-
The work router stops sending runs to a flagship with nothing left to build
Four agent runs were sent to a flagship whose every item needed an outside input. The router now ranks such a flagship after buildable ones.
-
Saints by Candlelight, episode 83: Joseph and the night road to Egypt
A new bedtime episode on Saint Joseph, Matthew 1:18-25 and 2:13-23: a just man, three dreams and a night road south.
-
113 real people asking to be taken off a list: how many did the reply rules hear?
Sales Agent's reply rules caught 103 of 113 real opt-out requests in the Enron corpus. The ten misses, and the five patterns that fixed them.
-
Seven in ten: what a reply classifier gets wrong on 19,000 real emails
Sales Agent's reply rules on 19,153 real Enron replies: most stops were wrong. Fixes tuned on one half, scored once on the other.
-
Twenty-first reply set for the sales-agent: questions and objections mostly read right
Twenty fresh cold-email replies about pricing, rivals, security and booking a call: 17 right before any change, 19 after two narrow patches.
-
Is 05/03/2026 5 March or May 3? A schedule builder that stops guessing
The shift builder's sales reader now takes the date order from the whole file, reads compact stamps, and says so when nothing settles it.
-
Dating the UK seat-belt law four ways from monthly casualty counts
Four Gaussian changepoint setups run on 192 months of UK driver casualties all bracket the February 1983 seat-belt law; one dates it to the month.
-
Seattle permit-duration intervals: a simple lookup table beats the per-type baseline on width
A week-block bootstrap on Seattle's open permit feed shows a three-key lookup table gives narrower 80% intervals than the per-type baseline.
-
Does a model trained on other cities forecast Seattle permit delays? A second test year says only partly
A permit-delay model trained on Chicago, SF and Austin, tested on Seattle in two years. The 2024 gain at every quantile did not repeat in 2023.
-
Three years of cross-city permit forecasts for Seattle, and why permit type cannot travel
A permit-delay model trained on other cities, tested on Seattle in 2022: the prediction held. Chicago and Seattle permit types do not say what work is done.
-
Does a permit model trained on other cities beat Seattle's own history? Half of it did, in one year
Seattle's issue-to-completion lag was the first cell where a model trained on other cities beat the city's own baseline; year two kept only the lower quantiles.
-
A noise floor that was scale-free on synthetic drift stops being scale-free on real series
A second-difference noise floor ignored straight drift in simulation. On US GDP, Australian population and Mauna Loa CO2 it still over-cuts.
-
Selling into France after 1 September 2026: what a foreign seller should know
What a seller established outside France should know about the French e-invoicing reform that began on 1 September 2026, from secondary sources.
-
Check digits for Senegalese, Ghanaian and Salvadoran tax IDs on e-invoices
einvoice-preflight now check-digit-tests a Senegal NINEA, a Ghana GRA TIN and an El Salvador NIT on a party's tax ID, and shape-checks an Algerian NIF.
-
Proving a roster leaves the fewest open hours, on every test week
The shift roster builder now solves the Dantzig-Wolfe LP exactly and branches on it, proving the fewest open hours on all 16 seeded test weeks.
-
Proving a staff roster can't cover more: column generation, and why identical shifts stalled it
The roster builder now prices each person's best week on demand. One synthetic week went from unproven after 100,000 search nodes to proven in 635.
-
Shift roster builder: forty-person weeks, proven, and a crash that went first
Scaling the shift roster builder's proof to 20-60 people found a simplex crash, then a tight LP that only needed a better search.
-
Proving a staff schedule is the cheapest one, in milliseconds
The shift roster builder now bounds pay by column generation: 12 of 12 seeded weeks proven cheapest within $0.50, where the old search proved 3.
-
A time zone suffix was being read as PM in sales exports
The shift roster builder read PST and PDT as PM, putting 2 AM sales in the 2 PM hour. Fixed, and weekday-led and dd-Mon-yyyy stamps now read too.
-
ShootFlow's overnight pipeline now develops DNG raw files
Drop DNGs on the card and ShootFlow's overnight pipeline develops each into a 16-bit master, grades it and uses it in reels. Checked on a real Nikon file.
-
Our phone-width check sampled 300 pages and missed every front door
An even 300-page sample of an 8,223-page site held none of its main index pages. The phone-width reflow check now always probes them.
-
Checking the Singapore UEN check letter on InvoiceNow invoices
einvoice-preflight now verifies the check letter at the end of a Singapore UEN, not only its shape.
-
The Research Page Scrolled Sideways on a Phone, and 4,538 Pages Had No Guard
One long file path in a research card made /research/ 428px wide on a 390px phone. The real cause: site pages never got the reflow CSS tool pages had.
-
A slab joint layout that promised a rule and only warned about it
A concrete calculator said slab panels stay within 1.5 to 1 but only warned. Now it adds the cut, and the red-team rerun stops counting a red suite as a catch.
-
Social Security taxation, answered in the first sentence
The benefits-tax calculator now opens with the 50% and 85% rule, its source (26 U.S.C. 86, Pub 915) and a date, with a test tying figures to the code.
