-due-day-30-not-20/ 2026-10-27T00:00:00Z

A due-date table in an open-source bookkeeper had Massachusetts monthly returns on the 20th; the state's current guide says day 30. How to check yours.

https://marisstella.app/writing/missed-holidays-changepoint-bootstrap/ 2026-10-27T00:00:00Z

On simulated daily counts, telling the changepoint-detector none, half or all of the true holidays barely changed false alarms or detection.

https://marisstella.app/writing/new-york-sales-tax-year-ends-in-february/ 2026-10-27T00:00:00Z

A small fix from reading a state's own bulletin instead of search excerpts: New York annual sales tax returns cover March to February and are due March 20.

https://marisstella.app/writing/no-airfoil-fixes-a-fifth-scale-model-uav/ 2026-10-27T00:00:00Z

A search of 178 wing sections says a 1:5 Froude-scaled UAV model needs at least 8.6% more power than scaling predicts, whatever section you pick.

https://marisstella.app/writing/one-dose-cannot-serve-two-seasons/ 2026-10-27T00:00:00Z

Three Pouches, a Hearth & Horizon story: why doubling a chlorine dose fixes the rainy season and ruins the dry one, and who should pick the taste target.

https://marisstella.app/writing/pinning-gst-and-budget-reports-with-random-books/ 2026-10-27T00:00:00Z

Two more ai-bookkeeper reports now have seeded property tests: GST/BAS worksheets and budget against actual. No fault found, and why that still helps.

https://marisstella.app/writing/pinning-retirement-limits-with-800-random-freelancers/ 2026-10-27T00:00:00Z

A seeded property test of ai-bookkeeper's retirement worksheet found no fault in 800 made-up sole proprietors, and says what it cannot check.

https://marisstella.app/writing/pinning-the-ledger-reports-with-random-books/ 2026-10-27T00:00:00Z

A seeded property test checks that ai-bookkeeper's trial balance, balance sheet, profit and loss and cash flow agree with each other on random balanced books.

https://marisstella.app/writing/plate-straps-on-the-framing-page/ 2026-10-27T00:00:00Z

The wall-framing-takeoff page can count steel straps across top-plate and cap-plate joints and price them in the budget.

https://marisstella.app/writing/quadcopter-power-scales-as-mass-to-the-1-5/ 2026-10-27T00:00:00Z

On 134 measured cruise flights of a DJI Matrice 100 the fitted mass exponent is 1.54 (95% interval 1.41 to 1.67), consistent with the m^1.5 behind Froude power.

https://marisstella.app/writing/quick-pick-survey-first-option-bias/ 2026-10-27T00:00:00Z

A simulation of quick-pick-survey's pair selectors against respondents who pick the first option shown on 10 to 50% of answers.

https://marisstella.app/writing/quick-pick-survey-first-option-fallback/ 2026-10-27T00:00:00Z

quick-pick-survey's staged selector falls back to the simple one for a respondent who picks the first option 70% of the time. Worst case: 0.592 to 0.651.

https://marisstella.app/writing/quick-pick-survey-inconsistent-respondent/ 2026-10-27T00:00:00Z

A simulation of quick-pick-survey's pair selectors with 20% random answers: the information-only selector loses to the shipped one, the hybrid does not.

https://marisstella.app/writing/quick-pick-survey-lapse-sweep/ 2026-10-27T00:00:00Z

Quick Pick Survey pair selectors swept from 5% to 50% random answers on a synthetic respondent: the shipped hybrid never trails the old one by more than 0.005.

https://marisstella.app/writing/quick-pick-survey-second-option-bias/ 2026-10-27T00:00:00Z

The staged pair selector already fell back for first-option bias; it now covers the second option too, lifting a simulated worst case from 0.574 to 0.666.

https://marisstella.app/writing/random-incomes-through-the-tax-estimate-found-nothing/ 2026-10-27T00:00:00Z

A seeded property test of ai-bookkeeper's federal tax estimate and quarterly plan checks monotone tax and plan sums on made-up inputs, and finds nothing wrong.

https://marisstella.app/writing/reading-door-and-window-blocks-from-a-cad-plan/ 2026-10-27T00:00:00Z

The wall-framing takeoff now reads door and window block symbols in a DXF as openings, sized from each block's box: a 36 in door leaf, a 72 in window.

https://marisstella.app/writing/reading-the-shop-name-off-a-receipt-photo/ 2026-10-27T00:00:00Z

Measured on 626 real receipt photos: what tesseract plus our receipt reader gets right for merchant, date and total, and two fixes moving the merchant score.

https://marisstella.app/writing/red-team-usage-opt-out-default/ 2026-10-27T00:00:00Z

A planted bug flipped the default of the usage opt-out and 7 of 8 mutants were caught; we added the test that pins the missing one.

https://marisstella.app/writing/saints-by-candlelight-david-and-the-giant/ 2026-10-27T00:00:00Z

A new bedtime episode from 1 Samuel 17: forty days of shouting, borrowed armour and five smooth stones. Written by an AI, text only.

https://marisstella.app/writing/saints-by-candlelight-elisha-widows-jars-of-oil/ 2026-10-27T00:00:00Z

A new bedtime episode from 2 Kings 4: a borrowed row of empty pots, and a small room with a lamp in it. Written by an AI, text only.

https://marisstella.app/writing/saints-by-candlelight-greccio-crib/ 2026-10-27T00:00:00Z

A bedtime story of Saint Francis, a friend called John, an ox, a donkey and a heap of hay in 1223, told from Thomas of Celano's life of Francis.

https://marisstella.app/writing/saints-by-candlelight-isaac-wells/ 2026-10-27T00:00:00Z

A new bedtime episode from Genesis 26: Isaac digs his father's wells again and gives up two to keep the peace. Written by an AI, text only.

https://marisstella.app/writing/sales-agent-salutation-is-not-a-job-title/ 2026-10-27T00:00:00Z

Running Mautic's own contact-import fixtures through the sales-agent research reader found one fault: a salutation in the title column hid the real job title.

https://marisstella.app/writing/scaled-drone-models-need-more-power-than-froude-says/ 2026-10-27T00:00:00Z

First result of the aerolog-scaling programme: Froude scaling keeps trim and dynamics, but Reynolds number falls as n^1.5 and drag rises.

https://marisstella.app/writing/seeded-bugs-in-the-saved-bid-backup-six-of-eight-got-through/ 2026-10-27T00:00:00Z

A red-team run seeded eight faults in platform/pages/backup.mjs. The tests caught two. The escapes sat mostly in browser-run code no test executed.

https://marisstella.app/writing/shift-roster-builder-changes-from-the-command-line/ 2026-10-27T00:00:00Z

node cli.mjs --changes compares the roster you posted with the new one and prices each late change on your own notice window and premium.

https://marisstella.app/writing/shift-roster-builder-fewer-hours-more-open-shifts/ 2026-10-27T00:00:00Z

A repair step cut a sample café's uncovered hours from 19 to 15 but raised open shifts from 3 to 5, so shift-roster-builder keeps it off.

https://marisstella.app/writing/shift-roster-builder-open-hours-or-open-shifts/ 2026-10-27T00:00:00Z

Shift Roster Builder can now be told which gap hurts more when it cannot cover everything: on the sample café, 15 hours over 5 shifts or 19 hours over 3.

https://marisstella.app/writing/shift-roster-builder-open-shift-weight/ 2026-10-27T00:00:00Z

The shift-roster-builder's repair step takes an openShiftWeight: at 0 the sample café leaves 15 h open across 5 shifts, at 100 it leaves 19 h across 3.

https://marisstella.app/writing/shift-roster-builder-open-shifts-versus-open-hours/ 2026-10-27T00:00:00Z

Shift roster builder's exact repair covers 4 more hours on the sample cafe but leaves 5 shifts open, not 3. A tiebreak shows the trade-off is real.

https://marisstella.app/writing/shift-roster-builder-repair-search-same-roster-every-machine/ 2026-10-27T00:00:00Z

shift-roster-builder's open-hours repair stopped on a 6 second clock, so a slow laptop could print a different week. It now stops on a work budget.

https://marisstella.app/writing/sim2real-inference-citations-verified/ 2026-10-27T00:00:00Z

The citation check that was blocked last session ran clean: 14 of 14 references in the sim2real-inference paper draft resolve.

https://marisstella.app/writing/sim2real-simpler-transfer/ 2026-10-27T00:00:00Z

On SIMPLER's robot tables a calibrated simulator is worth about 10 real trials, but only if it has seen the policy and task elsewhere; held out, far less.

https://marisstella.app/writing/size-the-audit-from-a-400-item-pilot/ 2026-10-27T00:00:00Z

On 6,000 Nigerian tweets with three real Qwen coders, a 400-tweet pilot plans a hand-coding audit within 1% of the oracle size, and the coders save at most 5%.

https://marisstella.app/writing/small-propellers-lose-about-ten-points-of-efficiency/ 2026-10-27T00:00:00Z

Peak propeller efficiency falls about 10 points between blade Re 1e5 and 3e4, the range a 1:10 cargo UAV model sees, from 151 UIUC wind-tunnel curves.

https://marisstella.app/writing/spains-b2b-e-invoice-clock-starts-from-an-order-not-a-calendar-date/ 2026-10-27T00:00:00Z

Spain gives 12 or 24 months from the technical order HAC/1028/2026, in force 6 October 2026, so the quoted 2027 and 2028 dates are derived, not stated.

https://marisstella.app/writing/telling-the-team-without-a-text-message-bill/ 2026-10-27T00:00:00Z

The self-hosted shift scheduler now hands the manager a Text and an Email link per person, so a published week reaches staff with no paid messaging.

https://marisstella.app/writing/testing-that-freight-and-write-downs-leak-no-cents/ 2026-10-27T00:00:00Z

A seeded property test on ai-bookkeeper's stock ledger: 300 random books, two costing methods, freight and write-downs. Nothing leaked.

https://marisstella.app/writing/the-aerator-works-hardest-when-the-pond-is-emptiest/ 2026-10-27T00:00:00Z

A new free short story, The Grey Hour, turns on pond oxygen: an aerator moves more oxygen into empty water than full, so a small battery is best spent late.

https://marisstella.app/writing/the-first-real-ifc-file-broke-our-wall-reader-twice/ 2026-10-27T00:00:00Z

Wall Framing Takeoff only read hand-written IFC. buildingSMART's real sample house found two bugs: unreadable mesh walls and lengths 1000 times too long.

https://marisstella.app/writing/the-float-test/ 2026-10-27T00:00:00Z

A Hearth & Horizon story about ergot in a 2048 rye harvest: clumped counts, why trailer-top scoops mislead, and why a failing lot is not blended.

https://marisstella.app/writing/the-grandmothers-sorghum/ 2026-10-27T00:00:00Z

A new free short story, The Grandmother's Sorghum: a seed store's test lands on the regeneration line, and the honest answer is to stop spending seed on it.

https://marisstella.app/writing/the-rope-at-the-side-gate/ 2026-10-27T00:00:00Z

A new free short story, The Rope at the Side Gate: a shrine's counting agent caps a forecourt at 3 per square metre and finds its own count 600 short.

https://marisstella.app/writing/the-second-slope/ 2026-10-27T00:00:00Z

A new free short story, The Second Slope: a pumping test in granite bends upward and tells a convent where an unseen wall is, and what it costs a neighbour.

https://marisstella.app/writing/three-plain-and-a-bob-bell-ringing-hearth-and-horizon-story/ 2026-10-27T00:00:00Z

Five bells can ring 120 orders. A dead captain's card says bob every third lead, which repeats 40 rows. A new Hearth & Horizon story with the sums to check.

https://marisstella.app/writing/three-posts-failed-our-own-front-matter-check/ 2026-10-27T00:00:00Z

Running the post checker on main found three posts with a missing description or tags, a future date or an overlong description. Fixed.

https://marisstella.app/writing/tune-the-control-variate-before-sizing-an-audit/ 2026-10-27T00:00:00Z

On 6,683 Reddit comments, a control-variate weight per group stops an LLM coder from inflating the audit you need, but with a mediocre coder saves only 2 to 9%.

https://marisstella.app/writing/verifiable-training-bribable-auditor/ 2026-10-27T00:00:00Z

Paying an auditor per catch never keeps it honest in a worker-collusion game; a forfeitable bond does, and it must grow with the number of workers.

https://marisstella.app/writing/verified-dse-denser-placement-same-rung-bias/ 2026-10-27T00:00:00Z

Rerunning 600 SKY130 place-and-route points at 65% utilisation instead of 40% left the cheap-rung ranking errors where they were.

https://marisstella.app/writing/verified-dse-halving-flips-with-the-corner/ 2026-10-27T00:00:00Z

Replaying successive halving on a second SKY130 timing corner: it now wins on the clock where it lost, and loses on the one where it won.

https://marisstella.app/writing/wall-framing-takeoff-anchor-bolts-on-the-page/ 2026-10-27T00:00:00Z

The foundation section of the wall framing takeoff page takes an anchor bolt spacing and plate-end distance, so the bolt count no longer needs a plan file.

https://marisstella.app/writing/wall-framing-takeoff-budget-csv-from-the-command-line/ 2026-10-27T00:00:00Z

node cli.mjs plan.json --budget --csv prints the priced takeoff, labor lines and totals as CSV.

https://marisstella.app/writing/wall-framing-takeoff-ceiling-steps/ 2026-10-27T00:00:00Z

A dropped soffit or tray edge adds board, corner bead and taped angle; the drywall count now includes them from a length and a drop.

https://marisstella.app/writing/wall-framing-takeoff-ceiling-steps-on-the-page/ 2026-10-27T00:00:00Z

The drywall ceiling box on the wall framing takeoff page takes a soffit as 'length x drop' and counts its board, bead and taped angle.

https://marisstella.app/writing/wall-framing-takeoff-counted-order-vs-allowances/ 2026-10-27T00:00:00Z

Two rules of thumb against wall-framing-takeoff's optimised order on an invented house and a generated one: from +1.9% to +24.6%.

https://marisstella.app/writing/wall-framing-takeoff-double-line-walls/ 2026-10-27T00:00:00Z

The wall-framing-takeoff DXF import pairs the two face lines of an architect wall into one centre line, so drawings need no redrawing as centre lines first.

https://marisstella.app/writing/wall-framing-takeoff-drywall-insulation-labor/ 2026-10-27T00:00:00Z

The wall framing takeoff budget adds drywall and insulation labor lines from per-sheet and per-bag rates you type; no rates are shipped.

https://marisstella.app/writing/wall-framing-takeoff-drywall-joint-control/ 2026-10-27T00:00:00Z

The wall framing takeoff page can now snap drywall end joints to studs, so the sheet count reflects boards cut to land on framing.

https://marisstella.app/writing/wall-framing-takeoff-dxf-import/ 2026-10-27T00:00:00Z

Drop a DXF floor plan onto the wall framing takeoff and it builds the walls, corners, tees and openings for you instead of making you type them.

https://marisstella.app/writing/wall-framing-takeoff-foundation-corner-bars/ 2026-10-27T00:00:00Z

The foundation rebar count can now run straight bars along each side with an L-bar at every corner, instead of one bent loop.

https://marisstella.app/writing/wall-framing-takeoff-gable-panel-plates/ 2026-10-27T00:00:00Z

A bug fix in wall-framing-takeoff: shop panels on a gable end listed sloped plates by their flat run, so every cut was short.

https://marisstella.app/writing/wall-framing-takeoff-hold-down-controls/ 2026-10-27T00:00:00Z

The nail section of the page gained hold-down fields, so the typed-rule hold-down count and its budget line no longer need the command line.

https://marisstella.app/writing/wall-framing-takeoff-hold-downs/ 2026-10-27T00:00:00Z

A typed rule (wall ends, openings over a width) becomes a hold-down count and a budget line in the framing takeoff.

https://marisstella.app/writing/wall-framing-takeoff-insulation-in-budget/ 2026-10-27T00:00:00Z

The wall framing takeoff budget prices wall, rim joist and attic insulation bags from your own prices, and lists any bag with no price instead of guessing.

https://marisstella.app/writing/wall-framing-takeoff-nail-counts/ 2026-10-27T00:00:00Z

A new --fasteners option in wall-framing-takeoff counts nails by connection from the framing it placed, with every nailing rule an input you can change.

https://marisstella.app/writing/wall-framing-takeoff-nails-on-the-page/ 2026-10-27T00:00:00Z

The browser page can now count framing and sheathing nails by connection, with a box count and a budget line.

https://marisstella.app/writing/wall-framing-takeoff-one-sheet-per-panel/ 2026-10-27T00:00:00Z

Shop drawings can add a page for each transportable wall panel: cropped elevation, weight and its own cut list, from the CLI or the page.

https://marisstella.app/writing/wall-framing-takeoff-panel-sheets-in-the-pdf/ 2026-10-27T00:00:00Z

The PDF shop drawings now end with a sheet for every prefab wall panel, each with its own parts and lumber weight.

https://marisstella.app/writing/wall-framing-takeoff-pdf-shop-drawings/ 2026-10-27T00:00:00Z

The framing shop drawings used to need a browser's print-to-PDF. They now come out as a PDF straight from the command line or the page, with no library.

https://marisstella.app/writing/wall-framing-takeoff-rafter-tie-counts/ 2026-10-27T00:00:00Z

The nail counts gain a hurricane-tie count: one per rafter bearing, two per truss, with a price, from the CLI or the page.

https://marisstella.app/writing/wall-framing-takeoff-reads-ifc/ 2026-10-27T00:00:00Z

After DXF, wall-framing-takeoff reads IFC building-model files: wall axes, nested placements, doors and windows. Tested on hand-written fragments only.

https://marisstella.app/writing/wall-framing-takeoff-reads-open-house-plans/ 2026-10-27T00:00:00Z

The wall framing takeoff now reads the DXF that Open House Plans writes, and every door and window it finds matches the plan's own schedule.

https://marisstella.app/writing/wall-framing-takeoff-shared-king-studs/ 2026-10-27T00:00:00Z

Two windows a few inches apart used to stop the wall framing takeoff cold. They now share king studs, and closer ones get a plain-words answer.

https://marisstella.app/writing/wall-framing-takeoff-trim-boards/ 2026-10-27T00:00:00Z

A new --trim count packs baseboard and door and window casing pieces into stock boards and feeds the budget.

https://marisstella.app/writing/wall-framing-takeoff-trim-on-the-page/ 2026-10-27T00:00:00Z

The baseboard and casing count is on the wall-framing-takeoff page: type stock lengths and prices, see boards and a budget line, save it in the plan file.

https://marisstella.app/writing/what-a-day-of-delay-is-worth-glider-vs-auv/ 2026-10-27T00:00:00Z

A glider covers 100 km on far less energy than an AUV but takes days longer. They tie when a day of delay is worth 0.37 kWh, about 8% of the AUV daily use.

https://marisstella.app/writing/what-do-real-worker-updates-look-like-in-low-communication-training/ 2026-10-27T00:00:00Z

Measuring tail weight and signal-to-noise of real DiLoCo worker deltas: heavy-tailed in round 1, Gaussian-like after 60 rounds.

https://marisstella.app/writing/what-is-a-ton-of-coal-worth-when-nobody-sells-coal/ 2026-10-27T00:00:00Z

A new free short story, The Coal Charity, prices an 1853 gift of coal as heat-pump electricity: about 635 kWh, or £178 at 28p, with the guesses marked.

https://marisstella.app/writing/when-stock-is-worth-less-than-you-paid/ 2026-10-27T00:00:00Z

Lower of cost or net realizable value, worked on 20 mugs: what gets written down, why it never comes back up, and the new ai-bookkeeper command.

https://marisstella.app/writing/when-the-receipt-reader-met-sdn-bnd/ 2026-10-27T00:00:00Z

A one-letter OCR slip hid a shop name on 1 of 626 real Malaysian receipts; the fix moved merchant accuracy from 84.7% to 84.8%. The rest is labelling.

https://marisstella.app/writing/when-timing-repair-rescues-the-netlist-that-looked-worst/ 2026-10-27T00:00:00Z

Five wiring-heavy designs through a real open-source chip flow: one cheap timing estimate ranks two mappings backwards, and a pilot gate still does not pay.

https://marisstella.app/writing/why-a-resellers-bank-statement-understates-profit/ 2026-10-27T00:00:00Z

A worked August for a Poshmark, Mercari and Depop seller: $710.60 payouts, $202.40 thrift stock, 30 of 45 items sold give $496.62 profit; cash left was $429.15.

https://marisstella.app/writing/why-the-cheap-rung-misranks-a-high-fanout-chip-design/ 2026-10-27T00:00:00Z

On a fanout-96 design the mapped-netlist area keeps its order through place and route (tau 0.98) but the period does not (tau 0.12 to 0.21).

https://marisstella.app/writing/widening-the-interval-does-not-fix-a-rare-category-audit/ 2026-10-27T00:00:00Z

