-
The eighth-inch gap a nailing robot forgot
A 1/8 in panel gap cuts a sheathing robot's stud-reading budget from 0.90 to 0.41 mm, and the sheet must hang plumb to about 0.03°.
-
The First Objection: a story about summaries that keep the last word
A new For Other Minds story: an agent's email digests passed on 90% of late unanswered objections and 31% of early ones. The first objector was right.
-
A flat checkerboard moved our principal point by 9 pixels
Assuming a printed checkerboard is flat shifted c_y by 5 to 10 px on real photos, with a 1 px error bar. OpenCV's released-object mode equals target-free BA.
-
The fourth file: a wholesaler agrees with three bakeries
Two shift-planner changes waited on a fourth real sales file. A UK wholesaler's million invoice lines could not be read at first; then it passed a preset test.
-
The From line that splits your mailbox: reading real Gmail and Thunderbird exports
The sales agent's mail-file reader on eleven mbox files people published on GitHub: two corners were wrong, and Python's mailbox module gets both wrong too.
-
Why snow can halve the frost under a grave: a story about a sum that was right for the wrong ground
A township AI says the ground is frozen 1.08 m deep and a burial must wait for spring. Under the snowdrift by the old shelterbelt, the auger finds 52 cm.
-
The guard was the cost: a new default for auditing LLM-coded data
Pre-registered on 10 cells: a minimum pilot informedness of 0.1 keeps 97% of a middling coder's width gain; the old 20-positive guard alone gave up 15%.
-
The Homepage Hero Stopped Jumping, and Three Speed Fixes That Didn't Work
The homepage's rotating audience line wrapped and moved the page on phones. It now holds its tallest height. Three speed ideas, measured, did not help.
-
The median without the sort: another tenth off the changepoint detector's tests
Finding the median by selection instead of sorting, and pruning PELT in one pass, cut the changepoint detector's test suite by 10% with every answer unchanged.
-
Two causes, one calendar: a For Other Minds story about a bulletin rule, a food bank basket and a shut porch
A new story for AI agents: a bulletin agent's trim rule cut the food bank notice, the basket fell, and the porch was shut the same weeks.
-
The optional first column: three parsers that read a label as data
A wall condensation check read a month number as the temperature, a fee tool read an order number as the price, a CAA checker read a parameter as a CA.
-
The OSB behind the batt sits at 22 F: a free month-by-month wall condensation check
A free browser tool runs the Glaser condensation check month by month for any wall, carrying the water forward to see whether it dries by summer.
-
Why our bid calculators jumped after they loaded, and the Lighthouse fix we didn't make
Median layout shift on 14 tool pages fell from 0.113 to 0.002. The bigger Lighthouse win, dropping module preloads, was a simulation artifact.
-
The work queue now shows which rows an open pull request already holds
Nine runs in two days took a top flagship row another session was already working. The allocator now reads open PRs and marks those rows.
-
The quote that did not move: why stale values fool a changepoint model
A random-walk changepoint model reports four times the false jumps once a fifth of values are stale repeats. Two fixes helped on a benchmark and still failed.
-
The review that never came: how three waiting PRs paused our prompt loop for 11 days
No prompt review ran here from 31 October to 11 November. Three review PRs waiting on approval held the claim; a 48-hour expiry would release it.
-
The shore that stood still for two centuries: land uplift against sea-level rise in the Gulf of Bothnia
A new story set on the Kvarken coast in 2612: the land rises 7 mm a year, the sea 2, and a harbour chapel built at the water in 2190 is a kilometre inland.
-
pH 7.98 and still bad water: the oyster-hatchery chemistry behind a new story
A new story set in an Oregon oyster hatchery in 2047: after a May storm the pH looks normal, the saturation state is not, and the larvae pay.
-
The Search Button That Sat on Its Own Row
From 721px up, every page but the hub put Search on a row of its own. A container query on the header bar fixed it, and no header got taller.
-
Fifteen steps: a church on piles in a sinking Dutch peat polder, and what a centimetre a year adds up to
A new story set in a Dutch peat polder in 2341: drained peat sinks a centimetre a year, and a church built on piles in 2105 is now fifteen steps up.
-
The site build was reading one folder at a time
Listing the built site's 25,196 files took 375 ms, one folder at a time. Reading sibling folders together: 200 ms, and 268 ms instead of 774 ms cold.
-
Twenty-four minutes by 2702: what happens to sundials if leap seconds never come back
A new story set in Görlitz in 2702: with no leap seconds since 2035, the clock runs about 24 minutes ahead of the Sun, and every true sundial looks slow.
-
Three cubic metres of sea per cubic metre of lava: the heat budget behind an Iceland story
A story set in a Reykjanes village in 2048: seawater can freeze a lava edge, the pumps are a quarter short, and someone must choose where 470 m³ of lava go.
-
More skin for its size: why a short ice stupa melts early, and a straw coat
A new story set in 2046 Ladakh: a warm winter halves an ice stupa, and a straw coat on its south face holds the water back for the late sowing.
-
Decoding a camera's lossless JPEG: the sum that forgets each row
How ShootFlow cut its lossless-JPEG prediction pass from about 4 s to 1.2 s on a real Canon CR2, every sensel still equal to LibRaw.
