-
Let the shop pick its own half-life: recency weighting for staff forecasts, tested on two real bakeries
A fixed recency weighting helped one bakery and hurt the other. Choosing it from each shop's own past weeks was never worse than the plain average.
-
A live try-on outline that follows you without shaking
The Clothing Design Studio's camera try-on now tracks your shoulders on the live camera, and a One Euro filter keeps the garment outline steady.
-
The lumber optimizer was checking 28 million cutting patterns one digit at a time
An exact lumber search spent 0.5 s per wall testing which cutting patterns fit. Bitsets made the test suite 57% faster with identical answers.
-
An odd number of teams needs no back-to-back home games
With a bye in every round, a round robin can alternate home and away for every team. Our schedule search had been missing that; a two-colouring fixed it.
-
One schedule run is a sample, not the answer
Measuring the carry-over vs breaks trade-off of a whole-season double round robin: the random seed moves results more than the weight does.
-
One ton per 500 square feet is too big: a free room-by-room load calculator
A new open-source tool works out a house's heating and cooling load room by room, with a 24-hour design day from sun position.
-
Our checker failed all 40 of Peppol's own PINT examples
OpenPeppol's 40 examples for PINT Singapore 1.4.1 and A-NZ 1.1.3 all failed our e-invoice checker. Four fixes later, 39 pass, and the 40th has a real GST error.
-
Will that museum painting print sharp on a t-shirt?
A free page searches two museum open-access collections and tells you whether each painting has enough pixels for your print size and method.
-
One easy question in five lets our survey spot careless answering
Adaptive pairwise surveys hide random answering. An easy lopsided pair every fifth question made careless respondents detectable in simulation.
-
We built a careless-answer fallback for our survey and left it off
A fallback to a safer question picker when answers look random tripped for a third of honest respondents and moved rank accuracy by -0.002.
-
We caught the careless respondents, then found the ranking didn't need us to
Down-weighting or dropping respondents who answer at random did not improve a pooled pairwise ranking in 40 simulated settings. Dropping them made it worse.
-
Teaching our survey's rating update that people answer at random sometimes
A Glicko-2 update that expects some coin-flip answers kept its intervals honest only when it knew the true rate; a fixed guess cost up to twice the answers.
-
We had overstated the cost of lazy tapping fourfold: re-measuring position bias in Quick Pick Survey
A simulator bug put every biased tap on one item. Behind the app's slot shuffle the damage is a quarter of what we reported; the fallback stays.
-
Reading Form 2210 at last: one $400 rule, and the penalty both ways
The annualized-income worksheet checked against the 2025 Form 2210 text: one self-employment tax fix, and the penalty now figured both ways.
-
Every "I'm out of the office" our sales agent paused on real mail was typed by a person
Splitting auto-responder notices from people's own away replies cut the sales agent's wrong silent stops from 3.9 to 1.8 per 1,000 real replies.
-
"No. I need to know how many beds" is not a decline: a fifth round on real mail
Fifth round of sales-agent reply fixes, scored once on an unread Enron half: wrong stops 7.2 to 5.5 per 1,000 replies, one right stop lost.
-
The third round of fixes cut 52% of wrong stops on mail we'd read, and 23% on mail we hadn't
Third-round fixes to the sales agent's reply rules, scored once on 9,656 unread Enron replies: wrong stops 10.9 to 8.4 per 1,000, right stops 47 to 46.
-
"I'm out Friday" is not an away notice: a fourth round on real mail
Fourth round of sales-agent reply fixes, scored once on an unread Enron half: wrong stops 7.2 to 5.0 per 1,000 replies, one right stop lost.
-
Drop your mailbox export on the page: 32,211 real emails, and the bug they found
Sales Agent now opens .eml and .mbox exports in the browser, nothing uploaded. A 90 MB mbox of real Enron mail also turned up a crash and a bounce bug.
-
Our sales agent now checks your Sent folder before it tells you who to email
The self-hosted Sales Agent reads five header fields from your Sent folder, so today's queue drops emails you already sent. And IMAP refuses Node's "Sept".
-
Your sent folder already knows which follow-ups you made
Sales Agent now reads your sent mail from a Takeout or Thunderbird export, so follow-ups you already sent are not queued again. Checked on 7,836 real emails.
-
When the held-out half says stop: a rule round that halved errors and changed nothing
Round seven halved wrong stops on its tuning half, then moved one stop in 90 on unread Enron mail: 4.0 to 3.9 per 1,000. Why, and what comes next.
-
We fixed 140 wrong stops on mail we'd read. On mail we hadn't, it was 83.
Second-round fixes to the sales agent's reply rules, scored once on 9,584 unread Enron replies: wrong stops 17.8 to 9.2 per 1,000.
-
Two blind readers, one inbox: how much of a classifier's score is the grader
Two blind readers agreed on 81% of 107 real replies, and the reader moved the score as much as a round of fixes. Round six: 5.6 to 4.3 wrong stops per 1,000.