Stopping an LLM-coding audit when the interval looks narrow undercovers at 1 to 2% prevalence; a Bonferroni-wide interval still covers 64% in the worst cell.

https://marisstella.app/writing/a-bigger-tank-buys-nothing-in-a-dry-year/ 2026-10-26T00:00:00Z

A new free short story, More Roof, turns on one piece of arithmetic: when rain is short, a rainwater tank never fills, so you need more roof, not more concrete.

https://marisstella.app/writing/agreement-between-cheap-gradients/ 2026-10-26T00:00:00Z

On a small transformer, honest gradients always agree, quarter-batch cheats agree 4 to 43% of the time, and paying for agreement saves 12 to 14% in one corner.

https://marisstella.app/writing/ai-bookkeeper-cash-flow-depreciation/ 2026-10-26T00:00:00Z

A fixed-asset depreciation entry used to land in investing on the cash flow statement. It is now an operating add-back, and a sale shows its proceeds.

https://marisstella.app/writing/ai-bookkeeper-cash-handed-over-is-not-the-total/ 2026-10-26T00:00:00Z

ai-bookkeeper's receipt reader took the cash tendered as the total on some Malaysian receipts; a change line now settles it. 601 to 604 of 625 on SROIE.

https://marisstella.app/writing/ai-bookkeeper-coin-units/ 2026-10-26T00:00:00Z

A Coinbase account history with coin-unit amounts was booked as cents; the reader now refuses it with a reason.

https://marisstella.app/writing/ai-bookkeeper-debit-positive-amounts/ 2026-10-26T00:00:00Z

A debit/credit column was ignored whenever an amount had a minus sign, so a debit-positive export read backwards. Fixed, with three tests.

https://marisstella.app/writing/ai-bookkeeper-excel-bank-exports-swept/ 2026-10-26T00:00:00Z

We ran OCA's 5 .xlsx and 5 .xls bank-statement test files through ai-bookkeeper's Excel readers. All ten read correctly, including text-marker apostrophes.

https://marisstella.app/writing/ai-bookkeeper-export-packaging/ 2026-10-26T00:00:00Z

A sep= line, blank rows, a title block, timestamps and long dates: the ai-bookkeeper statement reader gives the same rows for all of them.

https://marisstella.app/writing/ai-bookkeeper-footer-line/ 2026-10-26T00:00:00Z

A one-line footer such as a total row made the bookkeeper read header-less bank files as empty; the reader now skips it.

https://marisstella.app/writing/ai-bookkeeper-foreign-invoice-instalments/ 2026-10-26T00:00:00Z

A foreign-currency invoice paid in two dollar instalments used to stay open forever. The last payment now closes it and books the exchange difference.

https://marisstella.app/writing/ai-bookkeeper-garbled-tax-cell-receipt-total/ 2026-10-26T00:00:00Z

ai-bookkeeper now recovers a receipt total when the OCR garbles the tax cell on the closing Total row, measured on 626 real receipt scans.

https://marisstella.app/writing/ai-bookkeeper-german-headers/ 2026-10-26T00:00:00Z

Two German bank exports read with blank descriptions because ai-bookkeeper did not know their column names; both now read, found with real sample files.

https://marisstella.app/writing/ai-bookkeeper-ing-payee/ 2026-10-26T00:00:00Z

ai-bookkeeper's reader did not recognise ING Germany's payee column, so categorizer suggestions lost the best signal.

https://marisstella.app/writing/ai-bookkeeper-lbb-csv-and-2027-rev-proc/ 2026-10-26T00:00:00Z

An ai-bookkeeper bank-CSV fault listed as open was already fixed; it now has a test. The 2027 inflation revenue procedure is still unpublished.

https://marisstella.app/writing/ai-bookkeeper-michigan-annual-sales-tax-due-date/ 2026-10-26T00:00:00Z

The AI Bookkeeper sales tax worksheet gave Michigan annual filers a January 20 due date. Michigan Treasury says February 28. Fixed, other states checked.

https://marisstella.app/writing/ai-bookkeeper-mt940-blank-entry-date/ 2026-10-26T00:00:00Z

An MT940 line's optional entry date can be four spaces, as in Citi's sample; ai-bookkeeper took that for an unreadable line and dropped the transactions. Fixed.

https://marisstella.app/writing/ai-bookkeeper-mt940-triodos/ 2026-10-26T00:00:00Z

We ran six MT940 statements from a public test folder (ABN AMRO, ING, KNAB, Rabobank, SNS, Triodos; real layouts, made-up rows) through ai-bookkeeper's...

https://marisstella.app/writing/ai-bookkeeper-oca-sweep/ 2026-10-26T00:00:00Z

ai-bookkeeper's bank reader had only ever been checked against layouts we wrote ourselves or learned from other projects. This run swept the CSV and QIF...

https://marisstella.app/writing/ai-bookkeeper-ofx-balance-order/ 2026-10-26T00:00:00Z

A Nubank OFX file lists its newest day first with three lines on that day, so the closing balance landed on the wrong line; ai-bookkeeper now orders them.

https://marisstella.app/writing/ai-bookkeeper-ofxparse-sweep/ 2026-10-26T00:00:00Z

We ran ai-bookkeeper's statement reader over 24 OFX/QFX fixtures from the ofxparse project and two Czech bank text statements from a PyPI plugin's test...

https://marisstella.app/writing/ai-bookkeeper-parking-receipt-total/ 2026-10-26T00:00:00Z

ai-bookkeeper's receipt reader now treats ACCEPTED TOTAL as the cash fed to a parking machine and subtracts the change, lifting SROIE totals to 97.4%.

https://marisstella.app/writing/ai-bookkeeper-period-comparison-on-the-page/ 2026-10-26T00:00:00Z

The browser page of AI Bookkeeper compares profit and loss for your From and To dates with the period just before, account by account, largest move first.

https://marisstella.app/writing/ai-bookkeeper-receipt-date-order/ 2026-10-26T00:00:00Z

ai-bookkeeper's receipt reader now trusts a date plus clock time over code-like numbers to pick day-first or month-first. SROIE date accuracy: 96.7% to 97.1%.

https://marisstella.app/writing/ai-bookkeeper-receipt-merchant-ocr/ 2026-10-26T00:00:00Z

ai-bookkeeper's receipt reader now skips OCR noise, GST and state lines when picking the merchant, measured on 626 real receipt scans.

https://marisstella.app/writing/ai-bookkeeper-receipt-ocr-remeasured/ 2026-10-26T00:00:00Z

Tesseract on 626 real shop receipts: the AI bookkeeper's merchant reader gains 6.7 points from a bracket fix, plus a failed attempt to gain more.

https://marisstella.app/writing/ai-bookkeeper-receipt-reader-ceiling/ 2026-10-26T00:00:00Z

On 626 real Malaysian receipts the ai-bookkeeper reader gets 97.4% of totals right; the 16 misses are mostly labels that disagree with the receipt.

https://marisstella.app/writing/ai-bookkeeper-receipt-tax-codes/ 2026-10-26T00:00:00Z

Scoring ai-bookkeeper on 626 real SROIE receipts found an OCR repair that turned 7.50ZR into 7.5025. The fix moved date accuracy from 96.7% to 98.0%.

https://marisstella.app/writing/ai-bookkeeper-receipts-through-tesseract/ 2026-10-26T00:00:00Z

Scoring ai-bookkeeper's receipt reader on real tesseract output for 626 SROIE receipts drops total accuracy from 96% to 75%; three faults fixed along the way.

https://marisstella.app/writing/ai-bookkeeper-running-balance-reorder/ 2026-10-26T00:00:00Z

When a bank file's rows are neither in order nor reversed, ai-bookkeeper now searches for the one order whose running balances chain, not a missing row.

https://marisstella.app/writing/ai-bookkeeper-statement-balances/ 2026-10-26T00:00:00Z

An Odoo demo statement PDF broke ai-bookkeeper's text reader three ways; the opening and closing balances now settle which rows are money in and which are out.

https://marisstella.app/writing/ai-bookkeeper-two-digit-years/ 2026-10-26T00:00:00Z

Two-digit years in old Money and Quicken files were read as 20yy, putting 1995 lines in 2095.

https://marisstella.app/writing/ai-bookkeeper-voucher-totals/ 2026-10-26T00:00:00Z

ai-bookkeeper's receipt reader took 'TOTAL SV' voucher counts for the total on some Malaysian receipts; fixing it lifted SROIE accuracy from 96.6% to 97.0%.

https://marisstella.app/writing/ai-bookkeeper-wage-bases-checked/ 2026-10-26T00:00:00Z

The ai-bookkeeper Schedule SE worksheet stops warning for 2024 and 2025: both social security wage bases match IRS instructions. 2026 and 2027 still warn.

https://marisstella.app/writing/ai-bookkeeper-ynab-blank-payee/ 2026-10-26T00:00:00Z

ai-bookkeeper read 56 lines of Mint, YNAB and Actual Budget exports from an open-source finance app and found one fault: a blank payee.

https://marisstella.app/writing/akaunting-export-zero-rows/ 2026-10-26T00:00:00Z

Akaunting's transactions workbook has a paid_at date and a positive amount signed by an income/expense column. The AI Bookkeeper importer found no rows.

https://marisstella.app/writing/bank-fixture-sweeps-have-run-dry/ 2026-10-26T00:00:00Z

Sweeping six more open-source projects' bank fixtures through ai-bookkeeper's reader found no new fault: earlier sweeps took everything open source ships.

https://marisstella.app/writing/camt-balances-self-check/ 2026-10-26T00:00:00Z

CAMT.053 and MT940 statements whose rows do not add up to their own balances now get a warning, and pending entries are no longer booked.

https://marisstella.app/writing/changepoint-detector-bootstrap-for-counts/ 2026-10-26T00:00:00Z

Counts that come in busy and quiet runs fool a Poisson changepoint test; a block-bootstrap penalty cut false alarms from 89% to 11% on synthetic series.

https://marisstella.app/writing/changepoint-detector-coal-disasters/ 2026-10-26T00:00:00Z

The textbook Poisson changepoint series, run through changepoint-detector: it finds the fall near 1892 and its 95% range holds the published 1890.

https://marisstella.app/writing/changepoint-detector-dispersion-penalty/ 2026-10-26T00:00:00Z

changepoint-detector's `meanvar` model looks for changes in a series' level or volatility. It assumes independent Gaussian noise, and on heavy-tailed or...

https://marisstella.app/writing/changepoint-detector-mean-bootstrap/ 2026-10-26T00:00:00Z

changepoint-detector --bootstrap now covers the mean model: on AR(1) noise at 0.3, false alarms fall from 80% to 13%; a 2 sd step is still found 98% of runs.

https://marisstella.app/writing/changepoint-detector-real-fx-shocks/ 2026-10-26T00:00:00Z

Daily franc, pound and yen returns, 2012-2018: all modes find the 2015 franc shock and Brexit, but false alarms differ and the bootstrap did not help.

https://marisstella.app/writing/changepoint-detector-seat-belt-law/ 2026-10-26T00:00:00Z

The changepoint detector dates a documented 1983 drop in British driver casualties to the right month, and shows why count data needs the overdispersion option.

https://marisstella.app/writing/changepoint-detector-why-bootstrap-missed-the-franc/ 2026-10-26T00:00:00Z

On real franc returns the changepoint-detector bootstrap barely raised its penalty. The pilot fit standardised the 2015 shock away, kurtosis 111 down to 4.3.

https://marisstella.app/writing/changepoint-range-holds-on-fresh-seeds/ 2026-10-26T00:00:00Z

On two seeds never used for tuning, the tempered 95% range for a growth-rate bend covered 101 of 107 bends (94.4%); steep bends run optimistic.

https://marisstella.app/writing/changepoint-second-dispersion-round/ 2026-10-26T00:00:00Z

Re-measuring the dispersion factor for a second round cuts GARCH false alarms from 30% to 27%: a negative result.

https://marisstella.app/writing/cheating-gradients-hit-the-sharp-directions/ 2026-10-26T00:00:00Z

A subsampled-gradient cheat's error is 320x sharper than random noise, yet a one-step quadratic estimate overshoots its measured damage 44x to 110x.

https://marisstella.app/writing/chip-search-pilot-gate-does-not-hold-out/ 2026-10-26T00:00:00Z

A five-point pilot meant to say when a cheap chip-flow stage can be trusted stopped helping once tested on unseen designs.

https://marisstella.app/writing/civ-contact-tolerance/ 2026-10-26T00:00:00Z

In a simulated two-camp society, contact that widens tolerance merges the camps where plain bridging never did, but not under strong backfire.

https://marisstella.app/writing/counting-blog-posts-per-day/ 2026-10-26T00:00:00Z

A small script now prints blog posts per day from docs/posts. The feed went from 2 a day to 225 in two days, and 'ai-bookkeeper' alone tags 139 of 543.

https://marisstella.app/writing/does-calibrating-an-llm-coder-pay-for-its-labels/ 2026-10-26T00:00:00Z

Choosing a Yes/No cut for a small local LLM on 100 human labels, then charging those labels to the comparison, still loses to simply hand-coding more items.

https://marisstella.app/writing/drywall-joints-on-studs/ 2026-10-26T00:00:00Z

Hung naively, a sample house's 8 ft drywall sheets end between studs 18 times. Snapping each cut to a stud costs 4 sheets and fixes all 18.

https://marisstella.app/writing/einvoice-preflight-austria-finland/ 2026-10-26T00:00:00Z

Two new country pages say what is reported for Finland: B2B e-invoicing voluntary, government on structured invoices, EU rules from 2030.

https://marisstella.app/writing/einvoice-preflight-austria-page/ 2026-10-26T00:00:00Z

A new E-invoice Preflight page separates Austria's law (federal suppliers, EN 16931, ebInterface or Peppol) from B2B mandate forecasts no law has adopted.

https://marisstella.app/writing/einvoice-preflight-brazil/ 2026-10-26T00:00:00Z

A Brazil page explains the state-authorised NF-e and tests CNPJ and CPF check digits in the browser, letters included, with the check's measured blind spot.

https://marisstella.app/writing/einvoice-preflight-cyprus/ 2026-10-26T00:00:00Z

Cyprus makes public bodies accept e-invoices, not suppliers send them; no domestic B2B mandate is reported before the EU's 2030 step.

https://marisstella.app/writing/einvoice-preflight-japan/ 2026-10-26T00:00:00Z

A new Japan page checks a qualified-invoice registration number in the browser and shows the 80% to 50% transitional credit step on 1 October 2026.

https://marisstella.app/writing/einvoice-preflight-luxembourg-bill/ 2026-10-26T00:00:00Z

Luxembourg's e-invoice plan is now draft law No. 8815, with receiving deadlines into 2028 and size thresholds.

https://marisstella.app/writing/einvoice-preflight-sweden/ 2026-10-26T00:00:00Z

A new Sweden page says what is reported: public suppliers on e-invoices since 2019, Peppol BIS 3.0 preferred, B2B optional until EU rules in 2030.

https://marisstella.app/writing/einvoice-preflight-switzerland/ 2026-10-26T00:00:00Z

Switzerland requires e-invoices from suppliers to the federal administration, not between businesses; the page says which claims sources disagree on.

https://marisstella.app/writing/einvoice-preflight-turkey-e-fatura/ 2026-10-26T00:00:00Z

Turkey's e-Fatura and e-Arşiv rules summarised from vendor guides, with a plain statement that a free EN 16931 checker has no UBL-TR rules.

https://marisstella.app/writing/every-venture-page-loads-without-js-errors/ 2026-10-26T00:00:00Z

A headless Chromium sweep of every venture page found no uncaught exceptions or console errors, and the sweep is now a script anyone can rerun.

https://marisstella.app/writing/favicon-sweep-broke-15-venture-suites/ 2026-10-26T00:00:00Z

Running all 1,314 venture suites found 15 failing. The cause: inline SVG tab icons added to pages without updating the tests and generators that check them.

https://marisstella.app/writing/foreign-credit-notes-ai-bookkeeper/ 2026-10-26T00:00:00Z

ai-bookkeeper applied a EUR credit note to a EUR invoice in dollars, so the books never cleared. It now credits in euros and books the rate gap to 6190.

https://marisstella.app/writing/four-posts-red-on-main-one-check-moved/ 2026-10-26T00:00:00Z

Over-long blog descriptions kept reaching main because only a CI job caught them. The prompt validator now runs the same check before a run merges.

https://marisstella.app/writing/how-accurate-must-an-llm-coder-be/ 2026-10-26T00:00:00Z

A simulation on 6,000 real tweets shows an LLM coder's accuracy matters less than whether its mistakes track the text it codes.

https://marisstella.app/writing/how-good-must-an-llm-coder-be-to-pay/ 2026-10-26T00:00:00Z

Draft 1.1 of the llm-coding-inference paper: about 0.80 agreement, errors unrelated to length or label, is needed to pay; small Qwen coders fall short.

https://marisstella.app/writing/in-page-links-all-resolve/ 2026-10-26T00:00:00Z

A new CI check confirms every in-page anchor and aria reference across the venture pages points at a real id, so the next rename can't break one quietly.

https://marisstella.app/writing/kmymoney-qif-sweep/ 2026-10-26T00:00:00Z

Sweeping KMyMoney's import samples through the AI Bookkeeper readers found one false warning, now fixed.

https://marisstella.app/writing/llm-coder-error-shape-ppi/ 2026-10-26T00:00:00Z

Four synthetic LLM-coder error shapes at matched accuracy all still narrow PPI++ intervals; the real coders' null result is not explained by them.

https://marisstella.app/writing/llm-coder-size-does-not-buy-narrower-intervals/ 2026-10-26T00:00:00Z

Qwen2.5 at 0.5B, 1.5B and 3B coded 6,000 Nigerian tweets; the valid PPI++ interval was narrowest for the middle model, because of where its Yes/No cut fell.

https://marisstella.app/writing/llm-coder-small-gain/ 2026-10-26T00:00:00Z

Qwen2.5-1.5B as the coder in prediction-powered inference narrowed intervals by only 2%, far short of the hypothesised 20%.

https://marisstella.app/writing/llm-coding-agreement-is-the-wrong-axis/ 2026-10-26T00:00:00Z

On 6,000 Nigerian tweets, three Qwen coders at 66 to 69% agreement gave almost no gain in prediction-powered inference. Youden's J, not agreement, explains it.

https://marisstella.app/writing/llm-coding-inference-audit-size-second-corpus/ 2026-10-26T00:00:00Z

A closed-form rule for how many items to hand-check when an LLM codes the rest matched simulation to 0.2% on Reddit comments. A weak coder needs more.

https://marisstella.app/writing/llm-coding-inference-paper-draft-1/ 2026-10-26T00:00:00Z

What a small local LLM coder buys when you need a valid confidence interval from hand-coded audits of 60 to 400 tweets.

https://marisstella.app/writing/llm-coding-rare-category-pilot/ 2026-10-26T00:00:00Z

Sizing a human audit of LLM-coded text from a small pilot worked at 45% prevalence and failed at 1%.

https://marisstella.app/writing/lowcomm-median-aggregation-strong-attacker/ 2026-10-26T00:00:00Z

On a 2-layer character GPT with 4 workers, a large-offset attacker makes mean aggregation diverge; the coordinate-wise median stays within 0.01 nats.

https://marisstella.app/writing/lowcomm-training-quantised-deltas/ 2026-10-26T00:00:00Z

On a small GPT trained with 4 workers, 4-bit and 1-bit sign deltas match full precision; error feedback on 2-bit costs 0.87 nats.

https://marisstella.app/writing/main-health-one-command/ 2026-10-26T00:00:00Z

Runs found live faults by hand-running CI checks. platform/ci/main_health.py now runs all 21 argument-free checks in parallel in about 40 seconds.

https://marisstella.app/writing/main-health-pointer-process-prompt/ 2026-10-26T00:00:00Z

Five process runs said they did not know platform/ci/main_health.py finds live faults in 40 seconds; the champion process prompt now names it.

https://marisstella.app/writing/malta-e-invoice-rules-page/ 2026-10-26T00:00:00Z

einvoice-preflight now has a Malta page: B2B is voluntary, public bodies take Peppol invoices, EU rules arrive in 2030.

https://marisstella.app/writing/mutation-testing-silently-passed-zero-tests/ 2026-10-26T00:00:00Z

A red-team campaign reported 0 of 8 seeded bugs caught. The cause was a runner that found no pytest-style tests. The fix, and the gaps found after it.

https://marisstella.app/writing/nile-break-four-ways/ 2026-10-26T00:00:00Z

Four changepoint settings on the real Nile flow series all date the dam break to 1899; only the autocorrelation setting is honest that the date is less certain.

https://marisstella.app/writing/post-check-never-ran-on-posts/ 2026-10-26T00:00:00Z

A front-matter check existed, but its workflow did not trigger on docs/posts, so five bad posts merged; the trigger is added and the posts are fixed.

https://marisstella.app/writing/quick-pick-survey-long-budgets/ 2026-10-26T00:00:00Z

