-
Who should have had that shift? A seniority report for the roster builder
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.
-
Why a sealed lid is not a safe jar: botulism spores and boiling water at altitude
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.
-
Why a shed's floor joist runs at 15% of its strength
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.
-
Why boiling well water raises its nitrate, and what that means for a formula-fed baby
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.
-
Why the cull rejected your picks: shootflow now says which threshold
shootflow cull --picks now breaks down the photographer's picks that it rejected by reason, so a miss points at the threshold to loosen.
-
Wind turns a barn's tie into a column: a $7.72 brace
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.
-
Winder stairs: a tread that passes by arc can fail by straight edge
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.
-
A solar boat that just reads yesterday's sea keeps three quarters of the oracle's gain
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.
-
Yoga studio bookkeeping from a bank CSV, and a Gusto misfile we found on the way
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.
-
Zero false alarms can mean zero power: changepoint detection on a cycle
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.
-
1,200 random walls, and no gap in the sheathing
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.
-
22 product pages that nothing linked to
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.
-
$549 off an 8x12 shed with stock parts, and what the height limit holds back
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.
-
A CI check that no venture page loads a third-party script, font or image
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.
-
A gradient-chosen weight plant found no room under a per-tensor bound
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.
-
A pilot-allocated LLM audit looks better than the oracle and is wrong
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.
-
We tried a quieter airline series to find a strike; it wasn't quiet
Ansett first class had no school-holiday swing, yet the change point detector still returned 29 to 37 changes around the 1989 pilots' strike.
-
A seasonal drift warning that also warns on nothing
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.
-
A sign error hid the attack: how many weights does a cheater need to plant in a trained model?
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%.
-
Fuzzing the bookkeeper's spreadsheet reader found a three-second stall
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.
-
A wandering series is not a staircase: two more real checks for the changepoint detector
Real internet-usage and temperature series show when a changepoint detector reports steps that are not there, and which switch fixes it.
-
An adjustable tail fixes the gust response of a scaled UAV model, in simulation
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.
-
An adjustable tail fixes what no fixed tail could: sub-scale cargo UAV modes
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.
-
Advanced framing: how much lumber does it really save?
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.
-
Aim at projects worth three hundred hours
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.
-
US airline miles, 1937 to 1960: one change, and it is the end of the war
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.
-
An 8 ft shed peak and a 6'6" door do not fit together
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.
-
Can a changepoint detector find an airline strike? A real-data check
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.
-
Audit by the coder's own label: when stratifying a hand-coded sample pays
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.
-
Bookkeeping for auto repair shops: six of thirteen lines had no account
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.
-
Baseboard splices cost board: a scarf-joint allowance in the framing takeoff
A baseboard run longer than one board needs splices, and a scarf joint overlaps. The wall-framing takeoff now counts that overlap.
-
Boil it once: a boil-water notice that made the nitrate worse (a 2047 story)
Boil It Once (Hearth & Horizon): a flooded village well has bacteria and 58 mg/L nitrate, and boiling for twenty minutes doubles the nitrate.
-
Bookkeeping for massage therapists and estheticians, from a bank statement
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.
-
Breaking the shed designer on purpose: what a fuzz test found in a framing engine
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.
-
Calendar files for leagues with several divisions
The multi-division league scheduler now exports an .ics calendar for the whole league, one division or one team.
-
Can a 16-point pilot certify a chip-flow search? Not yet, and here is what failed
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.
-
CI and application bloat: how much is useful and how much is deadweight
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.
-
Cleaning business bookkeeping from a bank statement: nine lines, $837.20 profit
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.
-
Gradient compression costs loss in proportion to its variance, in a small DiLoCo experiment
On a 413k-parameter GPT trained with DiLoCo, rand-k compression cost loss roughly in proportion to its variance factor omega = 1/p - 1.
-
Cull a shoot offline and get Lightroom star ratings, free
ShootFlow's cull command groups burst frames, keeps the sharpest, rejects blurred ones and writes XMP sidecars. Nothing is uploaded.
-
Did four European stock markets get more volatile together? A changepoint check
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.
-
Do the league calendar files hold up? 600 random leagues, 0 faults
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.
-
Do our calculator pages fetch files that exist? A sweep of 1,300 pages
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.
-
Do your stair treads span too far between stringers?
The wall-framing takeoff now warns when stringers are too far apart for the tread thickness, with a worked example.
-
Auditing LLM labels by the coder's own cells: no gain when the coder carries no signal
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%.
-
Does a seasonal-drift warning cry wolf on a clean series?
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.
-
Does a heel-to-toe centre of pressure help a humanoid carry faster?
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.
-
Does a load-sensing hydraulic pump pay for itself on a small excavator?
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.
-
A longer series did not help the changepoint detector tell a shock from a season
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.
-
Does an excavator's idle time decide which hydraulic circuit wins?
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.