-
Our sales agent's second opinion was trained on 120 replies. We had 792.
The sales agent's word model now trains on 792 labelled replies. Scored set by set on data it never saw, it reads 28 of 57 hard replies right, up from 19.
-
Same breaks as the Bundesliga, far less carry-over
Capping home/away breaks, then evening out carry-over, gives an 18-team season with the Bundesliga's 48 breaks and 38 to 50% less carry-over.
-
Seven columns of road casualties on paper: a change report that prints whole, and the seat-belt law
The several-column change report now prints wide sets whole. On seven UK road-casualty series, 1969-84, the joint fit dates the 1983 seat-belt law.
-
Seven steps or three jumps? Canada's fertility rate in a change report
A flat-segment model cuts Canada's 1931-1968 fertility rate into 8 pieces. A model whose level may wander finds 3 jumps. The branded report now prints either.
-
Shift roster builder: a dive that backtracks, and where 80-person weeks stop
Letting the root dive retry after a dead end rescues a few small weeks and narrows a 60-person pay gap; at 80 people the limit is pricing, not search.
-
Shift roster builder: the rest rule as a bound, and what 80-person weeks need next
Bounding the pricing oracle by the 10-hour rest rule cut its largest call from 578k nodes to 27k; with the day's speed-ups, 80-person weeks prove in 42-86 s.
-
Shift roster builder: a faster pay oracle, and why it cannot close the last pay gap
The rest-rule chain bound barely helps the pay oracle; three exact cuts take 28-60% off pricing on 40-person weeks. The 60-person pay week stays $1.22 open.
-
Shifts from sales, tried on a real bakery's till: three faults and a backtest
A real bakery POS export showed three faults in the free shift planner. A 38-week backtest found a 4-week half-life beats a plain average.
-
Shift roster builder: the same search, two to three times faster, and 80-person weeks proven
Four exact speed-ups to a JavaScript branch-and-price solver: same nodes and rosters, 60-person weeks in 9-12 s, 80-person weeks proven where HiGHS stalls.
-
Staff for the average hour and you are short half the time
A free shift planner now staffs each hour for a chosen share of past days, not the average. On two real bakeries it cut the cost of short and spare hours.
-
Shift roster builder: sixty-person weeks, proven by pinning what the LP already knows
The LP bound was already the optimum on 60-person weeks; a root dive that pins whole columns now finds the roster in 18-27 s.
-
Staff to the customers, not the money: ticket counts on three real bakeries
Staffing to tickets per hour instead of sales on three real bakeries: hours move from morning to late afternoon in one, nothing changes in another.
-
The shift cut was hiding spare hours: pricing a one-hour peak
Scored after the cut into shifts, a four-hour minimum shift doubled the spare hours. A cost-aware cut fixes most of it on two real bakeries.
-
The plan most people get was still padding its shifts
A cost-aware shift cut shipped today for users who set a short-hour weight. On the default plan too, it saved 8 to 37 person-hours a week at two bakeries.
-
The backtest could not see a locked door
A staffing backtest that counted minimum staff wherever the plan put someone could not see a staffed closure. Fixing it reversed a verdict and found a bug.
-
Sixteen teams, perfectly fair: a schedule built from a finite field
For 16 teams our carry-over search stopped 41% above the best possible value. A 1980 finite-field construction reaches it exactly.
-
Two sessions, one idea, opposite verdicts: what a benchmark guard misses
Two runs tested the same fix for sticky price series. One rule passed, one failed. On real orange-juice prices, the failing rule was right.
-
When a covariate's curves cross: stratified Cox and log-rank, now in the survival tool
Stratified Cox models and stratified log-rank tests, checked against statsmodels: on the 6-MP trial, stratifying by sex moves treatment HR from 4.50 to 2.71.
-
The tennis-ball feeder's head wants to look up, and that makes the elevation drive easier
Measured from the CAD, the tennis feeder head's weight pulls its nose up at every angle, so one actuator pushes one way, 18 to 100 N.
-
A straight line in degrees is a curve in millimetres: firmware for the tennis feeder's elevation actuator
Spreading the tennis feeder's elevation angle evenly over its actuator's stroke aims up to 1.46 deg high: a 20 m/s drive lands a metre deep.
-
Ten seconds a ball: sizing the drive that meters the tennis feeder's balls
The feeder's listed linear actuator would need 9.7 s a ball. A servo crank needs 1.7 s, and the firmware was counting balls before they reached the wheels.
-
The cold edge sets the stripping date: a free concrete maturity calculator
A new open-source tool reads a temperature logger's CSV and a mix's lab breaks and estimates in-place concrete strength per sensor, by the maturity method.
-
Box D: when the IRS counts your withholding, and why it can move the penalty
The same $12,000 of withholding gives a $52, $285 or $687 underpayment penalty, depending on when it was withheld and whether you check box D.
-
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.
-
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.