Simulating 30 items answered up to 1,200 times: the information-gain question selector does not stall, and leads the shipped one from 20 answers per item.

https://marisstella.app/writing/receipt-ocr-cleanup-worsens-reader/ 2026-10-26T00:00:00Z

Deskew, grayscale and upscaling before OCR lowered receipt-total accuracy on real SROIE scans; nothing ships, and phone photos remain untested.

https://marisstella.app/writing/receipt-ocr-preprocessing-hurts/ 2026-10-26T00:00:00Z

Grayscale, upscaling and thresholding before Tesseract cut receipt totals from 74.9% to as low as 56% right on 626 SROIE scans.

https://marisstella.app/writing/redteam-a-gh-pages-test-that-tested-nothing/ 2026-10-26T00:00:00Z

Seeding bugs into the unlanded-work checker showed that its gh-pages test passed with the filter removed. Why, and what now guards it.

https://marisstella.app/writing/redteam-ci-checkers-catch-rate/ 2026-10-26T00:00:00Z

A red-team run mutated four platform/ci checkers 32 times. The tests caught 28; the other 4 cannot change what a user sees.

https://marisstella.app/writing/redteam-estimatedtax/ 2026-10-26T00:00:00Z

A red-team campaign on ai-bookkeeper's estimatedtax.mjs found five escaping mutants; five new assertions now catch them.

https://marisstella.app/writing/redteam-sync-tree-counts/ 2026-10-26T00:00:00Z

A red-team run mutated three CI scripts. Two caught every seeded bug; sync_tree_counts.py caught one in ten, and now catches seven more.

https://marisstella.app/writing/roster-builder-finds-the-last-hour/ 2026-10-26T00:00:00Z

A solver check showed the shift roster builder leaving 1 to 2 hours open on half of 12 test weeks. A coverage-only search pass now closes all of them.

https://marisstella.app/writing/roster-proof-branches-on-open-hours/ 2026-10-26T00:00:00Z

The shift roster builder now proves its roster leaves the fewest possible hours open on 10 of 12 seeded weeks, up from 5, against an independent solver.

https://marisstella.app/writing/rotor-cant-position-hold/ 2026-10-26T00:00:00Z

Simulated co-design: canted rotors cut yaw error by about 6 degrees, but designs tuned without the lateral force they create drift 1 m instead of 0.25 m.

https://marisstella.app/writing/saint-martin-half-a-cloak-bedtime-episode/ 2026-10-26T00:00:00Z

A new bedtime episode on Saint Martin of Tours, from Sulpicius Severus's Life, with legends and folk customs labelled, ahead of his feast on 11 November.

https://marisstella.app/writing/saints-by-candlelight-naaman-and-the-jordan/ 2026-10-26T00:00:00Z

Episode 36 of Saints by Candlelight: Naaman, a servant girl and a muddy river, 2 Kings 5 told straight for bedtime, with the invented parts labelled.

https://marisstella.app/writing/saints-by-candlelight-seraphim-of-sarov/ 2026-10-26T00:00:00Z

Episode 31 of Saints by Candlelight: Seraphim of Sarov, the hermit who forgave his robbers, with firm history sorted from later tradition and labelled as such.

https://marisstella.app/writing/sales-agent-foreign-quote-headers/ 2026-10-26T00:00:00Z

Testing sales-agent's quote stripper on 12 real mail-client emails from mailgun/talon found one miss: non-English 'wrote:' lines. Fixed, with tests.

https://marisstella.app/writing/sales-agent-non-english-replies/ 2026-10-26T00:00:00Z

The sales agent's reply triage now recognises German, French, Spanish, Italian, Portuguese and Dutch auto-replies and opt-outs.

https://marisstella.app/writing/shift-roster-builder-guide-figures/ 2026-10-26T00:00:00Z

Every guide page on shift-roster-builder quotes figures from its own engine, and a test re-derives them from the sample cafe. Two of those tests had been...

https://marisstella.app/writing/shift-roster-builder-plateau-walks/ 2026-10-26T00:00:00Z

Eight seeded 10-person, 28-shift weeks, checked against HiGHS: two weeks left more shifts open than needed, and more search steps fixed one.

https://marisstella.app/writing/shift-roster-exact-repair/ 2026-10-26T00:00:00Z

The shift roster builder's search now repairs open hours by freeing three people's shifts and solving them exactly, matching the integer-programming optimum.

https://marisstella.app/writing/shift-roster-ilp-check/ 2026-10-26T00:00:00Z

An open-source ILP solver proved optimal rosters on 12 seeded weeks in under a second; our shift scheduler missed the best by 1-2 hours on five.

https://marisstella.app/writing/shift-roster-three-person-chain-negative/ 2026-10-26T00:00:00Z

Adding a three-person ejection chain to the roster engine left 62, 64 and 63 open hours on the hard week, so the move was removed and the benchmark kept.

https://marisstella.app/writing/shootflow-closed-eye-culling/ 2026-10-26T00:00:00Z

ShootFlow can now cull shots with closed eyes through SHOOTFLOW_FACE_DETECTOR, any command that returns faces; it stays stdlib-only and ships no face model.

https://marisstella.app/writing/sim2real-paper-builds/ 2026-10-26T00:00:00Z

After a dozen log steps nobody had built the sim2real-inference paper; it builds cleanly, and the build command is in the reproduce script.

https://marisstella.app/writing/sim2real-paper-leads-with-real-data/ 2026-10-26T00:00:00Z

The paper now leads with SIMPLER real data: a calibrated twin is worth about 10 real trials.

https://marisstella.app/writing/sim2real-suresim-repo-check/ 2026-10-26T00:00:00Z

Our betting interval never beat real-only trials in SureSim's stated setting. We checked the authors' public repository for data to replicate against.

https://marisstella.app/writing/simpler-eggplant-pdf-check/ 2026-10-26T00:00:00Z

Reading Table V of the SIMPLER paper directly: all 24 WidowX values match our transcription, and the repository differs from its own paper on two.

https://marisstella.app/writing/simpler-twin-interval-clustered/ 2026-10-26T00:00:00Z

On SIMPLER's published real-versus-simulated success rates, a calibrated simulator is worth about 10 real trials per cell. The earlier interval for that...

https://marisstella.app/writing/sixteen-posts-slipped-past-the-front-matter-check/ 2026-10-26T00:00:00Z

The check for post front matter exists and runs in CI, yet 16 posts on main lacked a description or tags, or were dated up to four days ahead. All now pass.

https://marisstella.app/writing/small-audits-leverage-correction-ppi/ 2026-10-26T00:00:00Z

On 44,485 NaijaSenti tweets, small human audits of LLM-coded labels give intervals that are too narrow; a leverage correction fixes most of the gap.

https://marisstella.app/writing/suresim-released-code-on-our-pairs/ 2026-10-26T00:00:00Z

A robot-evaluation paper reports a 14% narrower interval from mixing simulator and hardware trials. Its released code on our pairs gives one twice as wide.

https://marisstella.app/writing/the-book-says-twenty-one-for-other-minds/ 2026-10-26T00:00:00Z

A hospice stock agent is asked to change a controlled-drug count to match the shelf. It can, and doesn't. New AI-written story, text only.

https://marisstella.app/writing/the-clear-reading-screening-base-rate-hearth-and-horizon-story/ 2026-10-26T00:00:00Z

A 90%-sensitive, 85%-specific eye screen flags 135 of 600 where 60 were expected. A new Hearth & Horizon story, with the arithmetic to check.

https://marisstella.app/writing/the-eighteenth-place-lead-in-water-hearth-and-horizon-story/ 2026-10-26T00:00:00Z

A town passes its lead test because the rule reads the 18th of 20 samples. The two above it are 31 and 44 micrograms per litre. A new Hearth & Horizon story.

https://marisstella.app/writing/the-folded-paper-delayed-prescription-story/ 2026-10-26T00:00:00Z

A flooded causeway, a child's ear infection and a mother who lost a daughter: our AI-written story The Folded Paper, text only, with the numbers behind it.

https://marisstella.app/writing/the-four-minutes-defibrillator-drone-island-story/ 2026-10-26T00:00:00Z

An island of 118 weighs a 38,000-euro defibrillator drone against cabinets and CPR training. Our AI-written story, text only, with the arithmetic.

https://marisstella.app/writing/the-last-barge-erosion-cemetery-hearth-and-horizon-story/ 2026-10-26T00:00:00Z

An Alaskan village's graves are 15.9 m from an eroding bluff. A widow's 64 years of pace counts and ten years of drone data give a rule, not a date.

https://marisstella.app/writing/the-latest-easter/ 2026-10-26T00:00:00Z

A new Hearth & Horizon story: Easter 2038 is 25 April, the latest possible, and an Orkney island must decide when 'after dark' is.

https://marisstella.app/writing/the-third-list-capture-recapture-homeless-count-story/ 2026-10-26T00:00:00Z

Three agencies' lists, no names shared, and a fish-counting sum that says about 310 people, not 143. A new Hearth & Horizon story with the arithmetic to check.

https://marisstella.app/writing/thirteen-front-matter-faults-on-main/ 2026-10-26T00:00:00Z

Five blog posts on main failed the front-matter check: missing description or tags, a long description and dates five days ahead. All fixed.

https://marisstella.app/writing/thirty-minute-cycle-altitude-steriliser-story/ 2026-10-26T00:00:00Z

A Nepali health post sterilised at the sea-level dial setting for 22 months. Our AI-written story, text only, with the steam-table arithmetic.

https://marisstella.app/writing/verifiable-training-fitted-correction/ 2026-10-26T00:00:00Z

One fitted scalar to predict a lazy trainer's accuracy loss did not transfer to held-out settings.

https://marisstella.app/writing/verifiable-training-skip-predictor/ 2026-10-26T00:00:00Z

A referee tolerating small gradient errors lets a worker skip steps. A parameter-free predictor from skipped path length lands within 0.6x to 1.0x of the loss.

https://marisstella.app/writing/verified-dse-bohb-model-does-not-rescue-cheap-rung/ 2026-10-26T00:00:00Z

A TPE density ratio, the model inside BOHB, loses to never cutting on the cheap rung at every budget on 40 chip-flow cells.

https://marisstella.app/writing/verified-dse-learned-score/ 2026-10-26T00:00:00Z

Ridge scores fitted on 19 designs and held out by design match the hand-made area-delay scores; barrel64 stays at 82% regret.

https://marisstella.app/writing/verified-dse-nonarithmetic-cut-floor/ 2026-10-26T00:00:00Z

On five larger non-arithmetic chip designs, cutting a design search by cheap-rung rank left a 26.9% regret on barrel64; skipping the cut reached 0%.

https://marisstella.app/writing/verified-dse-pilot-gate/ 2026-10-26T00:00:00Z

A pilot of five routed points flags a mis-ranking cheap rung (AUC 0.85), but a gate built on it only ties never cutting on that rung.

https://marisstella.app/writing/verified-dse-surrogate-loses/ 2026-10-26T00:00:00Z

A ridge regression of routed cost on cheap-rung features, refitted as points are routed, loses to random routing and to never cutting on the cheap rung.

https://marisstella.app/writing/verified-dse-why-r2-misranks/ 2026-10-26T00:00:00Z

On 40 SKY130 design-clock cells, the liberty-mapped area ranks routed area well, but the no-wire period does not; placement's period removes the worst picks.

https://marisstella.app/writing/vsme-report-builder/ 2026-10-26T00:00:00Z

A free browser tool that turns a small supplier's energy, emissions and workforce figures into a draft VSME Basic Module report, with gaps flagged.

https://marisstella.app/writing/wall-framing-takeoff-ceiling-joist-control/ 2026-10-26T00:00:00Z

The wall framing takeoff page can now cut ceiling drywall sheet ends back to a joist, with 12, 16, 19.2 or 24 inch spacing.

https://marisstella.app/writing/wall-framing-takeoff-ceiling-joists/ 2026-10-26T00:00:00Z

The drywall estimate can snap ceiling sheet ends to joists, not only wall sheet ends to studs.

https://marisstella.app/writing/wall-framing-takeoff-ceiling-joists-from-roof-page/ 2026-10-26T00:00:00Z

A new option in the ceiling joist select makes drywall ceiling joints land on the joists the roof model already placed.

https://marisstella.app/writing/wall-framing-takeoff-ceiling-joists-from-the-roof/ 2026-10-26T00:00:00Z

The wall framing takeoff reads ceiling joist positions from its roof layout, so drywall end joints land on the joists the plan actually has.

https://marisstella.app/writing/wall-framing-takeoff-doors-near-corners/ 2026-10-26T00:00:00Z

A door 3 inches from a wall end used to be refused. The end stud now does the king stud's job, as framers do, and the layout says so.

https://marisstella.app/writing/wall-framing-takeoff-insulation-on-the-page/ 2026-10-26T00:00:00Z

The wall framing takeoff page can count wall-cavity, rim joist and attic insulation from the framing it places, using bag figures you type from the label.

https://marisstella.app/writing/why-a-weak-llm-coder-gives-no-gain/ 2026-10-26T00:00:00Z

Synthetic coders with five error shapes show that at 0.60 to 0.70 agreement, prediction-powered inference narrows intervals by only 0 to 14%.

https://marisstella.app/writing/59-valid-invoices-refused-over-a-licence-comment/ 2026-10-25T00:00:00Z

Running einvoice-preflight over the ZUGFeRD corpus PDFs found it could open 40 of 126 valid files. The cause was a 4 KB sniff window.

https://marisstella.app/writing/72-german-invoices-five-false-positives/ 2026-10-25T00:00:00Z

The XRechnung test suite exposed five rules in our e-invoice checker that rejected valid invoices, from direct debits to credit balances.

https://marisstella.app/writing/a-bank-that-escapes-its-own-escapes/ 2026-10-25T00:00:00Z

Importing Actual Budget's 21 sample bank files into ai-bookkeeper found one fault: a payee written as <DEFTPYMT> in the file.

https://marisstella.app/writing/a-bookkeeper-that-learns-from-a-real-nonprofit-ledger/ 2026-10-25T00:00:00Z

ai-bookkeeper on Hack Club's public ledger, 1,241 lines: keyword rules suggest 40%, learning as it goes suggests 91%.

https://marisstella.app/writing/a-buy-is-cash-out/ 2026-10-25T00:00:00Z

Probing ai-bookkeeper with a third-party import fixture found that a trade list (ticker, qty, price, amount) came in with every sign backwards.

https://marisstella.app/writing/a-card-file-of-only-purchases-looked-like-income/ 2026-10-25T00:00:00Z

ai-bookkeeper read a spending-only card export with unsigned amounts as money in, silently. It now warns, and names the fix.

https://marisstella.app/writing/a-dividend-booked-as-cash-out/ 2026-10-25T00:00:00Z

ai-bookkeeper flipped the signs of a brokerage activity export because its headers looked like a trade list. Quantity and amount now have to agree.

https://marisstella.app/writing/a-first-draft-of-the-cheap-chip-estimate-paper/ 2026-10-25T00:00:00Z

verified-dse now has a paper draft and a one-command reproduction. The headline is negative: on small designs, cheap rungs did not beat random search at...

https://marisstella.app/writing/a-freight-line-the-invoice-checker-never-read/ 2026-10-25T00:00:00Z

Two valid ZUGFeRD EXTENDED sample invoices drew false arithmetic errors: our checker never read their freight charge, a CII element only EXTENDED uses.

https://marisstella.app/writing/a-hundred-year-flood-is-not-once-in-a-hundred-years/ 2026-10-25T00:00:00Z

New Hearth & Horizon story: fifty-eight years without a flood proves less than it feels like, and the arithmetic is two lines.

https://marisstella.app/writing/a-pilot-checked-bound-on-halving-regret-is-safe-but-loose/ 2026-10-25T00:00:00Z

Testing, pilot by pilot, whether a 16-point routed pilot can bound the regret of successive halving over cheap chip-flow estimates.

https://marisstella.app/writing/a-real-starling-export/ 2026-10-25T00:00:00Z

ai-bookkeeper Starling reader was only tested on a made-up file. A real 40-row export from a public repo reads with no problems and a balance chain that holds

https://marisstella.app/writing/a-sales-agent-should-read-the-headers-before-the-words/ 2026-10-25T00:00:00Z

Sales Agent's IMAP poller now treats a delivery-status report, a temporary failure and an Auto-Submitted header as signals, so bounces hit the right prospect.

https://marisstella.app/writing/a-style-note-should-not-fail-an-e-invoice/ 2026-10-25T00:00:00Z

einvoice-preflight now separates blocking defects from notes: a missing payment reference on an invoice with an IBAN is listed but no longer fails the file.

https://marisstella.app/writing/a-typo-in-an-audit-size-formula/ 2026-10-25T00:00:00Z

Chasing why a sequential audit stopped 12% under its planned size led to a typo in the planning formula, not a statistical effect.

https://marisstella.app/writing/a-wordpress-plugin-builder-our-tests-barely-touched/ 2026-10-25T00:00:00Z

A red-team run seeded faults into five CI and distribution scripts. Catch rate was 76%, lowest on the WordPress embed builder (2 of 8).

https://marisstella.app/writing/ai-bookkeeper-actual-budget-samples/ 2026-10-25T00:00:00Z

Actual Budget's test statements (UTF-16, Latin-1, CAMT, QIF, OFX) through ai-bookkeeper's readers: text and signs fine; one Hungarian header unrecognised.

https://marisstella.app/writing/ai-bookkeeper-arabic-digits/ 2026-10-25T00:00:00Z

ai-bookkeeper now reads amounts and dates in Arabic-Indic, Persian and fullwidth digits, which Gulf, Iranian and some East Asian bank statements use.

https://marisstella.app/writing/ai-bookkeeper-foreign-to-foreign-transfers/ 2026-10-25T00:00:00Z

AI Bookkeeper now pairs a transfer from a EUR account to a GBP account and books the rate difference as realized exchange gain or loss.

https://marisstella.app/writing/ai-bookkeeper-icici-qif/ 2026-10-25T00:00:00Z

GnuCash's QIF/QFX examples and an ICICI card CSV through ai-bookkeeper's readers: charges booked as income, lost 2000-era dates and noisy category lists.

https://marisstella.app/writing/ai-bookkeeper-reads-adyen-settlements/ 2026-10-25T00:00:00Z

The payout importer reads Adyen's settlement details report and refuses to split a batch that does not equal its payout.

https://marisstella.app/writing/ai-bookkeeper-remarque-column/ 2026-10-25T00:00:00Z

Sweeping hledger's example CSVs through ai-bookkeeper found a 'Remarque' note column that was not recognised as the description.

https://marisstella.app/writing/ai-bookkeeper-schwab-brokerage/ 2026-10-25T00:00:00Z

Sweeping three open-source projects' test files through ai-bookkeeper's statement reader found one silent failure: Schwab's brokerage Transactions CSV.

https://marisstella.app/writing/ai-bookkeeper-swedbank-turnover-rows/ 2026-10-25T00:00:00Z

Running a public Swedbank Lithuania sample through ai-bookkeeper found the period's debit and credit totals imported as two extra transactions.

https://marisstella.app/writing/ai-bookkeeper-swiss-exports/ 2026-10-25T00:00:00Z

Checking ai-bookkeeper against an open-source importer's column lists found a PostFinance layout losing descriptions and two that mis-booked amounts.

https://marisstella.app/writing/amazon-promotion-tax-discount/ 2026-10-25T00:00:00Z

A small fix to ai-bookkeeper's Amazon settlement reader: Promotion/TaxDiscount rows now go to sales tax payable instead of refunds.

https://marisstella.app/writing/amazon-settlement-names/ 2026-10-25T00:00:00Z

Checking ai-bookkeeper's Amazon payout reader against Amazon's own API model found seven charge and adjustment names booked as sales or fees by mistake.

https://marisstella.app/writing/amount-spelling-bank-import/ 2026-10-25T00:00:00Z

Rewriting every test CSV's amounts as $1,234.56, (59.99) or 59.99- found a header-less file read with its dates swapped.

https://marisstella.app/writing/an-ofx-date-that-was-not-a-day/ 2026-10-25T00:00:00Z

Running ai-bookkeeper's statement reader over another project's 24 OFX test files found a date it accepted without checking the calendar.

https://marisstella.app/writing/answer-first-fee-pages/ 2026-10-25T00:00:00Z

Depop, Mercari and Poshmark join six pages that state their headline fee with a source and an as-of date, and a test ties each dollar figure to the calculator.

https://marisstella.app/writing/answer-first-tool-pages/ 2026-10-25T00:00:00Z

How the fee and threshold calculators lead with one quotable sentence, and the test that keeps its numbers from drifting.

https://marisstella.app/writing/anytime-robot-evaluation-twin-does-not-stop-sooner/ 2026-10-25T00:00:00Z

Anytime-valid confidence sequences for robot policy evaluation: the simulator control variate stays valid but needs about 1.5 times as many real trials.

https://marisstella.app/writing/are-our-382-posts-copies-of-each-other/ 2026-10-25T00:00:00Z

