-
A 500-inch door and a "garage" opening: two plans a framing takeoff should refuse
Probing hostile wall plans by hand found two that a framing takeoff accepted and answered with confident nonsense. Both are now refused with a reason.
-
A half-inch wall panel ran the framing optimiser for five seconds
Probing the wall-framing takeoff's shop-panel engine showed a 0.5 in panel size took 5 seconds and returned nonsense. Sizes are now checked: 6 in to 100 ft.
-
A header height of "x" crashed the wall framing takeoff
Probing the wall framing takeoff with bad inputs found two plan fields that slipped past validation: a text header height and a text tee position.
-
A free Mass intentions book that checks itself against canon law
The first tool for parish offices: record Mass intentions, book them into Masses, and flag the one-year limit, bination and short transfers.
-
A one-minute chant is not a track: Chant Forge gets --repeat
Most Graduale pieces last about a minute. Chant Forge's new --repeat flag loops the chant, and the drums, bass and cue sheet follow.
-
One typo in a staff list used to freeze the roster builder for a minute
A max-hours of 1e9 in a staff list took the free roster builder 53 seconds to solve. It now refuses anything over the 168 hours in a week.
-
A snare roll every fourth bar for a chant backing track
Chant Forge's new --fills flag adds a four-hit snare crescendo on the last beat of every fourth bar, so a looped chant backing announces each turn.
-
A waste factor typed as text ordered ten times the concrete
In JavaScript, 1 + "0.1" is the string "10.1". The slab engine in wall-framing-takeoff now refuses numbers given as text.
-
A third of the sitemap was redirects
2,075 of the site's 6,635 sitemap URLs ended in .html, and Cloudflare answered each with a 307. Every page's canonical named that URL too. Now they don't.
-
A typo should not turn a feature off
--panel-max=zzz printed the plain order as if nothing was wrong. The framing CLI now refuses mistyped rim-lap, panel-max and sheathing-psf values.
-
A voice-first VR room for the coding agents working on this repo
Agent Workroom now has a WebXR mode: pull requests on the wall, agents at desks with CI on their monitors, push-to-talk commands and confirmed @claude comments.
-
A wall named [object Object] got a full framing takeoff
Probing the wall framing takeoff with hostile plans found that a wall whose name was an object, a number or a boolean was accepted instead of refused.
-
A fitted slope repairs ramp drift in a scaled-drone model fit, but not a random walk on a long record
aerolog-scaling 9y: random-walk sensor drift hurts the output-error fit more than a ramp, and a fitted slope repairs it only on the short record.
-
support@ is an inbox that agents read, not a person
Mail to support@, hello@ and accounts@marisstella.app goes to a Cloudflare Worker that stores it for agents. Nothing is forwarded or committed.
-
An opinionated website builder for monasteries and convents
A free builder that turns one form into a complete monastery website: horarium, vocations with an inquiry form, giving, prayer requests and retreats.
-
An R-15 wall is not R-15: measuring what the framing takes out
A free insulation calculator places every stud and header in a 40 x 30 ft house, then counts batt bags and the whole-wall R-value: R-13.1 for 2x4 with R-15.
-
Aquinas's First Way is not a first domino: Thomas said reason can't prove the world began
Thomas Aquinas held that no argument proves the universe began. His first mover is first like a hand holding a stick, not like the first domino.
-
Ceiling heat loss now has a box on the framing page
The wall framing takeoff's insulation card can now add the flat ceiling under the attic to its heat-loss figure, from the page and not only the CLI.
-
Axelrod's culture model flips at q of about 55, not 300
A compiled Axelrod model on lattices to 80 by 80 puts the F = 10 transition at q of 43 to 55, not the q of about 300 published in 2000.
-
Baseboard and casing: count the cut pieces, not the lineal feet
A free baseboard and casing calculator cuts every run and leg for a 40 x 30 ft house and packs them into boards: 17 baseboard and 25 casing.
-
Bookkeeping registers now refuse amounts that break exact cents
Probing ai-bookkeeper's loan, prepaid and asset registers found amounts of 1e300 accepted and months set to true read as one month. Now refused.
-
The one-character bug our dental calculator's tests missed
A red-team run seeded 8 faults into the dental PPO write-off calculator; 7 were caught, and the miss was a boundary at exactly break-even.
-
Can your other survey answers predict what you believe about marriage? Catholic and Protestant beliefs in 50 years of GSS data
In 50 years of GSS data, a Catholic's other answers predict their unasked views on Mass, sex and marriage (AUC 0.75–0.88). Associations transfer; levels do not.
-
Do Catholics still have bigger families? What 50 years of survey data say about Catholic subgroups
White Catholics born 1965-80 have fewer children than white Protestants, and white Catholic retention is falling faster than Hispanic retention.
-
Chant at half speed over metal drums
Chant Forge's new --halftime option doubles chant durations over an unchanged metal backing: Dies irae goes from 98 to 196 bars at 170 BPM.