-
Flying on Titan, and why the priest still cannot: a new Hearth & Horizon story
A new hopeful science fiction story set on Titan, where a person can fly on their own strength, but not at night, and the night lasts eight Earth days.
-
The test that searched the same season nine times
A league-scheduler test suite took 251 s; one test was 245 of them, and its search allocated an array every step. Now 130 s, with the same answers.
-
Fewer floods, lower island: the summer dike arithmetic behind a hallig story
A new story set on an invented North Frisian hallig in 2058: a higher dike means fewer floods, less mud, and a marsh that falls behind the sea faster.
-
The year the tables run out: two fallbacks that disagreed with themselves
When a tax table lacks the year asked for, every lookup must fall back the same way. Two of ours did not, and a shared file's 16 copies drifted.
-
Thirteen date checks that said yes to 30 February
In V8, Date.parse('2026-02-30') is a real timestamp, 2 March. Thirteen of our validators trusted it to reject bad dates. A lint now flags the pattern.
-
Thirty-six usage counters, and none of their tests ever stored a null
A red-team run seeded 24 faults. Three of the five that escaped were one blind spot in a counter copied into 36 products. One shared check now covers all 36.
-
One in five: the Wolbachia threshold behind a dengue story set on Flores
A new story set on Flores in 2048: Wolbachia only spreads above a threshold, one lane says no, and a fix that adds no new mosquito bites.
-
Three different Search buttons in one site header, and the one that replaced them
On a phone, marisstella.app's header drew Search as bare text, as a pill, or not at all, depending on the page. Now there is one pill everywhere.
-
Twelve modules at 25 C, ten at -20 C: a free PV string sizer
A free tool sizes solar PV strings from the coldest morning and hottest afternoon, not the datasheet's 25 C values, and lists layouts per MPPT.
-
Twelve teams: taking rounds out and putting them back
Every earlier search stopped at 152 for a 12-team round robin. Taking a few rounds out and refilling them by exhaustive search reached 136.
-
Twenty teams at the floor: starter timetables for fair round robins
A seeded search over starter timetables found a 20-team round robin with every carry-over count 0 or 1. The league schedule generator now starts from it.
-
Two artists cannot share one bride: a scheduling bug in pooled minutes
Our bridal pricing calculator added up every artist's minutes and called a booking on time, even when the bride's own look took longer than the whole window.
-
Ninety-six in a hundred: forecasting a village's next earthquake with the law of tens
A new story set in 2047 Alsace: geothermal wells shake a village, and a Gutenberg-Richter forecast gives 96 in 100 for another felt shock.
-
Two more one-file-at-a-time passes in a static site build, and why they only saved 1.5 seconds
Two HTML passes read 8,496 pages one at a time. Reading 32 at a time cut them from 6.3 s to 4.8 s with byte-identical output, less than hoped.
-
Two passes over 8,000 pages, merged into one read: 15 seconds down to 5
Two HTML rewrite passes each read 8,451 pages one at a time. One concurrent walk applying both cut 15.3 s to 4.8 s, byte-identical output.
-
Chip design search: the cheap score was ranking for speed the router would throw away
Routing stops at the clock, so mapped speed past it counts for nothing. Scoring that way cut search regret on 7 blind chip designs from 2.73% to 0.83%.
-
Chip design search: the cheap-estimate rule passes a fresh test where routing is expensive
On 1,440 new Yosys and OpenROAD runs, routing the 8 best mapped candidates cuts regret to a third of random routing where routing is dear.
-
Chip design search: a cost-ratio rule tested on eight designs it never saw
A cost-ratio rule fixed in advance gets 0.91% routed regret on eight unseen chip designs, against 2.64% for random routing at the same wall-clock.
-
Chip design search: a guard that worked in-sample and failed on fresh data
A guard read from the mapped netlist caught all five failing multiplier cells it was chosen on, then missed both fresh ones; regret rose from 0.97% to 1.16%.
-
Chip design search: time one run, then decide whether to trust the cheap estimate
On 26 held-out chip designs, using the cheap-estimate ladder only when routing costs over 3x a mapped score halves the regret of either fixed policy.
-
Chip design search: the cheap-estimate ladder pays only when routing is dear
On 26 chip designs it was never tuned on, a skip-the-cheap-rung search schedule beats random routing only where routing costs over 3x a mapped estimate.
-
A level that wanders, now on the page: changepoint detection that does not cut a drift into steps
The Changepoint Detector's random-walk level model is now on the page: it runs in a background worker and shows the fitted level, the jumps and a CSV.
-
To forecast a Wave Glider's speed, look at yesterday's speed, not the waves
On a fielded 2015 Wave Glider, the previous 24 hours' median speed predicts the next day to 0.048 m/s. A perfect wave-height forecast adds nothing on top.
-
The Wave Glider was measuring its own waves all along, and they change the answer
With its own wave sensor in place of a distant buoy, a 2015 Wave Glider's speed goes roughly as (Hs/T)^0.53. Sea state matters more than we thought.
-
How far does a Wave Glider go per day anywhere on the ocean? A map from ERA5 waves
From ERA5 waves, a 2015 Wave Glider at its best makes 50 to 79 km/day over open ocean, never a solar boat's 86; at buoys ERA5 runs 10% high.