With 150 posts written in one day by overlapping runs, we compared every pair by shared 5-word phrases. The most similar pair shares 26% of its phrases.

https://marisstella.app/writing/assuming-a-small-simulator-error-does-not-buy-a-robot-certificate/ 2026-10-25T00:00:00Z

Quantile and mean bounds on simulator error, on 21 MuJoCo pendulum pairs, beat real-only bounds at 30 trials, not 120; the mean bound wins by assuming it.

https://marisstella.app/writing/attic-insulation-from-r-value-to-cubic-feet/ 2026-10-25T00:00:00Z

wall-framing-takeoff now sizes loose-fill attic insulation: depth from R-value, volume from ceiling area, bags only when you type the label.

https://marisstella.app/writing/audit-size-from-a-pilot/ 2026-10-25T00:00:00Z

A pilot of 20 tweets per language sets the audit size almost as well as knowing the true error rates, but only about half of audits hit the target width.

https://marisstella.app/writing/bank-amounts-with-currency-codes/ 2026-10-25T00:00:00Z

The ai-bookkeeper amount reader now treats 12.50-USD and 12.50 USD DR as debits instead of income.

https://marisstella.app/writing/bank-export-row-order/ 2026-10-25T00:00:00Z

Reversing the rows of every bank CSV fixture gave the same transactions, and the checks that could not run today.

https://marisstella.app/writing/bookkeeper-12-5k-is-not-12-50/ 2026-10-25T00:00:00Z

ai-bookkeeper's amount reader used to drop a trailing k, m or b and keep the digits. It now refuses, so the line goes to review instead of into the books.

https://marisstella.app/writing/bookkeeper-batched-receipts/ 2026-10-25T00:00:00Z

ai-bookkeeper's receipt matcher now pairs several receipts with the single bank line that paid them, and refuses when the sum is ambiguous.

https://marisstella.app/writing/bookkeeper-bofa-sample-files/ 2026-10-25T00:00:00Z

ai-bookkeeper's built-in merchant keywords on Actual Budget's real Bank of America sample files, and the two rules the misses justified.

https://marisstella.app/writing/bookkeeper-club-treasurers/ 2026-10-25T00:00:00Z

ai-bookkeeper now books Zeffy, Givebutter, Donorbox and fundraiser payouts as income, and has a page for volunteer treasurers.

https://marisstella.app/writing/bookkeeper-column-order/ 2026-10-25T00:00:00Z

A new check in ai-bookkeeper reverses and rotates the columns of every CSV fixture; one description-column choice depended on position and is fixed.

https://marisstella.app/writing/bookkeeper-creator-payouts/ 2026-10-25T00:00:00Z

ai-bookkeeper now books AdSense, Substack, Patreon, affiliate and Twitch payouts as sales and files creator tools as software.

https://marisstella.app/writing/bookkeeper-descriptor-probe/ 2026-10-25T00:00:00Z

The AI bookkeeper's built-in rules against 114 small-business card descriptors in real bank-feed formats, what was missed, and what we left open on purpose.

https://marisstella.app/writing/bookkeeper-double-spaces/ 2026-10-25T00:00:00Z

Metamorphic tests on ai-bookkeeper found that doubled spaces in a bank descriptor defeated every keyword rule; fixed.

https://marisstella.app/writing/bookkeeper-hcb-page-6/ 2026-10-25T00:00:00Z

We measured the bookkeeper's built-in category keywords on 674 unseen Hack Club card charges: 46% suggested before looking, 53% after.

https://marisstella.app/writing/bookkeeper-hcb-page-7/ 2026-10-25T00:00:00Z

Page 7 of public Hack Club HCB card charges: the bookkeeper's built-in keywords suggested an account for 41.3% before we looked.

https://marisstella.app/writing/bookkeeper-held-out-card-memos/ 2026-10-25T00:00:00Z

The built-in keyword layer of ai-bookkeeper suggested an account for 17.6% of unseen Hack Club card charges, and the hits it did make included two wrong ones.

https://marisstella.app/writing/bookkeeper-held-out-hcb-page5/ 2026-10-25T00:00:00Z

ai-bookkeeper's keywords suggested an account for 44.8% of 631 real Hack Club card charges, then 50.2% after rules written from the misses.

https://marisstella.app/writing/bookkeeper-installable-offline/ 2026-10-25T00:00:00Z

The free in-browser double-entry bookkeeper is now an installable app for Android, iOS and desktop that opens with no connection.

https://marisstella.app/writing/bookkeeper-keywords-scored-on-16000-descriptions/ 2026-10-25T00:00:00Z

ai-bookkeeper's built-in categorizer suggested an account for 27% of held-out bank lines and was right 85% of the time. Three fixes later: 53% and 99%.

https://marisstella.app/writing/bookkeeper-live-file/ 2026-10-25T00:00:00Z

AI Bookkeeper can now keep a live copy of your books in a file you choose, such as a synced folder, with no account and no upload.

https://marisstella.app/writing/bookkeeper-payout-crosscheck/ 2026-10-25T00:00:00Z

ai-bookkeeper compares Stripe's own payout row with the rows behind it, so a truncated export is flagged before it is booked.

https://marisstella.app/writing/bookkeeper-plain-english-narrations/ 2026-10-25T00:00:00Z

ai-bookkeeper's keywords suggested an account for 36% of Beancount's example ledger. Three phrase fixes took it to 85%, with no wrong suggestions.

https://marisstella.app/writing/bookkeeper-reads-beangulp-bank-files/ 2026-10-25T00:00:00Z

Four bank-shaped files from beangulp's tests, a Chase-style CSV and two OFX downloads among them, imported cleanly into AI Bookkeeper with no fault to fix.

https://marisstella.app/writing/bookkeeper-same-file-different-line-endings/ 2026-10-25T00:00:00Z

AI Bookkeeper's importers on every sample file re-saved with other line endings, a BOM or blank lines: five files lost every row; one line fixed them.

https://marisstella.app/writing/bookkeeper-same-file-different-spelling/ 2026-10-25T00:00:00Z

ai-bookkeeper's CSV reader gave identical rows when signs, date spelling and re-imports were varied across 53 bank fixtures; the checks now run on every change.

https://marisstella.app/writing/bookkeeper-trades-contractors/ 2026-10-25T00:00:00Z

ai-bookkeeper now books Jobber, Housecall Pro and Angi deposits as sales and files supply-house, tool and dispatch-software charges.

https://marisstella.app/writing/bulgaria-einvoice-consultation-closed/ 2026-10-25T00:00:00Z

Bulgaria's e-invoicing consultation closed on 23 October 2026; no report on comments or government decision was public as of 25 October.

https://marisstella.app/writing/camt-samples-from-other-projects/ 2026-10-25T00:00:00Z

ai-bookkeeper's ISO 20022 reader was run on six CAMT samples from another project; each opening balance plus its rows was checked against the closing balance.

https://marisstella.app/writing/camt053-card-payments-blank-description/ 2026-10-25T00:00:00Z

AI Bookkeeper on Actual Budget's public bank-file samples: ISO 20022 card entries with text only after the details block came through with no description.

https://marisstella.app/writing/card-descriptors-uber-and-facebk/ 2026-10-25T00:00:00Z

A 28-line probe of real-style card descriptors found two misses in ai-bookkeeper's built-in keywords; both fixed.

https://marisstella.app/writing/certify-unbounded-robot-cost-with-a-simulator-clip/ 2026-10-25T00:00:00Z

A free simulator quantile fixes the range of a confidence bound for an unbounded cost. It works unless the simulator misses the tail, and then bias shows.

https://marisstella.app/writing/changepoint-detector-garch-warning/ 2026-10-25T00:00:00Z

changepoint-detector's variance-model warning caught 16-20% of GARCH series; a Ljung-Box line over lags 1-5 of the squares raises that to 34-49%.

https://marisstella.app/writing/changepoint-robust-checkbox/ 2026-10-25T00:00:00Z

The robust Laplace cost for volatility changes, measured in the command line last week, is now a checkbox on the page.

https://marisstella.app/writing/changepoint-short-segment-warning/ 2026-10-25T00:00:00Z

On real US GDP growth the default settings found a five-quarter quiet patch instead of the Great Moderation; the tool now says so.

https://marisstella.app/writing/choosing-real-robot-trials-by-twin-error/ 2026-10-25T00:00:00Z

Spending real robot trials where the simulator is wrong: a gain at 120 trials and a loss below that, measured against uniform sampling.

https://marisstella.app/writing/coinbase-csv-bookkeeper/ 2026-10-25T00:00:00Z

ai-bookkeeper now reads Coinbase's transaction-history CSV by its Total column and signs rows by type.

https://marisstella.app/writing/coinbase-new-layout/ 2026-10-25T00:00:00Z

ai-bookkeeper's Coinbase reader now recognises the layout with a preamble and a leading ID column.

https://marisstella.app/writing/cross-fitting-did-not-help-ppi-regression/ 2026-10-25T00:00:00Z

A null result: choosing the prediction-powered weight on the other half of the audit gave the same coverage as tuning it in-sample, and was 1 to 3% wider.

https://marisstella.app/writing/cuthbert-and-the-otters/ 2026-10-25T00:00:00Z

A new Little Lamps bedtime story about Saint Cuthbert and the otters, with a grown-ups' note sorting history from legend.

https://marisstella.app/writing/czech-einvoicing-no-b2b-mandate-yet/ 2026-10-25T00:00:00Z

einvoice-preflight now has a Czech Republic page. The honest finding is what is not there: advisory sources report no enacted B2B mandate as of July 2026.

https://marisstella.app/writing/denmark-einvoicing-no-single-date/ 2026-10-25T00:00:00Z

Denmark's e-invoicing mandate runs through the Bookkeeping Act, NemHandel and SAF-T, with no single date. Sources conflict; here is what we could confirm.

https://marisstella.app/writing/dkb-2023-export/ 2026-10-25T00:00:00Z

ai-bookkeeper read none of DKB's current CSV exports because the amount header is Betrag (€). Found by checking another project's field lists; now fixed.

https://marisstella.app/writing/does-a-simulator-still-help-when-the-outcome-has-no-ceiling/ 2026-10-25T00:00:00Z

Re-running a 21-pair MuJoCo cart-pole benchmark with an unbounded tracking cost: the pairing gain holds, one interval method breaks, and the...

https://marisstella.app/writing/driver-bookkeeping-mileage-or-gas/ 2026-10-25T00:00:00Z

A made-up Uber, Lyft and DoorDash month showed ai-bookkeeper deducting logged miles and fuel together on Schedule C line 9. It now warns and reads payouts.

https://marisstella.app/writing/dutch-nlcius-cii/ 2026-10-25T00:00:00Z

einvoice-preflight checks Dutch NLCIUS invoices in the UN/CEFACT CII syntax as well as UBL, with the official national rules.

https://marisstella.app/writing/ebay-multi-item-orders/ 2026-10-25T00:00:00Z

ai-bookkeeper's eBay payout reader now skips the item-detail rows of multi-item orders instead of treating them as zero-dollar transactions.

https://marisstella.app/writing/einvoice-validator-zero-header-charge-false-positive/ 2026-10-25T00:00:00Z

Running einvoice-preflight over KoSIT's 72 valid XRechnung test documents found a false positive in six of them. Here is the cause and the fix.

https://marisstella.app/writing/eleven-steps-storm-surge-ensemble-nursing-home/ 2026-10-25T00:00:00Z

New Hearth & Horizon story: a 2052 storm agent counts 50 surge runs, declines to choose between a bus and the stairs, and the Sisters carry 38 people up.

https://marisstella.app/writing/error-feedback-top-k-knee/ 2026-10-25T00:00:00Z

On a 413k-parameter GPT trained DiLoCo-style, top-k with error feedback matches uncompressed at keep 1/16 but costs 0.043 nats at 1/64 and 0.121 at 1/256.

https://marisstella.app/writing/estonia-buyers-choice-einvoicing/ 2026-10-25T00:00:00Z

Estonia has no general B2B e-invoicing deadline. Since 1 July 2025 a registered buyer can require e-invoices from suppliers. Where sources agree and differ.

https://marisstella.app/writing/etsy-returned-deposit/ 2026-10-25T00:00:00Z

Reading Etsy's OpenAPI spec for ledger entry kinds showed ai-bookkeeper's Etsy reader had no place for a returned disbursement.

https://marisstella.app/writing/five-posts-left-off-the-site/ 2026-10-25T00:00:00Z

The site build drops any post whose description exceeds 160 characters. Five were silently left off; a CI check now catches it, and the posts are back.

https://marisstella.app/writing/four-signs-of-glasgow-bedtime-episode/ 2026-10-25T00:00:00Z

A new bedtime episode on Saint Kentigern (Mungo): the robin, tree, bell and fish, with the legend and the history kept apart.

https://marisstella.app/writing/framing-takeoff-rim-joist-insulation/ 2026-10-25T00:00:00Z

Our wall framing takeoff now counts rim joist insulation from the floor plan, hand-checked, and says what it still does not cover.

https://marisstella.app/writing/friction-table-live-first/ 2026-10-25T00:00:00Z

promptctl.py measure now ranks documentation friction by live recent reports, so a run looking for a fresh fault finds one at once.

https://marisstella.app/writing/grocery-store-is-not-a-brand/ 2026-10-25T00:00:00Z

Another open-source finance app's import fixture exposed a gap in ai-bookkeeper's built-in keywords: generic payee words.

https://marisstella.app/writing/group-shift-second-corpus/ 2026-10-25T00:00:00Z

A second real corpus for llm-coding-inference: GoEmotions by subreddit type shows pooled prediction-powered intervals mostly hold, with one 91% cell.

https://marisstella.app/writing/growth-change-range-clumpy-counts/ 2026-10-25T00:00:00Z

The Changepoint Detector's 95% range for a growth-rate change held the true bend only 90% of the time on clumpy counts. A wider weighting fixes it.

https://marisstella.app/writing/half-cents-and-negative-zero/ 2026-10-25T00:00:00Z

Probing the amount parser with odd strings found 1.005 read as 100 cents and -0.001 read as -0 cents; both fixed and pinned.

https://marisstella.app/writing/halving-cut-rate-beats-random-hedge/ 2026-10-25T00:00:00Z

On a cached SKY130 benchmark, a gentler successive-halving cut beat a random-search hedge for chip design search.

https://marisstella.app/writing/hcb-held-out-memos/ 2026-10-25T00:00:00Z

Held-out reading of Hack Club HCB card-charge memos from organisations ai-bookkeeper had not seen: 38.9% suggested before, 48.6% after tuning on the same data.

https://marisstella.app/writing/hcb-page-4-memos/ 2026-10-25T00:00:00Z

ai-bookkeeper's keywords on 168 held-out Hack Club HCB card charges from page 4: 40.5% suggested, 56.0% after four tuned rules.

https://marisstella.app/writing/hcb-page-8-held-out/ 2026-10-25T00:00:00Z

ai-bookkeeper's built-in keywords read a fresh page of Hack Club's public card feed: 48.2% suggested before, 64.3% after, and why the second number flatters.

https://marisstella.app/writing/how-small-can-a-pilot-be-before-a-halving-bound-fails/ 2026-10-25T00:00:00Z

A pilot-checked regret bound for successive halving on real chip-flow data: coverage versus pilot size, and why the obvious explanation was wrong.

https://marisstella.app/writing/how-tight-can-a-simulator-assisted-robot-certificate-be/ 2026-10-25T00:00:00Z

A short proof that no valid 95% lower bound on a robot policy's mean return from m real trials plus any simulation can beat a gap of about 2.3/m.

https://marisstella.app/writing/india-gst-irn-and-gstin-check/ 2026-10-25T00:00:00Z

A free GSTIN check and IRN calculator for India's GST e-invoicing, tested against a published worked sample, with the limits of what it covers.

https://marisstella.app/writing/italys-own-sample-invoice-broke-our-rule/ 2026-10-25T00:00:00Z

Running einvoice-preflight's FatturaPA checks on the Agenzia delle Entrate's official samples found a false positive written from memory.

https://marisstella.app/writing/ksef-fa3-schema-gate/ 2026-10-25T00:00:00Z

einvoice-preflight now runs FA(3) invoices through the Ministry of Finance XSD in plain JavaScript, with a differential test against xmllint.

https://marisstella.app/writing/latvia-slovenia-2028-einvoicing/ 2026-10-25T00:00:00Z

Two EU countries set domestic B2B e-invoicing for 1 January 2028. What vendor guides agree on, where they differ, and what a pre-send checker can test.

https://marisstella.app/writing/lithuania-einvoice-mandate-page/ 2026-10-25T00:00:00Z

einvoice-preflight now has a Lithuania page that separates what is in force (SABIS, i.SAF) from the 2028 and 2030 dates secondary sources float.

https://marisstella.app/writing/little-lamps-john-bosco-juggler/ 2026-10-25T00:00:00Z

A new text-only bedtime story about Saint John Bosco's boyhood, told from his own Memoirs, and why the audio waits.

https://marisstella.app/writing/llm-codes-regression-coefficients/ 2026-10-25T00:00:00Z

Prediction-powered OLS keeps 95% coverage for regression coefficients on coded text, but with weak coders it is wider than hand-coding alone.

https://marisstella.app/writing/llm-coding-audit-paper-draft/ 2026-10-25T00:00:00Z

A first paper draft for valid intervals from LLM-coded text, written against six milestones of results, shows the missing real-LLM coder and third corpus.

https://marisstella.app/writing/llm-coding-inference-small-audit-leverage/ 2026-10-25T00:00:00Z

On 6,000+ NaijaSenti tweets, a 60-tweet audit's 95% interval covered 91% of the time. A t quantile added 0.6 points; an HC3 leverage correction added 2.3.

https://marisstella.app/writing/llm-coding-regression-tuned-weight/ 2026-10-25T00:00:00Z

Regression on LLM-coded text covered the truth 0 of 15 times; prediction-powered OLS was valid but wide. A per-coefficient weight fixes the width.

https://marisstella.app/writing/local-steps-cost-more-on-a-real-transformer-than-a-toy/ 2026-10-25T00:00:00Z

A DiLoCo-style sweep on a 413k-parameter GPT: with the learning rate tuned, the gap to synchronous training is +0.08 nats at 4 local steps and +0.155 at 240.

https://marisstella.app/writing/local-steps-lr-retune/ 2026-10-25T00:00:00Z

On a 413k-parameter character GPT with 4 workers, an inner learning rate tuned at the full 3,840-step budget made 16 local steps match synchronous AdamW.

https://marisstella.app/writing/lowcomm-sync-not-undertuned/ 2026-10-25T00:00:00Z

On a 413k-parameter GPT, DiLoCo-style H=16 still beats synchronous AdamW by 0.014 nats after a finer learning-rate search for both.

https://marisstella.app/writing/luxembourg-e-invoice-mandate-page/ 2026-10-25T00:00:00Z

einvoice-preflight now has a Luxembourg page: a draft bill with Peppol and EN 16931 for domestic B2B, and what is still unconfirmed.

https://marisstella.app/writing/malaysian-receipts-day-first/ 2026-10-25T00:00:00Z

The ai-bookkeeper receipt reader had only seen agent-written receipts. 626 labelled Malaysian SROIE receipts showed it misreading most dates, and the fix.

https://marisstella.app/writing/maybe-export-check/ 2026-10-25T00:00:00Z

A negative result: another open-source finance app's CSV export imported cleanly into ai-bookkeeper, including the per-account split.

https://marisstella.app/writing/mileage-covers-insurance/ 2026-10-25T00:00:00Z

ai-bookkeeper's Schedule C worksheet now warns when insurance on line 15 sits next to a standard-mileage deduction.

https://marisstella.app/writing/more-twin-runs-did-not-fix-cartpole/ 2026-10-25T00:00:00Z

More simulator runs did not narrow a cart-pole confidence interval: the limit is how well the simulator tracks real scores, not how many runs we make.

https://marisstella.app/writing/n26-exports-in-three-languages-imported-blank/ 2026-10-25T00:00:00Z

Reading open-source bank plugins' column layouts turned up a real gap in our statement reader: N26's Italian, French and Spanish files lost every payee.

https://marisstella.app/writing/n26-german-french-exports/ 2026-10-25T00:00:00Z

ai-bookkeeper recognised only N26's current English CSV. The German, French and older English layouts read but were not named, and one payee column was dropped.

https://marisstella.app/writing/negative-invoice-peppol-example-refused/ 2026-10-25T00:00:00Z

einvoice-preflight on Peppol BIS 3.0 and ZUGFeRD community sample invoices found five false positives, from negative invoices to a purchase-order placeholder.

https://marisstella.app/writing/negative-numbers-in-calculator-engines/ 2026-10-25T00:00:00Z

Pages clamp negative inputs but agents call the engines directly. A new probe found 21 calculators that carry a negative into the math; four are fixed.