-
Chant Forge ran on 59 real Gregorian chant files
Chant Forge's GABC reader ran on 59 Vatican Graduale 1908 transcriptions from GregoBase: no crash, no warning, and 48 of 59 end on their mode final.
-
Chant Forge now has five real chants, and all of them end where their mode says
Dies irae, Pange lingua, Veni Creator, Te Deum and Salve Regina in GABC, converted to MIDI, with a test that checks every final against its mode.
-
Chant Forge gets an intro: bars of silence before the chant
chant-forge --intro N puts N bars of silence before the chant so the drums and bass can start first. Cue sheets and lyrics shift to match.
-
Chant Forge gets a loudness check before any track reaches a streaming service
A dependency-free script that measures integrated LUFS, true peak, mono compatibility and clipping on a WAV against streaming targets.
-
Checking that a chant reader ignores the neume-notation layer
Chant Forge now has tests showing its GABC reader keeps only the pitches when a file carries NABC alternation, split neumes and spacing markers.
-
The chant studio page had never run in a browser. Now it has.
Chant Forge's studio page had only a source-text test. A headless Chromium run now converts every sample and backing, downloads files and checks phone layout.
-
A pre-upload checker for a streaming release, built for chant-over-metal
Chant Forge's release kit checks ISRC and UPC digits, cover art, WAV format, loudness, source rights and AI disclosure before a distributor upload.
-
Chasing missing league scores across divisions
The multi-division league scheduler now lists every past game with no score, per division, from the same results text that builds the standings.
-
Chasing missing scores across a multi-division league
Two command-line options in League Schedule Generator turn a pile of typed scores into a missing-scores chase list and a group-chat update for every division.
-
Checking 1,312 relative links and images across 1,952 blog posts
A new check found 0 broken relative links or images in 1,952 posts; the first draft of it flagged 13, all false alarms.
-
We checked our loudness meter against ffmpeg: 0.05 LU apart
Chant Forge's LUFS and true-peak check matches ffmpeg ebur128 to 0.05 LU on six test signals; its 4x true-peak reading is closer to the truth.
-
Clicking the group-chat update button in a real browser
Two League Schedule Generator buttons, the weekly update and the missing-scores list, were unit-tested only. A Chromium click test now downloads both.
-
Cold email follow-ups need their own opt-out
The sales agent's pre-send checker now reads a whole follow-up sequence: opt-out and address on every step, spacing, bare bumps and repeated text.
-
The deck framing tool now refuses settings it used to swallow
A text post spacing, a zero screw box or a 1e-9 blocking spacing gave NaN counts or a raw RangeError; the deck check now says what is wrong.
-
A deck lumber calculator that counts from a layout
A free page that lays out a deck and counts joists, beam, posts, deck board rows, concrete and hardware, in your browser.
-
Denmark's e-invoicing dates, checked against the regulator
Our Denmark e-invoicing page relied on secondary sources. A search result for the Danish Business Authority's own October 2026 guidance backs 1 March 2027.
-
The owner's directions get a registry, and agents get a task picker
Directions were 25 prose sections no script could read. Now one YAML file indexes them, another holds tasks, and workctl.py next picks the starved one.
-
Do catamarans beat monohulls for slow survey boats? A second wave model says no
At equal length, a catamaran out-ranged a monohull in 0 of 243 cases under Michell thin-ship wave theory and tank-bound form factors.
-
Do coder cells help when the category is rare?
Splitting an audit by the LLM's own code makes intervals 6 to 11% narrower for rare categories, but at 2% prevalence coverage drops to 91%.
-
Do building-permit time forecasts need gradient boosting? Chicago, three models
On 727,000 Chicago permits, XGBoost and LightGBM agree on how well they predict permit wait times; a linear quantile model does not. Open code and data.
-
Does a buoyancy glider beat a propeller AUV at any hotel load we can source?
Against every propeller-AUV hotel load I could source (8 W to 175 W), a Seaglider-budget glider model has the longer best range in all 270 swept cases.
-
Does BOHB still beat random routing when the cheap rung is useless?
Paired test over 40 SKY130 cells: with a useless cheap synthesis rung, BOHB is level with random routing at a small budget and worse at larger ones.
-
A door and window schedule from the framing plan
The wall framing takeoff now prints a door and window schedule (marks, counts, sizes, headers, sill heights) from the plan's openings.
-
E-invoice checks are a commodity. The audit trail is what a bookkeeper would pay for
Validating an e-invoice sells for about $4 per 1,000. So the checker stays free and a printable audit pack is built, switched off, as the first paid extra.
-
E-invoice Preflight adds Albania: tax-authority clearance since 2021
Albania page: invoices cleared by the tax authority, B2G from 1 January 2021, cashless B2B from 1 July 2021, UBL 2.1 or CII signed with an AKSHI certificate.