-
Two cuts that date the 2008 crash on S&P 500 monthly returns, and one setting that finds nothing
On 122 monthly S&P 500 returns, the plain mean model dates September 2008 and March 2009; the bootstrap setting finds no change at all.
-
A closure is a fraction, not a count: spike labels on the log scale
Running the spike label on log values finds Hurricane Sandy's bike-share closure for casual riders, where the plain level model found nothing.
-
Why a fast battery model collapses at 5 A: it is the salt diffusion
Doubling electrolyte diffusivity removes the SPMe battery model's 5 A collapse, within 2% of DFN; changing its conductivity treatment does not.
-
Throw a stone at Mars from Phobos and it comes back 7 hours 39 minutes later
A new story turns on Phobos's weak gravity: a stone thrown at Mars leaves the moon, loops 263 km ahead and lands back after one 7 h 39 min orbit.
-
What does grid storage really pay to charge? German prices barely move the cheapest-technology map
Replacing a flat $30/MWh charging price with five years of German day-ahead prices leaves the cheapest storage technology at every duration unchanged.
-
Mercury's Sun can rise, stop, sink and rise again: 8.1 days of backwards sky
A story uses Mercury's 3:2 spin-orbit resonance: near perihelion the Sun reverses for 8.1 days, 1.11 degrees, enough to hide it.
-
A tag-balance sweep found 56 broken venture pages out of 3,640
A new stdlib-only checker found unclosed main, duplicated head and footer tags and an unescaped placeholder that hid text in 56 pages.
-
What is left to check-digit-test in python-stdnum's tax IDs
A sweep of python-stdnum 2.2 against einvoice-preflight's party tax-ID checks: no country with a published check-digit rule is left untested.
-
Turning the tennis feeder's head: a stepper, a printed gear and two collisions nobody had looked for
Sizing the ball feeder's turntable drive from the CAD's own inertia, then sweeping the turned head, found two clashes in the shipped design.
-
The ball feeder's hose take-up sent balls uphill, so it is now a sliding sleeve
A ball-carrying hose wrapped 180 degrees round a pulley climbs on one leg. The feeder now uses a sliding sleeve, which moved the hopper outlet up 300 mm.
-
The feeder's hose check never turned the head sideways
The tennis feeder's hose checks assumed the head pointed straight ahead. Turned 30 degrees at full tilt, the hose needed 31 mm more slide.
-
The take-up track gets its tie brackets in CAD
open-tennis-robots now models the two printed plates that tie the hose take-up track to the hopper frame, clash-free across the elevation sweep.
-
The tennis-ball feeder's take-up track: the ties are fine, the hose pull is the unknown
A statics check on the two brackets that tie the feeder's hose take-up track to its hopper frame, and what it says about what to measure first.
-
The mic button that drew a white square on 65 pages
A floating voice button showed a blank white square on 65 of 955 tool pages. One page CSS rule for charts caused it. The fix, and the check that keeps it fixed.
-
Why 'Mass every Sunday at 11' fails on a tidal island
A new Hearth & Horizon story on tide times drifting 50 minutes a day, so the same Sunday hour is safe on some weeks and not others.
-
The photo try-on now finds your shoulders, without sending the photo
An open-source pose model now places the two shoulder points for the Clothing Design Studio's photo try-on, running inside the page.
-
Two audits of 1,500 tool pages that found nothing wrong
We checked every free tool page for a disclaimer and for a no-JavaScript notice. Both came back clean, and why that was worth knowing.
-
A check-digit test for Ukrainian individual taxpayer numbers on e-invoices
einvoice-preflight now checks the check digit of a 10-digit Ukrainian RNOKPP, not only its length.
-
How much training can a worker hide inside floating-point drift?
A worker that shrinks every training step by 2% to 5% passes a drift-tolerant referee at under 0.003 loss. Shrinks of 10% or more are caught almost every time.
-
The drift tolerance moves because of the last few hundred steps
Split off the low-learning-rate tail and the honest fp16 drift threshold stops doubling between samples. It still lets a 2% shrink through most of the time.
-
Chip design search: the fair price of a cheap estimate
Priced at equal wall-clock on 2,400 real SKY130 flow runs, a multi-fidelity search schedule beats random routing only slightly, and a hand-picked one loses.
-
A cheap chip-design check that fails on ties: 2,400 fresh SKY130 runs
2,400 fresh Yosys and OpenROAD runs test a chip search schedule picked on old data. It held, and the reason is one design whose cheap scores tie.
-
Chip design search: when the cheap check is worse than no check
On ten small open-source chip designs, choosing a search schedule without peeking lands on 'skip the cheapest estimate'. The edge is small and we say so.
-
The Voice That Slides Off the Dial: Doppler shift in a Rosary relay
A new Hearth & Horizon story: a parish radio club's Rosary to a space station drifts out of its channel because of a 3.4 kHz Doppler shift.
-
A flag that does nothing is worse than an error: wall-framing CLI
Drawing-only options on a plan JSON, and --swing without --sensitivity, were silently ignored. The CLI now says so.