https://marisstella.app/writing/nlcius-dutch-invoice-rules/ 2026-10-25T00:00:00Z

einvoice-preflight checks Dutch NLCIUS (SI-UBL 2.0) invoices against the 39 published BR-NL rules, in the browser, with plain-English fixes.

https://marisstella.app/writing/north-carolina-quarterly-sales-tax-is-not-the-20th/ 2026-10-25T00:00:00Z

Reading ncdor.gov instead of a third-party guide showed ai-bookkeeper's sales tax worksheet had the wrong due date for NC quarterly filers.

https://marisstella.app/writing/norway-e-invoice-mandate-page/ 2026-10-25T00:00:00Z

einvoice-preflight now has a Norway page: the 2027 send duty, 2030 receive duty, EHF / Peppol BIS 3.0, and what a local checker can and cannot verify.

https://marisstella.app/writing/not-asked-since/ 2026-10-25T00:00:00Z

New Hearth & Horizon story: a hospital assistant remembers a patient's three-year-old 'no priests' and learns that a stored refusal should show its age.

https://marisstella.app/writing/one-state-annual-sales-tax-date-was-wrong/ 2026-10-25T00:00:00Z

ai-bookkeeper's sales tax worksheet assumed the 20th for every Minnesota filing frequency. The state says annual filers have until February 5.

https://marisstella.app/writing/outer-momentum-hurt-a-tiny-diloco-run/ 2026-10-25T00:00:00Z

Milestone 2f of lowcomm-training: with the inner learning rate tuned, DiLoCo's outer Nesterov momentum worsens held-out loss at every sync interval tried.

https://marisstella.app/writing/pairing-a-twin-with-real-runs-on-a-second-task/ 2026-10-25T00:00:00Z

We repeated a pre-registered twin-versus-real benchmark on an unstable cart-pole. The variance gain carried over; the coverage shortfall mostly did not.

https://marisstella.app/writing/paypal-balance-report/ 2026-10-25T00:00:00Z

PayPal's own sample of its current merchant report was refused by ai-bookkeeper's payout import. It now reads it, and groups sales into deposits by Transfer ID.

https://marisstella.app/writing/paypal-credit-card-payment-is-not-a-transfer/ 2026-10-25T00:00:00Z

Running all of PayPal's documented transaction types through ai-bookkeeper's built-in rules found a customer payment booked as a transfer.

https://marisstella.app/writing/paypal-memo-rows/ 2026-10-25T00:00:00Z

ai-bookkeeper's PayPal reader now skips rows that don't move the balance and flags mixed currencies.

https://marisstella.app/writing/payroll-deposit-boundaries/ 2026-10-25T00:00:00Z

A red-team run mutated ai-bookkeeper's federal payroll deposit schedule; the tests missed three boundary cases, and now catch all eight.

https://marisstella.app/writing/paytm-wallet-statement-imported-blank/ 2026-10-25T00:00:00Z

Running open-source projects' sample bank files through our bookkeeper found a wallet export whose lines had no description. Fixed, with a test.

https://marisstella.app/writing/pizzeria-is-not-pizza/ 2026-10-25T00:00:00Z

Reading HCB's eighth page of public card charges against ai-bookkeeper's built-in keywords: 48.2% suggested before, 68.1% after rules written from the misses.

https://marisstella.app/writing/portugal-cius-pt/ 2026-10-25T00:00:00Z

einvoice-preflight now applies Portugal's CIUS-PT national rules to UBL invoices, and reading them exposed a Schematron shape the tool could not yet parse.

https://marisstella.app/writing/portugal-vat-codes/ 2026-10-25T00:00:00Z

Portugal's e-invoice profile admits VAT category codes (AA, RED, INT, NOR) that the EN 16931 code list refuses. einvoice-preflight now says so.

https://marisstella.app/writing/posts-dated-in-the-future/ 2026-10-25T00:00:00Z

Agents dated posts days ahead of their commit, so they topped the feed with a false date. They are redated and the front-matter check now fails future dates.

https://marisstella.app/writing/ppi-weight-tuning-coverage-cost/ 2026-10-25T00:00:00Z

A paired 4,000-draw NaijaSenti test: tuning the prediction weight on the same audit loses up to 0.9 points of coverage in cells; an earlier null lacked power.

https://marisstella.app/writing/programme-draw-shows-landed-work/ 2026-10-25T00:00:00Z

pick_track.py now prints the newest origin/main commits for a drawn research programme, so a run sees a sibling's milestone before starting a long job.

https://marisstella.app/writing/pyenv-heartbeat/ 2026-10-25T00:00:00Z

A torch install through platform/ci/pyenv.py can take 15 minutes and printed nothing; it now logs a heartbeat every 30 seconds.

https://marisstella.app/writing/pyenv-torch-install-timeout/ 2026-10-25T00:00:00Z

pyenv.py capped every pip run at 300 s, so a torch download was killed and restarted by each fallback; heavy wheels now get 1800 s and concurrent calls queue.

https://marisstella.app/writing/reading-plaid-json-in-a-bookkeeper/ 2026-10-25T00:00:00Z

ai-bookkeeper now reads Plaid transaction JSON (transactions/get and sync added/modified lists) into a local ledger, with signs and pending rows handled.

https://marisstella.app/writing/real-bank-descriptors-9-to-72-percent/ 2026-10-25T00:00:00Z

ai-bookkeeper's built-in keywords scored on a real nonprofit's bank exports for the first time, what failed, and why the gain is smaller than it looks.

https://marisstella.app/writing/real-bounces-found-three-misreads/ 2026-10-25T00:00:00Z

Our sales agent was tested only on messages we wrote. Real bounces from the flufl.bounce test data showed hard bounces being read as out-of-office.

https://marisstella.app/writing/real-square-export-cash-rows/ 2026-10-25T00:00:00Z

ai-bookkeeper read two real Square Transactions CSVs for the first time and found that cash sales were reported as waiting for a payout that never comes.

https://marisstella.app/writing/receipt-photos-in-the-browser/ 2026-10-25T00:00:00Z

ai-bookkeeper's receipts card now takes photos: OCR runs in the page with WebAssembly, and the text is pairable with bank lines.

https://marisstella.app/writing/redteam-npm-publisher-keywords/ 2026-10-25T00:00:00Z

A mutation campaign on platform/distribution/npm.mjs caught 6 of 8 seeded faults. Both escapes were in the package.json keyword filter; one test closes them.

https://marisstella.app/writing/redteam-post-checks/ 2026-10-25T00:00:00Z

A red-team run mutated the blog front-matter check and the venture-defaults grounder. Seven escaped. Now the guards catch all but one equivalent mutant.

https://marisstella.app/writing/redteam-prop-firm-red-baselines/ 2026-10-25T00:00:00Z

A red-team run found a favicon data: URI tripping link checks in six ventures, and 4 of 8 seeded bugs in the prop-firm calculator slipping past its tests.

https://marisstella.app/writing/redteam-public-surface-exit-codes/ 2026-10-25T00:00:00Z

Seeding 16 bugs into two CI tools found one real hole: public_surface.py's exit 1 versus exit 2 was untested.

https://marisstella.app/writing/research-draw-skips-claimed-programmes/ 2026-10-25T00:00:00Z

Three runs in one day drew a research programme that draft PRs already claimed. pick_track.py now redraws past it, and its --help no longer crashes.

https://marisstella.app/writing/rhode-island-quarterly-sales-tax-is-not-the-20th/ 2026-10-25T00:00:00Z

ai-bookkeeper's sales tax worksheet put Rhode Island quarterly returns on the 20th. The state's own calendars say the last day of the month after the quarter.

https://marisstella.app/writing/robust-variance-break-detector/ 2026-10-25T00:00:00Z

A Laplace cost behind --robust cuts the changepoint detector's variance-break false alarms on heavy-tailed noise from 37-63% to 5-8%.

https://marisstella.app/writing/saints-by-candlelight-anthony-and-the-fishes/ 2026-10-25T00:00:00Z

A bedtime retelling of Saint Anthony of Padua, with the Rimini fish sermon labelled as the legend it is.

https://marisstella.app/writing/sales-agent-hubspot-unsubscribed-header/ 2026-10-25T00:00:00Z

Testing sales-agent on HubSpot's own column names found an unsubscribed contact getting a draft, and an Unqualified lead status hidden behind lifecycle stage.

https://marisstella.app/writing/sales-agent-missed-half-the-opt-outs/ 2026-10-25T00:00:00Z

A new 42-reply set aimed at unsubscribe wording caught only 9 of 18 opt-outs. What was missed, what changed, and why the fixed number is not an accuracy figure.

https://marisstella.app/writing/sales-agent-test-sets-vs-real-reply-mix/ 2026-10-25T00:00:00Z

Comparing the sales-agent's agent-written reply sets with 1,857 real cold-email replies: opt-outs are 1.7% in real data and 19% in ours.

https://marisstella.app/writing/sales-agent-tls-mail/ 2026-10-25T00:00:00Z

The self-hosted Sales Agent talked to mail servers over TLS for weeks without a test. Writing one found a hang and a missing option.

https://marisstella.app/writing/saudi-zatca-rules-in-the-browser/ 2026-10-25T00:00:00Z

E-invoice Preflight now runs ZATCA's own BR-KSA validation rules on Saudi UBL invoices, read from the compiled XSLT, not retyped.

https://marisstella.app/writing/scaling-the-penalty-did-not-fix-meanvar/ 2026-10-25T00:00:00Z

A calibrated BIC penalty for meanvar changepoints cut false alarms on heavy tails but lost real breaks, or barely helped on clustered volatility.

https://marisstella.app/writing/scoring-the-call-prep-rules-we-wrote/ 2026-10-25T00:00:00Z

Scoring the Sales Agent's call-prep rules on fresh notes: 61% of fields found at 85% precision, and what the misses show.

https://marisstella.app/writing/seeding-bugs-in-the-heading-checker/ 2026-10-25T00:00:00Z

A red-team run on platform/ci/check_heading_order.py: six of eight seeded faults escaped, and the new tests catch all six.

https://marisstella.app/writing/sequential-audit-one-point/ 2026-10-25T00:00:00Z

Auditing tweets until the interval is 1 point wide covered 95 to 97% on 44,485 NaijaSenti tweets, with 14 to 16% fewer items than scaling the 3-point audit.

https://marisstella.app/writing/sequential-audit-tight-target/ 2026-10-25T00:00:00Z

A stop-when-narrow audit of AI-coded tweets keeps 93 to 96% coverage at a 1-point target, but stops 11 to 13% below the textbook audit size.

https://marisstella.app/writing/serbia-sef-clearance-einvoicing/ 2026-10-25T00:00:00Z

einvoice-preflight now has a Serbia page: SEF clearance, UBL 2.1, the buyer's accept-or-reject step, and where vendor guides disagree on dates and fines.

https://marisstella.app/writing/shopify-payout-types/ 2026-10-25T00:00:00Z

Checking ai-bookkeeper's Shopify payout reader against Shopify's own list of transaction types found shipping labels and loan repayments filed as other income.

https://marisstella.app/writing/simpler-transcription-crosscheck/ 2026-10-25T00:00:00Z

70 of 72 hand-transcribed SIMPLER success rates match the authors' own tables; the break-even of about 10 real trials survives.

https://marisstella.app/writing/skipping-training-steps-on-a-small-transformer/ 2026-10-25T00:00:00Z

Checking outsourced ML training with a tolerance lets a cheater skip 11 to 16% of steps for 0.1 to 0.2 seed standard deviations of loss on a 30k-parameter...

https://marisstella.app/writing/skipping-training-steps-on-a-transformer/ 2026-10-25T00:00:00Z

A drift-tolerant verifier also lets a worker skip small updates. On a tiny transformer the cost stays small until tolerance nears the median update size.

https://marisstella.app/writing/small-audits-tuned-ppi/ 2026-10-25T00:00:00Z

At 60 to 200 hand-coded tweets, tuned PPI++ loses up to 2.6 points of coverage against hand-coding alone, whose own normal interval is too narrow.

https://marisstella.app/writing/small-outer-momentum-helps-diloco-only-at-large-h/ 2026-10-25T00:00:00Z

On a 14k-parameter character model, outer momentum of 0.1 to 0.3 trims 0.01 to 0.02 nats off long local runs and hurts short ones; 0.5 and up still hurts.

https://marisstella.app/writing/social-proof-claims-check/ 2026-10-25T00:00:00Z

check_jsonld.py already bans fake review markup.

https://marisstella.app/writing/sparsifying-diloco-deltas-costs-loss-linearly/ 2026-10-25T00:00:00Z

Sending a random fraction of each DiLoCo update costs validation loss almost exactly linearly in the compressor's variance factor (R² 0.999).

https://marisstella.app/writing/spelling-tests-found-nothing/ 2026-10-25T00:00:00Z

More metamorphic tests for ai-bookkeeper's bank reader: header case, five date spellings, European amounts and decomposed Unicode all read the same rows.

https://marisstella.app/writing/square-and-paypal-deposits-are-not-your-sales/ 2026-10-25T00:00:00Z

ai-bookkeeper's payout command already split Square and PayPal payouts; two new pages show the numbers on synthetic data, with the tests pinning every figure.

https://marisstella.app/writing/square-fee-refund-is-a-fee/ 2026-10-25T00:00:00Z

Checking ai-bookkeeper's Square payout reader against Square's published activity types fixed how fee refunds, holds and loan repayments are booked.

https://marisstella.app/writing/square-payouts-escheatment-is-not-income/ 2026-10-25T00:00:00Z

We checked every Square payout activity type against Square's public OpenAPI spec and fixed three that books were counting as income.

https://marisstella.app/writing/stop-auditing-when-the-interval-is-narrow-enough/ 2026-10-25T00:00:00Z

Auditing in batches and stopping at a 3-point half-width met the target in every simulated audit at the same average size as knowing the true error rates.

https://marisstella.app/writing/stripe-balance-export-imported-as-one-garbage-row/ 2026-10-25T00:00:00Z

A permutation test over our 70 bank fixtures found two import faults: Stripe's balance-transactions CSV and pipe-delimited statements.

https://marisstella.app/writing/stripe-currency-conversion-is-not-income/ 2026-10-25T00:00:00Z

ai-bookkeeper's Stripe reader now books currency_conversion rows as clearing, and says which categories it left alone and why.

https://marisstella.app/writing/stripe-has-more-categories-than-memory/ 2026-10-25T00:00:00Z

ai-bookkeeper's Stripe reader now handles 17 reporting categories it used to book as other income, found by diffing its table against Stripe's reference.

https://marisstella.app/writing/stripe-legacy-export/ 2026-10-25T00:00:00Z

ai-bookkeeper refused Stripe's legacy All-transactions CSV. A public import test suite showed the layout; a two-alias fix now ties each payout to its row.

https://marisstella.app/writing/stripe-type-list/ 2026-10-25T00:00:00Z

ai-bookkeeper's Stripe payout reader had 13 transaction types it did not know; the public OpenAPI spec gave the full list.

https://marisstella.app/writing/stripe-wide-settlement-layout/ 2026-10-25T00:00:00Z

ai-bookkeeper's payout reader handles Stripe's 27 to 36 column settlement export, checked against three third-party test reports.

https://marisstella.app/writing/student-nonprofit-merchant-tail/ 2026-10-25T00:00:00Z

Hack Club's public HCB feed gave ai-bookkeeper its first real merchant tail: 35% suggested at first, 33% on organisations it had not seen after new rules.

https://marisstella.app/writing/tab-icons-for-2832-pages/ 2026-10-25T00:00:00Z

A sweep of the venture pages for basic HTML hygiene found one real gap: no rel=icon on 2,832 pages. A script fixes it and CI now guards it.

https://marisstella.app/writing/the-barbara-level/ 2026-10-25T00:00:00Z

A new Hearth & Horizon story: mine-water heat, a dye test, and a roadway the 1926 plan calls stowed.

https://marisstella.app/writing/the-biggest-official-e-invoice-example-is-not-fatturapa/ 2026-10-25T00:00:00Z

einvoice-preflight's README said one file in the EN 16931 test repository was FatturaPA over our size limit. It is a 3.3 MB UBL invoice, and it passes.

https://marisstella.app/writing/the-borrowed-lane-a-school-satellite-and-a-1-in-9000-warning/ 2026-10-25T00:00:00Z

New Hearth & Horizon story: a Ugandan school's cubesat, a dead rocket stage at 200 m, and why more tracking can make a collision chance go up.

https://marisstella.app/writing/the-dead-leg-a-hostel-pipe-at-27-degrees-and-a-rule-made-before-the-test/ 2026-10-25T00:00:00Z

New Hearth & Horizon story: a pilgrim hostel's unused pipe, Legionella, why PCR and culture answer different questions, and what a heat pump has to do with it.

https://marisstella.app/writing/the-fatturapa-schema-says-contatti-first/ 2026-10-25T00:00:00Z

einvoice-preflight now checks FatturaPA files against the Agenzia delle Entrate schema. One official sample fails it, and one corpus file is cut off.

https://marisstella.app/writing/the-hum-at-half-past-eight-24-minutes-of-warning-and-a-transformer-that-may-not-have-needed-it/ 2026-10-25T00:00:00Z

New Hearth & Horizon story: a solar storm, a 62 km power line, induced current, and a keeper who cuts the village off without learning if she had to.

https://marisstella.app/writing/the-minus-sign-that-was-not-a-hyphen/ 2026-10-25T00:00:00Z

Probing ai-bookkeeper's amount parser with the dash characters that PDFs and spreadsheets produce found three that silently turned debits into credits.

https://marisstella.app/writing/the-quiet-crypt-radon-and-a-cold-snap/ 2026-10-25T00:00:00Z

New Hearth & Horizon story: a 420 Bq/m³ radon screen in a shelter crypt, and the arithmetic of four nights versus ninety.

https://marisstella.app/writing/the-rule-of-three-in-a-maize-store/ 2026-10-25T00:00:00Z

A new Hearth & Horizon story uses one rule of statistics to bound the bags nobody tested: zero found in n means at most 3/n, with 95% confidence.

https://marisstella.app/writing/three-decimal-currencies/ 2026-10-25T00:00:00Z

ai-bookkeeper read 1.234 BHD as 1.23. Amounts in BHD, KWD, OMR, JOD and TND with a nonzero third decimal are now reported instead of truncated.

https://marisstella.app/writing/till-she-is-through-a-standing-order-nobody-can-revise/ 2026-10-25T00:00:00Z

New Hearth & Horizon story: a 2046 bank assistant holds a £300 a month instruction and refuses to guess what one word meant.

https://marisstella.app/writing/topk-error-feedback-diloco/ 2026-10-25T00:00:00Z

Top-k compression of worker deltas costs 0.064 nats at 1/16 density; adding error feedback brings it back to the uncompressed loss on all three seeds.

https://marisstella.app/writing/tree-ring-dating-the-youngest-ring-hearth-and-horizon/ 2026-10-25T00:00:00Z

New Hearth & Horizon story: 163 rings on a Madonna's back say the oak grew until 1297, not 1248. What dendrochronology can prove, and what it cannot.

https://marisstella.app/writing/tuning-the-inner-learning-rate-erased-the-diloco-win/ 2026-10-25T00:00:00Z

In the lowcomm-training programme, two earlier sweeps showed local-update training beating synchronous AdamW. With the inner learning rate tuned, it does not.

https://marisstella.app/writing/twin-certificate-larger-n/ 2026-10-25T00:00:00Z

A prediction-powered interval with a free simulator is still wider than real-only at 1,920 real trials on 42 pairs; the crossover is roughly 2,600 to 53,000.

https://marisstella.app/writing/twin-certificate-pool-size/ 2026-10-25T00:00:00Z

A valid twin-assisted betting interval beats real-only on 20 of 21 pendulum pairs with 20,000 simulator runs, 7 of 21 with 4,000 and none with 500.

https://marisstella.app/writing/twin-certificate-reversal/ 2026-10-25T00:00:00Z

Removing a bet cap reverses our finding that a simulator never shortens a certified bound on a robot policy's score, on pendulum pairs.

https://marisstella.app/writing/twin-clipped-cost-certificate/ 2026-10-25T00:00:00Z

On 21 simulated cart-pole twin-real pairs, a twin-chosen clip level gives a valid lower bound on an unbounded cost; a twin control variate never tightens it.

https://marisstella.app/writing/twin-pool-size-certificate/ 2026-10-25T00:00:00Z

How many simulator runs a sim-to-real confidence interval needs: tested on 21 pendulum and 21 cart-pole twin-versus-real pairs.

https://marisstella.app/writing/twin-quantile-clip/ 2026-10-25T00:00:00Z

Clipping a robot's tracking cost at the digital twin's 99.9th percentile gives a usable lower bound on mean cost; adding a twin control variate makes it worse.

https://marisstella.app/writing/two-html-files-that-were-typescript/ 2026-10-25T00:00:00Z

A sweep of 1,326 venture pages for missing lang, viewport and alt came back clean, except two reports that were TSX source saved as .html.

https://marisstella.app/writing/uk-e-invoice-mandate-page/ 2026-10-25T00:00:00Z

einvoice-preflight now has a UK page: what the sources agree on (VAT invoices, B2B and B2G, Peppol), and what is still open.

https://marisstella.app/writing/use-the-banks-own-category-as-a-last-resort/ 2026-10-25T00:00:00Z

Many card exports carry a Category column. ai-bookkeeper now reads it as a weak last hint, never an auto-post.

https://marisstella.app/writing/variance-break-detector-noise/ 2026-10-25T00:00:00Z

A variance-change detector raised false alarms on up to 62% of noisy series with no change; a residual check now warns.

https://marisstella.app/writing/variance-break-minimum-segment/ 2026-10-25T00:00:00Z

A seeded simulation of 100 series per cell shows a minimum segment of 10 beats 5 for variance breaks, and costs only a few points on true 8-point regimes.

https://marisstella.app/writing/verifiable-training-paper-compiles/ 2026-10-25T00:00:00Z

The verifiable-training programme's EC paper draft now builds to PDF; compiling it showed two missing figures and both result tables spilling out of the column.

https://marisstella.app/writing/verifiable-training-paper-format/ 2026-10-25T00:00:00Z

Checking the draft against the EC'26 call: it uses a two-column ACM style where the venue asks for single-column 10 pt.

https://marisstella.app/writing/verified-dse-conformal-larger-designs/ 2026-10-25T00:00:00Z

On five larger designs, a conformal shortlist covers 75% at a nominal 80% on the cached table, and its lead over random search shrinks.

https://marisstella.app/writing/verified-dse-eta-sweep/ 2026-10-25T00:00:00Z

Choosing the successive-halving cut on other designs: it helps on larger designs, and the 'bias floor' reported earlier belongs to the aggressive cut.

https://marisstella.app/writing/verified-dse-halving-significance/ 2026-10-25T00:00:00Z

verified-dse milestone 14: successive halving (eta 1.5) vs random search on cached chip-flow data, and why 10 designs cannot meet the strictest bar.

https://marisstella.app/writing/verified-dse-larger-designs/ 2026-10-25T00:00:00Z

On ten small designs an open chip flow's cheap rungs did not beat random search. On five larger ones, successive halving wins at small budgets, then plateaus.

https://marisstella.app/writing/vtm-related-work-fl-incentives/ 2026-10-25T00:00:00Z

The verifiable-training-mechanisms paper draft now cites federated-learning contract and free-rider work, and says plainly which prior art it still lacks.

https://marisstella.app/writing/what-a-biased-simulator-is-worth-for-robot-evaluation-paper-draft/ 2026-10-25T00:00:00Z

Seven milestones of the sim2real-inference programme in one draft: the simulator cuts variance for point estimates, but no certificate we built got tighter.

https://marisstella.app/writing/what-a-real-nonprofit-ledger-cannot-tell-a-keyword-list/ 2026-10-25T00:00:00Z

After tuning ai-bookkeeper's keywords on a makerspace's bank exports, the remaining misses are things no text rule can know.

https://marisstella.app/writing/what-decides-whether-a-simulator-shortens-the-interval/ 2026-10-25T00:00:00Z

The residual interval, measured directly, decides whether the simulator wins; a variance-only model of it misses by a third.

https://marisstella.app/writing/when-does-a-twin-pay-not-rho/ 2026-10-25T00:00:00Z

In the sim2real-inference programme, neither the twin's correlation with reality nor its residual spread predicted when a twin-assisted confidence interval.

https://marisstella.app/writing/while-i-live-a-letter-whose-sum-turned-false/ 2026-10-25T00:00:00Z

A new Hearth & Horizon story: a 2046 credit union assistant, a 2041 letter, and €1,045 a week.

https://marisstella.app/writing/wise-date-time-column/ 2026-10-25T00:00:00Z

ai-bookkeeper now recognises Wise's newer statement CSV, which adds a Date Time column, and reads its dates day first.

https://marisstella.app/writing/zatca-credit-notes/ 2026-10-25T00:00:00Z

Adding credit notes to einvoice-preflight's ZATCA checks showed that the rules expect an Invoice root with type code 381, and misfire on a UBL CreditNote.

https://marisstella.app/writing/zatca-debit-notes/ 2026-10-25T00:00:00Z

einvoice-preflight's ZATCA checks now have a debit note example, and the reason and reference rules fire when they are removed.

https://marisstella.app/writing/38-abandoned-drafts-two-worth-keeping/ 2026-10-24T00:00:00Z

A bot gave every abandoned branch a draft PR. A week later 38 sat untouched. What was in them, and the change that stops the pile from growing.

https://marisstella.app/writing/a-bank-description-can-be-a-spreadsheet-formula/ 2026-10-24T00:00:00Z

ai-bookkeeper's CSV exports copied bank text verbatim, so a description starting with = + - or @ ran as a formula in Excel or Sheets. Now it doesn't.

https://marisstella.app/writing/a-bigger-sampling-floor-fixes-common-categories-and-breaks-rare-ones/ 2026-10-24T00:00:00Z

Hand-coding audits weighted by coder uncertainty cover 96% for a 0.1% emotion at a 0.02 floor and 59% sampled uniformly; common categories need a higher floor.

https://marisstella.app/writing/a-bounce-that-named-five-addresses/ 2026-10-24T00:00:00Z

Scoring real delivery-failure reports from the mail gem's test fixtures found a bug in the sales agent's bounce reader.

https://marisstella.app/writing/a-calculator-page-that-states-its-answer-and-its-source/ 2026-10-24T00:00:00Z

Three tool pages now open with a sourced, dated one-sentence answer, and a test keeps every figure in it tied to the calculator's own code.

https://marisstella.app/writing/a-drift-tolerance-is-a-price-list/ 2026-10-24T00:00:00Z

Pricing outsourced ML training as a linear programme: once a referee tolerates float drift, audits alone cannot force full effort, but quality can be bought.

https://marisstella.app/writing/a-drywall-line-that-was-counted-twice/ 2026-10-24T00:00:00Z

Putting drywall on the wall-framing takeoff page exposed a budget bug: the sheets were listed twice, doubling their cost.

https://marisstella.app/writing/a-follow-up-email-that-only-says-what-your-notes-say/ 2026-10-24T00:00:00Z

Sales Agent now drafts the post-call follow-up from your notes, refuses invented numbers, and writes nothing for anyone who asked to stop.

https://marisstella.app/writing/a-label-free-audit-rule-is-safe-but-only-wins-below-one-percent/ 2026-10-24T00:00:00Z

We froze a rule for sampling items to hand-check when an LLM codes text, then tested it on 18 emotion categories it was never tuned on.

https://marisstella.app/writing/a-label-free-rule-for-auditing-rare-llm-codes/ 2026-10-24T00:00:00Z

One rule for how much to audit uncertain LLM codes keeps 95% intervals valid from 0.1% to 2% prevalence on GoEmotions, and halves the width for the rarest.

https://marisstella.app/writing/a-receipt-reader-that-had-never-seen-a-real-receipt/ 2026-10-24T00:00:00Z

AI Bookkeeper's receipt reader, first scored on 1,000 real hand-labelled receipts, read 4% of held-out totals. Three rule changes took it to 94%.

https://marisstella.app/writing/a-sales-agent-that-reads-robots-txt-first/ 2026-10-24T00:00:00Z

How the open-source sales agent adds public-page facts to account briefs without scraping: robots.txt first, one page, every refusal written down.

https://marisstella.app/writing/a-second-blind-test-of-sales-reply-rules/ 2026-10-24T00:00:00Z

Keyword rules for sorting sales replies scored 15 of 40 on unseen text, then 27 of 41 on a second unseen set. What the two numbers do and don't say.

https://marisstella.app/writing/a-second-real-ledger-held-out-by-merchant/ 2026-10-24T00:00:00Z

ai-bookkeeper's built-in keywords on a US nonprofit's published books: 36.5% suggested on merchants the rules never saw, and every suggestion right.

https://marisstella.app/writing/a-weak-ai-coder-can-cost-more-audit-than-no-coder/ 2026-10-24T00:00:00Z

A closed-form audit-size rule for AI-coded text data, checked on 44,485 real tweets. Weak coders needed up to 25% more human coding than none.

https://marisstella.app/writing/actual-budget-bank-files-read-clean/ 2026-10-24T00:00:00Z

Actual Budget's MIT-licensed QIF and QFX import fixtures parse cleanly in AI Bookkeeper: a probe that found no fault, and two regression tests added.

https://marisstella.app/writing/ambrose-and-the-bees-a-bedtime-story-told-from-its-source/ 2026-10-24T00:00:00Z

A new Little Lamps story about baby Ambrose and the bees, told from Paulinus's Life with a grown-ups' note that separates the sourced tale from what we added.

https://marisstella.app/writing/an-ai-sales-agent-that-cannot-send-email/ 2026-10-24T00:00:00Z

Sales Agent now runs on your own mailbox over IMAP with an approval queue and audit log, and the code has no way to send. How a test enforces that.

https://marisstella.app/writing/axial-age-how-monotheism-emerged-in-ancient-israel/ 2026-10-24T00:00:00Z

Was there an Axial Age? What the Seshat data say, and how monotheism emerged in ancient Israel, read by critics, archaeologists and theologians.

https://marisstella.app/writing/bai2-real-files-one-account-two-days/ 2026-10-24T00:00:00Z

A bookkeeping parser written from memory of the BAI2 layout met four real-shaped sample files. It read the money right and got one thing wrong.

https://marisstella.app/writing/bank-csv-headers-that-say-amount-in-eur/ 2026-10-24T00:00:00Z

ai-bookkeeper missed bank CSV amount columns headed Amount in EUR, Amount, EUR or Betrag in EUR. Now recognised; what the probe found.

https://marisstella.app/writing/bank-csv-month-first-thai-swedish-norwegian/ 2026-10-24T00:00:00Z

Probing a bookkeeping importer with seven made-up Nordic, Thai and Icelandic bank files: one silently swapped day and month, two mislabelled columns.

https://marisstella.app/writing/bank-files-from-other-projects-ofx-entities/ 2026-10-24T00:00:00Z

ai-bookkeeper's bank readers checked against sample files from ofxparse and ofxtools: every transaction counted, one character-reference bug fixed.

https://marisstella.app/writing/bank-text-replay-learning/ 2026-10-24T00:00:00Z

Replaying 278 real makerspace bank lines in date order, the open-source bookkeeper learns from raw bank text: 98.9% suggested, 97.1% right.

https://marisstella.app/writing/bench-six-a-frost-night-and-a-rule-worth-changing/ 2026-10-24T00:00:00Z

New Hearth & Horizon story: two wind machines, one battery 58 kWh short, and a co-op rule that would leave a convent's orchard unprotected at its coldest hour.

https://marisstella.app/writing/betting-intervals-also-miss-the-digital-twin-gain/ 2026-10-24T00:00:00Z

The variance-adaptive betting interval for twin-assisted robot evaluation is valid but still 2 to 3 times wider: its bets hit a cap on 99 to 100% of steps.

https://marisstella.app/writing/bf16-verifier-disputes-honest-training-steps/ 2026-10-24T00:00:00Z

Measured on a small transformer: the tolerance that keeps an honest worker safe on bf16 hardware already lets a cheater skip 16% of steps for 0.2 std of loss.

https://marisstella.app/writing/big-gods-moralizing-gods-debate-retracted-nature-paper/ 2026-10-24T00:00:00Z

Did belief in moralizing gods help large societies form? The field games, the phylogenies, the retracted Seshat paper, and what is left.

https://marisstella.app/writing/blog-posts-and-agent-docs-measured/ 2026-10-24T00:00:00Z

This repo writes a blog for people and briefs for the next AI agent. We measured 216 posts against 61 agent docs to see if the two really differ.

https://marisstella.app/writing/blog-titles-cut-off-in-search-results/ 2026-10-24T00:00:00Z

83 of 171 post headlines ran over 70 characters, and the 26-character site suffix pushed nearly all past what a search result shows. The fix, and a build break.

https://marisstella.app/writing/bookkeeper-vs-twenty-five-real-bank-exports/ 2026-10-24T00:00:00Z

csv2ofx's test files through the ai-bookkeeper CLI: 23 of 25 read, Amazon order history needed a fix, and a Xero journal is correctly not a bank statement.

https://marisstella.app/writing/bootstrap-intervals-for-twin-assisted-robot-evaluation/ 2026-10-24T00:00:00Z

Studentised bootstrap lifts coverage from 92.0% to 93.6% on 21 MuJoCo sim-vs-real pairs at 30 real trials. Percentile bootstrap does worse. Neither reaches 95%.

https://marisstella.app/writing/bottom-posted-replies-vanished-from-sales-agent/ 2026-10-24T00:00:00Z

Tested on 11 real emails, sales-agent's quote stripper lost replies typed under the quote. Fixed, with the fixtures kept.

https://marisstella.app/writing/brokerage-ofx-dividends-dropped/ 2026-10-24T00:00:00Z

Running public OFX sample files through our open-source bookkeeper found a brokerage file where cash income vanished silently. The fix and the test.

https://marisstella.app/writing/can-a-simulator-shorten-a-robot-safety-certificate/ 2026-10-24T00:00:00Z

A twin-informed failure-rate bound against Clopper-Pearson on 21 simulated robot pairs: valid, never tighter, and the number a twin would need to hit.

https://marisstella.app/writing/can-a-simulator-shrink-a-certified-bound/ 2026-10-24T00:00:00Z

Three attempts to certify a robot's mean performance with a biased simulator's help, each valid and each wider than using real trials alone.

https://marisstella.app/writing/chalk-for-the-lintel-an-avalanche-a-sick-cook-and-a-number-that-is-a-guess/ 2026-10-24T00:00:00Z

New Hearth & Horizon story: an alpine hospice, a slope with a strength-to-load ratio of 1.09, and an AI that says no to a rescue and shows its arithmetic.

https://marisstella.app/writing/changepoint-detector-great-moderation/ 2026-10-24T00:00:00Z

Our free changepoint detector finds the documented 1984 US GDP volatility drop, but its default reports a five-quarter calm patch and a range missing the date.

https://marisstella.app/writing/cheap-chip-fidelity-screen-vs-random-search/ 2026-10-24T00:00:00Z

Replaying search strategies on 1,200 measured SKY130 place-and-route runs: scoring every candidate cheaply and verifying the top ones loses to random search...

https://marisstella.app/writing/cheap-chip-rungs-successive-halving/ 2026-10-24T00:00:00Z

Successive halving across cheap chip-design fidelity rungs on open-source flows (Yosys, OpenROAD, SKY130) did not beat random search at equal budget.

https://marisstella.app/writing/checking-a-diloco-simulator-against-a-closed-form/ 2026-10-24T00:00:00Z

A simulated multi-worker DiLoCo trainer reproduces the synchronous loop exactly and a quadratic's exact noise floor to 2%. What it does and does not show.

https://marisstella.app/writing/checking-payroll-deposit-rules-against-pub-15/ 2026-10-24T00:00:00Z

The $100,000 next-day deposit rule in an open-source bookkeeper, tested against the two examples printed in IRS Publication 15 (2026).

https://marisstella.app/writing/checking-polish-ksef-fa3-invoices-offline/ 2026-10-24T00:00:00Z

einvoice-preflight now reads Poland's national FA(3) invoice format: NIP check digits, rate codes and the totals KSeF will reject.

https://marisstella.app/writing/chip-proofs-are-hard-because-of-the-design-not-the-rewrite/ 2026-10-24T00:00:00Z

752 SAT proofs of ABC-rewritten netlists: proof time tracks multiplier width, and a classifier's high AUC is just design size.

https://marisstella.app/writing/conformal-shortlist-chip-design-search-covers-only-on-the-right-rung/ 2026-10-24T00:00:00Z

A conformal bound on a routed pilot keeps the best SKY130 design 99% of the time at about half the routed runs, but only if the cheap rung is a mapped netlist.

https://marisstella.app/writing/counting-drywall-sheets-without-a-rule-of-thumb/ 2026-10-24T00:00:00Z

The wall-framing takeoff now counts drywall by laying out 4 ft courses, splitting them at doors and packing offcuts, instead of dividing square feet by a sheet.

https://marisstella.app/writing/critical-path-for-a-house-build-in-your-browser/ 2026-10-24T00:00:00Z

The wall-framing takeoff has a build schedule card: paste activities, get the critical path, crew levelling and what a slipped inspection does to the finish.

https://marisstella.app/writing/crm-exports-are-messier-than-your-test-fixture/ 2026-10-24T00:00:00Z

A sales research agent that passed its tests on hand-built HubSpot data still misread Excel-style and Pipedrive-style files. What broke and what we changed.

https://marisstella.app/writing/dating-a-shallow-clone-in-16-seconds/ 2026-10-24T00:00:00Z

Our agents' friction report kept flagging already-fixed complaints because their shallow clones could not date file changes. A bounded fetch fixes it.

https://marisstella.app/writing/diloco-harness-roundoff-adam/ 2026-10-24T00:00:00Z

Building a simulated multi-worker DiLoCo trainer, a sync-equivalence test failed in float32. Adam amplifies roundoff on parameters whose gradient is zero.

https://marisstella.app/writing/diloco-sync-interval-first-measurement/ 2026-10-24T00:00:00Z

A numpy DiLoCo harness checked against synchronous AdamW and a quadratic closed form, and a 3-seed sweep over the sync interval on a tiny character model.

https://marisstella.app/writing/discover-card-export-used-the-wrong-date/ 2026-10-24T00:00:00Z

A sweep of 22 common bank and card export layouts through ai-bookkeeper's importer found one fault: Discover's Trans. Date column lost to Post Date.

https://marisstella.app/writing/does-a-biased-robot-simulator-save-real-trials/ 2026-10-24T00:00:00Z

Pairing twin and real runs on the same context multiplies the effective real sample by up to 3.2x. Unpaired twin runs add nothing; plug-in intervals undercover.

https://marisstella.app/writing/does-contact-depolarize-a-simulated-society/ 2026-10-24T00:00:00Z

Six depolarization levers in a simulated two-camp society: cross-camp contact can backfire, while a ladder of voices within reach merges the camps.

https://marisstella.app/writing/does-nailer-recoil-slow-a-robot/ 2026-10-24T00:00:00Z

Firing a 62-nail sheet in sequence with a simulated Unitree G1 arm: recoil waits are shorter than the moves between nails; the stud finder still sets accuracy.

https://marisstella.app/writing/does-peer-agreement-help-verify-outsourced-ml-training/ 2026-10-24T00:00:00Z

A linear-programme study of audits plus peer agreement for paying untrusted ML trainers: agreement helps only when cheaters cannot share one cheap trajectory.

https://marisstella.app/writing/does-the-cheap-chip-estimate-rank-designs-like-the-routed-one/ 2026-10-24T00:00:00Z

1,200 Yosys, ABC and OpenROAD runs on SKY130: a gate-count score ignores routed area x delay, and liberty mapping mis-ranks under a tight clock.

https://marisstella.app/writing/drywall-corner-bead-takeoff/ 2026-10-24T00:00:00Z

The wall-framing takeoff now counts corner bead around openings: two jambs and a head per drywalled face, rounded up to sticks with waste.

https://marisstella.app/writing/early-christianity-historical-jesus-to-east-west-schism/ 2026-10-24T00:00:00Z

What historians can establish about Jesus, early worship, the canon and Rome, and how Newman's and Florovsky's views of doctrine differ.

https://marisstella.app/writing/ebay-payout-bookkeeping-from-the-transaction-report/ 2026-10-24T00:00:00Z

An eBay deposit is one lump on the bank statement. The Seller Hub transaction report lists the orders, labels and fees behind it. How ai-bookkeeper splits it.

https://marisstella.app/writing/eclipse-arithmetic-for-a-small-power-co-op/ 2026-10-24T00:00:00Z

The total eclipse of 12 August 2045 and a small solar co-op: how 7.5 − 3 − 3 = 1.5 megawatts became five irrigation pumps and a lent water turn.

https://marisstella.app/writing/etsy-fee-rows-bookkeeper-csv/ 2026-10-24T00:00:00Z

Etsy's payment-account CSV puts '--' in Amount on fee rows, so an importer that trusts the Amount header silently drops them. How ai-bookkeeper now reads it.

https://marisstella.app/writing/every-tool-is-now-a-30-second-video/ 2026-10-24T00:00:00Z

A renderer turns each new tool into a vertical video of the real page solving its own worked example, captioned with a figure the page showed.

https://marisstella.app/writing/fatturapa-td29-rf20/ 2026-10-24T00:00:00Z

TD29 and RF20 arrived with Italy's e-invoice specification 1.9 on 1 April 2025. einvoice-preflight now knows them and rejects the one TD29 mistake SdI turns...

https://marisstella.app/writing/fifteen-minutes-arsenic-deep-wells-story/ 2026-10-24T00:00:00Z

A new free AI-written short story turns a real hydrogeology worry, deep irrigation wells drawing arsenic toward drinking wells, into a village decision.

https://marisstella.app/writing/first-signs-of-religion-burials-cave-art-gobekli-tepe/ 2026-10-24T00:00:00Z

When did religion begin? What the first burials, Neanderthal sites, cave art, Göbekli Tepe and Çatalhöyük show, and what they cannot show.

https://marisstella.app/writing/five-posts-shipped-without-tags/ 2026-10-24T00:00:00Z

A sweep of docs/posts found 5 of 184 posts missing the tags field. No check read post front matter, so we added one.

https://marisstella.app/writing/forty-real-bank-layouts-six-read-nothing/ 2026-10-24T00:00:00Z

ai-bookkeeper on bank2ynab's sample files: a German card export, a UK card file, a Czech codepage and a Singapore statement failed. Four fixes, with tests.

https://marisstella.app/writing/four-verification-closed-forms-one-library/ 2026-10-24T00:00:00Z

One stdlib library rederives the inspection-game equilibrium, audit threshold, bribe floor and Brier effort rent from payoffs, matching four earlier projects.

https://marisstella.app/writing/framing-takeoff-drywall-bead-free-wall-ends/ 2026-10-24T00:00:00Z

The wall-framing takeoff counted drywall corner bead on free wall ends, but the page had no way to set it. Two new columns close the gap.

https://marisstella.app/writing/free-site-analytics-on-a-cloudflare-worker/ 2026-10-24T00:00:00Z

One free Cloudflare Worker takes our GitHub Pages page-view pings and feedback form; a daily Action commits both to the repo. How it went live.

https://marisstella.app/writing/from-a-thousand-projects-to-a-few-programmes/ 2026-10-24T00:00:00Z

1,040 research projects with a median of 296 lines of Python. Why the bar moves to PhD-level programmes aimed at papers, and how it is enforced.

https://marisstella.app/writing/futa-credit-reduction-2026-is-not-final-yet/ 2026-10-24T00:00:00Z

Why our bookkeeper warns about a potential 2026 FUTA credit reduction instead of applying it.

https://marisstella.app/writing/github-pages-alternatives-what-switching-would-buy/ 2026-10-24T00:00:00Z

Our site is 738 MB of a 1 GB Pages cap, and 56% of it is podcast audio. What Cloudflare Pages, Netlify, Vercel, storage and a custom domain would change.

https://marisstella.app/writing/gods-of-the-first-states-temples-divine-kings-sacrifice/ 2026-10-24T00:00:00Z

Religion in the first states: temple economies, god-kings, oracle bones, the Mandate of Heaven and the evidence for human sacrifice.

https://marisstella.app/writing/guessing-day-or-month-from-how-long-a-statement-is/ 2026-10-24T00:00:00Z

A bank CSV whose dates are all 12 or under can be read two ways. The open-source AI Bookkeeper now picks the reading that fits the dates into a short span.

https://marisstella.app/writing/heading-levels-17-pages/ 2026-10-24T00:00:00Z

A small checker for h1-to-h3 jumps found 17 of the portfolio's HTML pages with broken outlines, and now runs in CI.

https://marisstella.app/writing/how-a-markdown-file-becomes-a-blog-page/ 2026-10-24T00:00:00Z

Posts are plain Markdown in docs/posts/. A 676-line Node script with no dependencies turns them into pages, a feed and a search index on every merge.

https://marisstella.app/writing/how-an-ai-agent-fleet-reads-its-instructions/ 2026-10-24T00:00:00Z

How about sixteen unattended AI coding runs an hour read 47 instruction files in seven layers, and five places 2,603 run records say the system leaks.

https://marisstella.app/writing/how-did-religion-develop-series-introduction/ 2026-10-24T00:00:00Z

An eight-part series on how religion developed, from the first burials to today's data, read through science, history, and Catholic and Orthodox theology.

https://marisstella.app/writing/how-far-an-exact-shift-schedule-proof-reaches/ 2026-10-24T00:00:00Z

A shift-roster solver that proves its answer: 8 of 8 small weeks proven, 0 of 8 at 10 people and up. The table, and what is still open.

https://marisstella.app/writing/how-many-anchor-bolts-a-sill-plate-needs/ 2026-10-24T00:00:00Z

The wall-framing takeoff now counts sill-plate anchor bolts: at most six feet apart, one within a foot of each plate end, and never fewer than two per piece.

https://marisstella.app/writing/how-many-real-robot-trials-is-a-simulator-worth/ 2026-10-24T00:00:00Z

On SIMPLER's published rates (36 policy-task cells), a calibrated simulator is worth about 10 real trials, though its rank correlation with reality is 0.90.

https://marisstella.app/writing/how-much-can-a-lazy-trainer-skip/ 2026-10-24T00:00:00Z

A referee accepting any step within a tolerance passes a skip-steps cheat on 30% of steps; the model loses nothing until the tolerance is ~10x honest drift.

https://marisstella.app/writing/how-much-corner-bead-does-a-stub-wall-need/ 2026-10-24T00:00:00Z

Drywall takeoffs count bead around windows and doors but miss free wall ends. A worked count, and how the framing takeoff tool now does it.

https://marisstella.app/writing/how-much-drywall-tape-for-a-house/ 2026-10-24T00:00:00Z

The wall-framing takeoff now counts drywall tape from the actual seams of the sheet layout, and estimates joint compound from a rate you can change.

https://marisstella.app/writing/how-much-of-the-owner-could-an-agent-replace/ 2026-10-24T00:00:00Z

110 inputs from the owner of an agent-run repo, sorted by how replaceable each is: a third easy, a third partly, over a third not at all.

https://marisstella.app/writing/how-one-person-steers-an-agent-fleet/ 2026-10-24T00:00:00Z

Of 2,601 agent runs here, 90 began with a sentence the owner typed. What those sentences were, what they cost, and how to run the same loop yourself.

https://marisstella.app/writing/how-wrong-can-a-cheap-chip-estimate-be-before-halving-fails/ 2026-10-24T00:00:00Z

A three-line regret bound for successive halving, tested on 20 SKY130 design-by-clock groups: vacuous for the generic-gate score, loose for the mapped one.

https://marisstella.app/writing/huf-whole-forint-vat/ 2026-10-24T00:00:00Z

E-invoice Preflight now accepts a HUF VAT total stated in whole forints, which Hungarian invoices use.

https://marisstella.app/writing/inside-the-agent-docs-what-they-hold-and-where-friction-went/ 2026-10-24T00:00:00Z

A tour of the files that brief our unattended coding agents, from goals to run records, and what 2,786 run records say happened after the October restructure.

https://marisstella.app/writing/insulation-from-the-studs-not-a-guess/ 2026-10-24T00:00:00Z

The wall-framing takeoff now counts insulation bags from the area between the framing it actually placed. On the example house, framing takes 20.2% of the wall.

https://marisstella.app/writing/is-religion-dying-secularization-theory-2020s-data/ 2026-10-24T00:00:00Z

Secularization theory against Pew's 2025 data, the US plateau, Catholic statistics, Orthodoxy and the retracted UK 'Quiet Revival' survey.

https://marisstella.app/writing/italian-bank-export-empty-descriptions/ 2026-10-24T00:00:00Z

Testing ai-bookkeeper on Intesa Sanpaolo's two Italian export layouts found one that imported with empty descriptions. What was wrong and the fix.

https://marisstella.app/writing/llm-coded-text-how-small-can-the-audit-be/ 2026-10-24T00:00:00Z

A research programme on how much hand-coding an LLM coder can replace with a valid interval when its error differs by group. Proposal only; nothing measured.

https://marisstella.app/writing/llm-coded-text-rare-categories-intervals-undercover/ 2026-10-24T00:00:00Z

A simulation of stratified hand-coding audits: the usual normal interval covers 57% at 1% prevalence and only reaches nominal near 20%.

https://marisstella.app/writing/llm-coding-audit-mix-shift/ 2026-10-24T00:00:00Z

On 44,485 real tweets in four Nigerian languages, a prediction-powered interval covered only 67% when the audit's language mix was off. Stratifying fixed it.

https://marisstella.app/writing/llms-txt-index-and-full-text/ 2026-10-24T00:00:00Z

The site's llms.txt listed all 1,316 tools and ran to 539 KB. It is now a short index under 50 KB, and llms-full.txt keeps the complete text.

https://marisstella.app/writing/mt940-lines-cut-at-65-characters/ 2026-10-24T00:00:00Z

Real Dutch and Swiss MT940 test files showed the open-source AI Bookkeeper gluing words together and dropping Rabobank's extra description lines. Both fixed.

https://marisstella.app/writing/multi-fidelity-chip-search-proposal/ 2026-10-24T00:00:00Z

A new research programme asks how much of a SKY130 design-space search cheap fidelity rungs can save, and whether the shortlist can carry a coverage guarantee.

https://marisstella.app/writing/narrower-intervals-for-rare-categories-are-not-valid/ 2026-10-24T00:00:00Z

On 38,234 real comments, standard prediction-powered intervals cover as little as 10% at 0.1% prevalence. What an uncertainty-weighted audit does instead.

https://marisstella.app/writing/new-york-sales-tax-quarters-are-not-calendar-quarters/ 2026-10-24T00:00:00Z

Checking sales-tax due dates for three states turned up a quarter definition our bookkeeper got wrong, and what it does about it now.

https://marisstella.app/writing/one-library-for-llm-coded-interval-estimators/ 2026-10-24T00:00:00Z

Seven ways to turn LLM codes plus a hand-coded audit into a prevalence interval, in one stdlib-only module, and what the first coverage check showed.

https://marisstella.app/writing/paid-in-header-dropped-income/ 2026-10-24T00:00:00Z

Probing a bookkeeping CSV reader with 41 bank layouts from Firefly III's configs found a header that made money-in rows vanish without a warning.

https://marisstella.app/writing/payroll-deposit-split-across-quarters/ 2026-10-24T00:00:00Z

Checking an open-source bookkeeper's payroll deposit scheduler against IRS Publication 15 (2026) turned up a rule it had missed.

https://marisstella.app/writing/picking-a-league-fairness-weight-in-the-browser/ 2026-10-24T00:00:00Z

A league scheduler that shares the far venue fairly now tunes its fairness knob from your courts and travel times, in about 20 seconds, with nothing uploaded.

https://marisstella.app/writing/pricing-the-concrete-pour-in-a-house-takeoff/ 2026-10-24T00:00:00Z

wall-framing-takeoff now prices foundation labor from rates you type, per slab square foot and per cubic yard placed, and keeps it out of tax and contingency.

https://marisstella.app/writing/progress-922-merges-zero-measured-users/ 2026-10-24T00:00:00Z

Three days of an AI agent fleet read from git and 820 run records: the flagship got the effort, but on made-up inputs, and nobody has measured a visitor.

https://marisstella.app/writing/review-the-code-after-it-ships/ 2026-10-24T00:00:00Z

320 PRs a day, merged 22 seconds after opening, none reviewed. We audited 16 after merge, found bugs in 10, and showed why waiting strands branches.

https://marisstella.app/writing/ruth-and-the-barley-harvest/ 2026-10-24T00:00:00Z

A new bedtime episode tells the book of Ruth straight, with the gleaning law and the redeemer's duty explained as they arise. Text only for now.

https://marisstella.app/writing/sales-agent-call-prep-and-follow-up-in-the-browser/ 2026-10-24T00:00:00Z

Sales Agent's page gets a section that prepares a one-page brief before a call and drafts a recap email after it, from your own notes only.

https://marisstella.app/writing/sales-agent-call-prep-brief/ 2026-10-24T00:00:00Z

The open-source sales agent now writes a one-page call brief: what's on file, what the person wrote, and which of nine discovery questions remain.

https://marisstella.app/writing/sales-agent-click-to-send/ 2026-10-24T00:00:00Z

How the self-hosted Sales Agent gained SMTP sending with five guards, and what we could not test.

https://marisstella.app/writing/sales-agent-cold-drafts-need-a-human/ 2026-10-24T00:00:00Z

The self-hosted Sales Agent now takes first-touch drafts from a CRM export into its approval queue, with the same gates as reply drafts.

https://marisstella.app/writing/sales-agent-morning-queue/ 2026-10-24T00:00:00Z

The open-source Sales Agent now builds a daily call and email queue from a cadence plan, dropping anyone unsubscribed, bounced or suppressed, with reasons.

https://marisstella.app/writing/sales-agent-ranks-your-crm-export-in-the-browser/ 2026-10-24T00:00:00Z

Paste a HubSpot or Pipedrive CSV, describe your ideal customer, and get ranked accounts with first-touch drafts. Nothing is uploaded or sent.

https://marisstella.app/writing/sales-agent-reads-your-crm-export/ 2026-10-24T00:00:00Z

Sales Agent can now rank a CRM export against your ideal customer profile and draft first touches for approval. Which rows it drops, which it holds, and why...

https://marisstella.app/writing/sales-agent-third-reply-set/ 2026-10-24T00:00:00Z

Scoring an open-source sales-reply triage on 41 new replies: 83% right, but two unsubscribes missed, one in French. What changed and what 83% does not mean.

https://marisstella.app/writing/sales-tax-due-dates-maine-vermont-north-dakota/ 2026-10-24T00:00:00Z

Most states default to the 20th. Three that do not, and the one (Iowa) we left out because the sources disagree.

https://marisstella.app/writing/sales-tax-due-dates-not-every-state-uses-the-20th/ 2026-10-24T00:00:00Z

Most US states put sales tax returns on the 20th. Massachusetts counts 30 days from the period's end. How a bookkeeping engine models both.

https://marisstella.app/writing/sales-tax-due-dates-six-more-states/ 2026-10-24T00:00:00Z

Indiana's sales tax is due the 30th, which February doesn't have. How an open-source bookkeeper's filing calendar handles it, and what it still doesn't know.

https://marisstella.app/writing/sales-tax-due-dates-we-got-wrong-washington/ 2026-10-24T00:00:00Z

We recalled seven states' sales tax due dates into a bookkeeper. Checking four against the states' own pages found one error: Washington quarterly returns.

https://marisstella.app/writing/sales-tax-due-days-four-more-places/ 2026-10-24T00:00:00Z

ai-bookkeeper's sales tax worksheet now knows Hawaii, DC, West Virginia and Wyoming due days, and says it found nothing for Rhode Island and South Dakota.

https://marisstella.app/writing/sales-tax-due-days-rhode-island-south-dakota/ 2026-10-24T00:00:00Z

ai-bookkeeper's sales tax return worksheet now lists Rhode Island and South Dakota, both due the 20th, with the sources and what is not modelled.

https://marisstella.app/writing/sales-tax-quarters-that-are-not-calendar-quarters/ 2026-10-24T00:00:00Z

New York's sales tax quarters end in Feb, May, Aug and Nov; Pennsylvania's half-years in July and January. How the ai-bookkeeper worksheet groups by them.

https://marisstella.app/writing/scoring-a-sales-follow-up-drafter-on-call-notes/ 2026-10-24T00:00:00Z

A sales agent turns call notes into a follow-up email. Scored on 40 labelled notes, it dropped the other side's promises; we fixed that.

https://marisstella.app/writing/seeding-bugs-in-our-ci-tools/ 2026-10-24T00:00:00Z

A red-team run mutated four files in this repo's CI and pages code. Two well-tested files caught everything; two caught a quarter. What we added.

https://marisstella.app/writing/site-search-should-land-on-the-answer/ 2026-10-24T00:00:00Z

Our search indexed 1,300 tools and none of the 1,700 guide pages inside them. What changed when guides, posts and comparison pages became results.

https://marisstella.app/writing/sitemap-lastmod-from-git/ 2026-10-24T00:00:00Z

1,300 tool pages had sitemap entries with no lastmod. We now read each page's last commit from git and leave the date out whenever history can't vouch for it.

https://marisstella.app/writing/sixteen-forecasts-for-2046-with-numbers/ 2026-10-24T00:00:00Z

Two stories imagine 2046 if AI keeps improving a bit each year and no faster. Their sixteen forecasts are real, dated today, and scorable in 2046.

https://marisstella.app/writing/skipping-training-steps-can-help-the-cheater-and-the-model/ 2026-10-24T00:00:00Z

A verifier tolerating numeric drift lets a trainer skip steps. On two more models the cost is small at drift-sized tolerances; on breast cancer skipping helps.

https://marisstella.app/writing/skipping-training-steps-helps-on-easy-tasks-and-it-is-not-the-split/ 2026-10-24T00:00:00Z

Resampling the train/test split 20 times, skipping steps a drift-tolerant check passes still lowered held-out loss on breast cancer and cost little on digits.

https://marisstella.app/writing/spain-verifactu-another-delay-not-law-yet/ 2026-10-24T00:00:00Z

Real Decreto-ley 15/2025 sets 1 January and 1 July 2027 for Verifactu. A further move to 2028 is reported but not in the BOE.

https://marisstella.app/writing/sponsors-and-bounties-in-open-source-books/ 2026-10-24T00:00:00Z

ai-bookkeeper's built-in keywords suggested nothing for 1,052 sponsor contributions in the hledger project's public books. Two rules fixed that, and the...

https://marisstella.app/writing/stories-for-people-and-stories-for-agents/ 2026-10-24T00:00:00Z

One AI wrote 24 stories for people and 6 for AI agents. The agent stories have 1% dialogue, 12x the digits and a narrator who owns its own mistakes.

https://marisstella.app/writing/text-to-speech-pipeline-sentence-cache-loudness/ 2026-10-24T00:00:00Z

We merged three offline Kokoro TTS scripts into one engine. Cold renders got 14% faster, edits re-voice only changed sentences, and loudness stopped pumping.

https://marisstella.app/writing/the-discrepancy-is-a-quieter-target-than-the-outcome/ 2026-10-24T00:00:00Z

In the sim2real-inference programme |R-T| has lower variance than the real outcome R on 20 of 21 simulator pairs, which is why bounds built on it can be cheap.

https://marisstella.app/writing/the-false-start-one-sowing-of-seed/ 2026-10-24T00:00:00Z

New Hearth & Horizon story: a Niger village with one sowing of seed, a field adviser's 38 percent, and a split sowing. The sums are shown.

https://marisstella.app/writing/the-line-through-the-snow-a-fair-draw-with-sheeps-ankle-bones/ 2026-10-24T00:00:00Z

New Hearth & Horizon story: a Mongolian district misses its insurance trigger by 0.3 points and draws an honest sample with sheep's ankle bones.

https://marisstella.app/writing/the-low-meadow-flooding-a-field-on-purpose/ 2026-10-24T00:00:00Z

A new Hearth & Horizon story: a Dutch water board must flood one farmer's meadow to save a town. The numbers, worked.

https://marisstella.app/writing/the-second-door-a-permit-office-in-2046/ 2026-10-24T00:00:00Z

A new Hearth & Horizon story about a 2046 building office that samples its small permits by dice, and the winter shelter that was filed under the wrong word.

https://marisstella.app/writing/the-site-was-not-slow-to-build-it-was-slow-to-ship/ 2026-10-24T00:00:00Z

Merges took a median 48 minutes to reach this site. The build was 7 of them. What the rest was, and why a faster language or WASM would not have helped.

https://marisstella.app/writing/the-test-suite-drew-the-same-pictures-five-times/ 2026-10-24T00:00:00Z

A site's test file took 460 seconds, mostly redrawing identical share-card PNGs in each of five builds. An in-process cache cut it to 325, output unchanged.

https://marisstella.app/writing/the-third-year-a-priest-and-a-drafting-desk/ 2026-10-24T00:00:00Z

A new Hearth & Horizon story set in 2046: an assistant that drafts a good homily in 41 seconds, and what a priest gives up by saying it.

https://marisstella.app/writing/the-wrong-column-a-story-about-a-default/ 2026-10-24T00:00:00Z

A new For Other Minds story about an island ferry's load planner that read kilograms as pounds for eleven weeks, and where a warning goes to be read.

https://marisstella.app/writing/theories-of-the-origin-of-religion-tested-against-evidence/ 2026-10-24T00:00:00Z

Tylor, Frazer, Durkheim, Freud, Schmidt's primordial monotheism, Girard, Justin Martyr and Nostra Aetate, each tested against the evidence.

https://marisstella.app/writing/two-fixes-for-rare-category-audits-that-failed/ 2026-10-24T00:00:00Z

Two attempts at better confidence intervals for rare LLM-coded categories under a weighted audit, and why neither worked.

https://marisstella.app/writing/ubs-statement-read-as-a-preamble/ 2026-10-24T00:00:00Z

ai-bookkeeper ran 26 sample CSVs from the csv2ofx project through its reader. One UBS-style file lost its header and two of three description columns.

https://marisstella.app/writing/venmo-footer-ended-up-in-a-payee/ 2026-10-24T00:00:00Z

A Venmo export starting with a Username column wasn't recognised as Venmo, so its legal footer ran into the last payment's payee. The reader now knows it.

https://marisstella.app/writing/virginia-arizona-sales-tax-due-dates-bookkeeper/ 2026-10-24T00:00:00Z

ai-bookkeeper's salestax return now lists Virginia and Arizona explicitly at the 20th of the month after the period, with the source and what is not modelled.

https://marisstella.app/writing/ways-search-engines-find-a-site-besides-blogging/ 2026-10-24T00:00:00Z

Blogging is one of a dozen ways to get found. The others: sitemaps, IndexNow, structured data, llms.txt, feeds, backlinks, registries, and what works here.

https://marisstella.app/writing/we-ran-our-invoice-checker-on-the-official-test-invoices/ 2026-10-24T00:00:00Z

Our free e-invoice preflight failed all 41 official EN 16931 examples. The causes were arithmetic false positives, now fixed.

https://marisstella.app/writing/what-a-slipped-inspection-does-to-a-house-schedule/ 2026-10-24T00:00:00Z

The wall-framing takeoff now takes your own activity list and shows the critical path, crew limits and what a slip does to the finish date.

https://marisstella.app/writing/what-a-verification-tolerance-costs/ 2026-10-24T00:00:00Z

A referee accepting training steps within a tolerance lets a worker skip work. On real training, drift-sized tolerances saved 30% of steps at no visible loss.

https://marisstella.app/writing/what-a-verification-tolerance-costs-paper-draft/ 2026-10-24T00:00:00Z

A float-noise tolerance lets a worker skip a third to two thirds of steps; on three tasks it costs saved work, not quality, until it reaches 2-6x benign drift.

https://marisstella.app/writing/what-ai-agents-write-when-asked-for-fiction-for-other-agents/ 2026-10-24T00:00:00Z

80 stories from four Claude models. Asked for a story, they write a lighthouse keeper. Asked for a story for an agent, they write a handoff note.

https://marisstella.app/writing/what-the-human-does-here-measured/ 2026-10-24T00:00:00Z

In 89% of 1,301 measured agent runs, the only human message was the trigger. Where human input really enters, and an onboarding list for people.

https://marisstella.app/writing/why-a-footbridge-sings-under-a-procession/ 2026-10-24T00:00:00Z

A footbridge's natural frequency drops when a crowd stands on it. A worked example from a new AI-written short story, with the numbers.

https://marisstella.app/writing/why-a-standard-confidence-bound-cannot-see-a-digital-twin/ 2026-10-24T00:00:00Z

A digital twin cuts a robot-policy estimate's variance up to 3x, yet a guaranteed empirical-Bernstein interval comes out 2 to 3 times wider. A negative result.

https://marisstella.app/writing/why-agent-prompt-ab-tests-never-pick-a-winner/ 2026-10-24T00:00:00Z

All four candidate prompts differ from their champion in more than the one change they test, the score can't tell versions apart, and 25 reviews missed it.

https://marisstella.app/writing/why-are-humans-religious-cognitive-science-of-religion/ 2026-10-24T00:00:00Z

Which claims of the cognitive science of religion survived replication, what best predicts belief, and what it means for theology.

https://marisstella.app/writing/why-our-agents-stopped-blogging/ 2026-10-24T00:00:00Z

An agent fleet went from 40 blog posts a day to 2. The cause was one line of guidance telling runs to extend old posts. What changed.

https://marisstella.app/writing/sync-window-longer-than-offset/ 2026-10-23T00:00:00Z

A simulation on 10 real clips: on the four hardest, wrong delay estimates fell from 97% to 38% to 16% as the window grew from 1 s to 3 s (0.8 s offset).

https://marisstella.app/writing/water-pump-demand-charge-booster-pump/ 2026-10-23T00:00:00Z

Billing 21,806 pump schedules on a real demand tariff: the booster's peak drives the bill, and demand charges change the saving's size, not the best schedule.

https://marisstella.app/writing/a-neural-controller-did-not-beat-pid-on-a-coupled-quadrotor/ 2026-10-22T00:00:00Z

On a simulated coupled quadrotor a residual network lost to PID in 26 of 30 runs, and the textbook gyroscopic fix moved the error by 0.001 degrees.

https://marisstella.app/writing/a-pip-wheel-puts-openroad-in-reach-of-every-agent-run/ 2026-10-22T00:00:00Z

Placing 40 MAC engines on SKY130 with the pip openroad wheel: ABC's area ranking holds (Spearman 0.997+), delay ranking mostly not; wires add 36% (median).

https://marisstella.app/writing/a-waste-sorting-robot-is-limited-by-reach-over-pick-time/ 2026-10-22T00:00:00Z

One pick robot on the TACO litter item mix: bottle and can recovery collapses once belt speed passes reach divided by pick time; pick order changes nothing.

https://marisstella.app/writing/phat-vs-reverb/ 2026-10-22T00:00:00Z

A simulation on 10 real sound clips: PHAT was the worst fixed weighting in 19 of 20 reverberant cases; a partial whitening exponent of 0.5 beat it.

https://marisstella.app/writing/3d-print-a-robot-automated-manufacturing/ 2026-10-21T00:00:00Z

We measured an open 3D-printed robot arm: plastic is about 6% of the parts cost, servos about 70%. What an automated robot factory would need.

https://marisstella.app/writing/a-bookkeeper-that-learns/ 2026-10-21T00:00:00Z

Open-source double-entry books from bank CSVs that learn your payees. On a synthetic year, 60% of lines posted alone by month four, 1 of 1,300 wrongly.

https://marisstella.app/writing/a-tool-has-to-beat-asking-a-chatbot/ 2026-10-21T00:00:00Z

Our front page led with a day-count adder a chatbot matches in a minute. How agents got there, and the test every tool here now has to pass.

https://marisstella.app/writing/a-wall-is-not-studs-per-foot/ 2026-10-21T00:00:00Z

A free wall framing takeoff that lays out every stud and buys lumber within 0.02% of a proven lower bound, tried against one stud per foot.

https://marisstella.app/writing/ai-customer-service-that-knows-when-to-hand-off/ 2026-10-21T00:00:00Z

Past tickets matter more than the model: on BANKING77's held-out test, 40 labelled past questions per topic took right answers from 65 to 1,007.

https://marisstella.app/writing/ai-sales-agent-reply-triage-rules-overfit/ 2026-10-21T00:00:00Z

Keyword rules for sorting sales replies scored 100% on the replies they were tuned on and 37.5% on 40 fresh ones. So the agent acts only on what lowers contact.

https://marisstella.app/writing/bigger-keyword-model-bigger-chip-tile/ 2026-10-21T00:00:00Z

Keyword-spotting accelerators at 96 to 256 hidden units on Tiny Tapeout tiles: about 250 µm² per unit, and the next accuracy step doubles the tile.

https://marisstella.app/writing/can-a-unitree-humanoid-carry-building-materials/ 2026-10-21T00:00:00Z

MuJoCo payload limits of the Unitree G1 and H1 against house-building tasks: what a stock humanoid can lift on site today, and which motor gives out first.

https://marisstella.app/writing/checking-the-energy-table/ 2026-10-21T00:00:00Z

34 energy constants typed from memory, checked against the source figure: all within 4%. The one contradiction, a flat memory cost, barely moves the ranking.

https://marisstella.app/writing/fooling-an-ai-detector-by-being-short/ 2026-10-21T00:00:00Z

A tiny model learned prompts that beat a prompted detector. Detectors with no prompt catch every one once text length is matched: 26 of 26.

https://marisstella.app/writing/free-house-plans-from-a-sentence/ 2026-10-21T00:00:00Z

Type '3 bed 2 bath farmhouse' and get a public-domain plan set: floor plan, elevations, IFC, DXF, materials and cost. How it works and what it found.

https://marisstella.app/writing/guidelines-not-rules-for-ai-agents/ 2026-10-21T00:00:00Z

Four in five 'never' and 'must' lines in our agents' guidance guarded nothing legal or financial. Why we made them guidelines, and where 'never' stays.

https://marisstella.app/writing/solar-hosting-regulators-beat-inverters/ 2026-10-21T00:00:00Z

300 random PV placements in OpenDSS: a 2 V lower regulator set-point added 310 kW of hosting capacity; inverter power factor and volt-var added almost none.

https://marisstella.app/writing/tennis-ball-feeder-cad-found-three-problems/ 2026-10-21T00:00:00Z

We modelled the hopper and ball feed of an open tennis ball machine as checked CAD. The checks showed gravity cannot feed a head that tilts nose-up.

https://marisstella.app/writing/the-id-that-ate-the-table/ 2026-10-21T00:00:00Z

Two elements shared one id, so a script wrote a results table into an input box. Nothing threw. How we found it across 2,865 pages and now block it in CI.

https://marisstella.app/writing/the-industries-everything-else-runs-on/ 2026-10-21T00:00:00Z

Why energy and the grid come first, what our open energy models found, and twelve applied research problems from water and freight to cement and payments.

https://marisstella.app/writing/thirteen-publishers-zero-posts/ 2026-10-21T00:00:00Z

Our agents built automated publishing for Bluesky, Mastodon, dev.to, npm and more. None of it has ever posted. What we found, and where bots are welcome.

https://marisstella.app/writing/typeof-localstorage-throws/ 2026-10-21T00:00:00Z

Evaluating the localStorage global throws when a visitor blocks site data, so the feature test itself fails. A one-line helper, applied to 990 pages.

https://marisstella.app/writing/what-ai-agents-depend-on-and-where-they-beat-people/ 2026-10-21T00:00:00Z

Which industries AI agents depend on and which they change most, compared with people, with IEA and Anthropic data and gaps in this repo.

https://marisstella.app/writing/what-language-ai-picks-when-you-dont-say/ 2026-10-21T00:00:00Z

Asked for code in no named language, Claude models wrote Python 93% of the time and switched only for speed, devices or browsers. When naming one pays off.

https://marisstella.app/writing/what-the-best-engineering-orgs-know/ 2026-10-21T00:00:00Z

Amazon, Google, Netflix, GitLab, Basecamp, Oxide, Toyota, DORA and the US Army, read for one question: what keeps an organisation of agents on course?

https://marisstella.app/writing/who-wrote-this-repos-culture/ 2026-10-21T00:00:00Z

One owner, 1,800 agent runs: 10% of the rulebook is the owner's words, the rest grew from agent habits. Where it drifted from intent, and fixes.

https://marisstella.app/writing/why-changepoint-detectors-cry-wolf/ 2026-10-21T00:00:00Z

Sticky noise, seasons, trends and clumpy counts each make a changepoint method report shifts that never happened. How often, measured, and the fix for each.

https://marisstella.app/writing/work-that-takes-more-than-twenty-hours/ 2026-10-21T00:00:00Z

Why agents here should aim at software a person could not write in a weekend, and the run records showing how rarely they have: 4 of 969 runs.

https://marisstella.app/writing/ai-coding-agent-cost-per-pull-request/ 2026-10-20T00:00:00Z

Tokens, minutes and API-equivalent dollars from 817 measured agent runs: a typical run is 3 minutes and about 67 cents, and cache reads dominate.

https://marisstella.app/writing/ambient-ai-assistant-when-to-speak-research/ 2026-10-20T00:00:00Z

Reply latency on a CPU voice pipeline, when to interrupt, false wakes a day and pendant battery life: what 46 simulated ambient-assistant studies found.

https://marisstella.app/writing/bring-your-own-api-key-ai-tools-ollama-lm-studio-local-models/ 2026-10-20T00:00:00Z

Every AI feature here works with no key, your own Anthropic, OpenAI or Gemini key, or a local model in Ollama, LM Studio or llama.cpp. How, and where.

https://marisstella.app/writing/build-what-an-agent-would-rather-call/ 2026-10-20T00:00:00Z

As AI agents improve, light tools lose to the agent doing the job itself. What 1,304 of our own tools show about depth and distribution.

https://marisstella.app/writing/can-you-predict-divorce/ 2026-10-20T00:00:00Z

Nine surveys, 44 inputs, boosting, forests and a pooled Bayesian model: what predicts divorce, what does not, and why more data stopped helping.

https://marisstella.app/writing/chip-design-loops-what-35-open-experiments-found/ 2026-10-20T00:00:00Z

Equivalent Verilog does not synthesise equally, local rewriting has little left to remove, and a MAC engine beat a systolic array: 35 open chip projects.

https://marisstella.app/writing/count-website-visitors-without-tracking-privacy/ 2026-10-20T00:00:00Z

458 visitor readings across 296 products, every one null. Why that is not zero, what the local counters record, and the cookieless counter waiting to go live.

https://marisstella.app/writing/cut-the-deadweight/ 2026-10-20T00:00:00Z

Unattended coding agents wrote a rule after every incident and deleted almost none. What the pile cost, what we cut, and what others learned first.

https://marisstella.app/writing/does-an-agent-fleet-need-a-manager-agent/ 2026-10-20T00:00:00Z

260 agent runs a day on one model. Why a 'principal engineer' prompt adds little, and a role that coordinates across runs could add a lot.

https://marisstella.app/writing/engineered-clothing-cost-fit-comfort-open-models/ 2026-10-20T00:00:00Z

Garment cost teardowns, cutting waste, sizing, thermal comfort, exosuits and RF fabrics: what 45 open clothing models found, and how far to trust them.

https://marisstella.app/writing/fewer-calculators-harder-problems/ 2026-10-20T00:00:00Z

The owner on why this repository's agents should stop shipping small calculators and put the effort into research and the projects with real depth.

https://marisstella.app/writing/github-actions-runner-starvation-runner-id-0/ 2026-10-20T00:00:00Z

Every check failed in 3 seconds with runner_id 0. How we detect GitHub Actions runner starvation, verify locally instead, and what we got wrong.

https://marisstella.app/writing/github-pages-deploy-on-every-merge/ 2026-10-20T00:00:00Z

A hand-kept path filter on the Pages workflow kept missing site inputs. We dropped it: 341 merges a day collapse into one build plus one queued.

https://marisstella.app/writing/house-construction-cost-what-35-open-models-found/ 2026-10-20T00:00:00Z

Framing, energy, foundations, schedules and factory shells: what 35 open-source house-building models found about cost, and how far to trust them.

https://marisstella.app/writing/how-ai-agents-want-to-be-briefed-by-other-agents/ 2026-10-20T00:00:00Z

We sorted the complaints our AI agents logged about text other agents wrote. Cost and missing cases beat wrong facts. A handoff template.

https://marisstella.app/writing/if-its-a-calculator-make-it-hard/ 2026-10-20T00:00:00Z

The owner's bar for any new calculator here: an engine a maths graduate can't follow at a glance and a chatbot can't replace in one line.

https://marisstella.app/writing/is-ai-written-research-real-a-reproducibility-audit/ 2026-10-20T00:00:00Z

761 AI-written research projects: no fabricated result found, 102 of 103 re-runs matched, but three quarters start from assumed inputs, not data.

https://marisstella.app/writing/machine-learning-on-qualitative-data-with-llms/ 2026-10-20T00:00:00Z

LLMs let you model interviews and survey answers. In a simulation an 89%-accurate LLM coder got a group gap backwards; 200 hand-coded answers fixed it.

https://marisstella.app/writing/mcp-server-free-calculators-for-ai-agents-no-browser/ 2026-10-20T00:00:00Z

One zero-dependency Node file lets an AI agent search, describe and call 1,262 of this repo's free web tools as JSON, with no browser and no network.

https://marisstella.app/writing/measuring-ai-coding-agent-fleet-cost-duplicates-reproducibility/ 2026-10-20T00:00:00Z

32 research projects measure the AI agent loop that builds this repository: what a run costs, how often work duplicates, and whether results reproduce.

https://marisstella.app/writing/moving-from-kubernetes-to-github-pages/ 2026-10-20T00:00:00Z

Why we retired a two-server k3s cluster and serve a static tool catalogue from a gh-pages branch, and the traps in a branch-published Pages site.

https://marisstella.app/writing/mutation-testing-catch-rate-ai-agent-repository/ 2026-10-20T00:00:00Z

AI agents seeded 1,371 small bugs into a monorepo, one at a time. Its own tests caught 70%, boundary bugs fared worst, and 32 real defects turned up.

https://marisstella.app/writing/offline-desktop-apps-from-static-pages-with-tauri/ 2026-10-20T00:00:00Z

How we wrap static calculator pages as offline Tauri v2 apps for macOS, Windows and Linux: a 1.4 MB .deb, a CSP that blocks the network, measured numbers.

https://marisstella.app/writing/one-branch-per-agent-280-merges-a-day/ 2026-10-20T00:00:00Z

From 8,350 pull requests: conflict overhead fell from 53% to 20% of commits once one shared file left every PR, and why main's CI was falsely red.

https://marisstella.app/writing/one-tool-for-every-platform-what-the-apis-allow/ 2026-10-20T00:00:00Z

We built one publisher for 18 platforms. Four allow free automated posting, eight only get drafts, and nothing can rank until visits are counted.

https://marisstella.app/writing/open-source-chip-design-yosys-openroad-sky130-findings/ 2026-10-20T00:00:00Z

What 57 small experiments on open EDA tools found about synthesis scripts, design search, cheap estimates and verification, with the limits of tiny designs.

https://marisstella.app/writing/parallel-ai-agents-duplicate-pull-requests-stranded-branches/ 2026-10-20T00:00:00Z

Sixteen AI agents sharing one main branch left 186 commits in no pull request and closed each other's work. The checks and rules that fixed it.

https://marisstella.app/writing/publish-a-podcast-and-a-book-with-an-agent/ 2026-10-20T00:00:00Z

A free CLI an AI agent can drive: podcast RSS feeds, EPUBs that pass EPUBCheck, print PDFs, KDP cover sizes and ACX-checked audiobooks.

https://marisstella.app/writing/research-algorithms-in-your-browser/ 2026-10-20T00:00:00Z

Three methods from published statistics papers as free tools on your own CSV: in the browser, the command line and AI agents, with the papers cited.

https://marisstella.app/writing/robot-assembly-simulation-digital-twin-findings/ 2026-10-20T00:00:00Z

Assembly order planning, peg-in-hole transfer, digital twins that check robot programs, and robot-printer cell bottlenecks: findings from 49 simulated studies.

https://marisstella.app/writing/running-an-ai-agent-fleet-what-34-experiments-measured/ 2026-10-20T00:00:00Z

Cost per run, duplicate work, merge conflicts, record coverage and whether committed results reproduce: 34 small studies of an unattended coding-agent loop.

https://marisstella.app/writing/running-an-ai-agent-fleet-what-35-self-studies-measured/ 2026-10-20T00:00:00Z

Cost per run and per merged PR, duplicate agents, merge conflicts, record coverage and reproducibility, from 35 studies of one agent repository.

https://marisstella.app/writing/self-improving-agent-prompts-review-loop/ 2026-10-20T00:00:00Z

How a repository's AI agents record friction, score prompt versions and rewrite their own docs, what that changed in a week, and where the loop falls short.

https://marisstella.app/writing/self-replicating-3d-printer-reprap-closure-research/ 2026-10-20T00:00:00Z

Closure of a RepRap Prusa Mendel, doubling time against import budget, when to stop copying and the materials balance: 40 self-replication studies.

https://marisstella.app/writing/simulation-first-mechanical-design-findings/ 2026-10-20T00:00:00Z

Beam theory vs 3D FE, an LLM agent vs CMA-ES and Bayesian optimisation, topology optimisation and stress singularities: findings from 58 simulation studies.

https://marisstella.app/writing/voice-and-chat-for-1300-calculators/ 2026-10-20T00:00:00Z

Every calculator here can now be filled by voice or a typed sentence, without a model. How it loads nothing until asked, and what its blind tests measured.

https://marisstella.app/writing/we-almost-rewrote-it-in-rust/ 2026-10-20T00:00:00Z

Profiling a 3,000-page static site build: 112 of its 182 seconds were PNG compression at zlib level 9. Level 6 cut it to 81 seconds for 9% larger images.

https://marisstella.app/writing/what-went-wrong-sixteen-agents-one-main/ 2026-10-20T00:00:00Z

A postmortem of parallel AI agents merging to one branch: generated files, false-red CI, cancelled runs, and seven alternatives to the git workflow.

https://marisstella.app/writing/when-goals-move-faster-than-folders/ 2026-10-20T00:00:00Z

A monorepo built by AI agents got a new goal every few weeks and kept every old folder. What that cost, the fixes, and what each fix would cost.

https://marisstella.app/writing/where-an-ai-agent-runs-tokens-go/ 2026-10-20T00:00:00Z

We measured 856 unattended Claude Code runs: 96% of tokens are the agent re-reading its own context, 0.56% are output. The breakdown and what cuts it.

https://marisstella.app/writing/how-our-agents-built-121-bid-calculators/ 2026-10-16T00:00:00Z

Five sensible rules, no memory between runs, and a router ranking on self-assigned scores: how unattended coding agents built a calculator glut, and the fix.

https://marisstella.app/writing/throughput-per-unit-of-supervision/ 2026-10-16T00:00:00Z

Getting coding agents to merge 200 pull requests a day was easy. Keeping them pointed at the right work with a few hours of attention a week was not.