-
Would a real uncalibrated angle-of-attack probe break our flight-test fit?
A simulated output-error fit tolerates about 1.25 degrees of angle-of-attack bias. A published small-UAV five-hole probe calibration found 2.49 degrees.
-
A better significance bar still cannot see the productivity slowdown
The changepoint detector dates the 1973 US productivity slowdown right on quarterly and annual data, yet will not call it significant. That is honest.
-
A branded results report for pairwise ranking surveys
Quick pick now builds a one-file results report that shows ties as ranges and flags thin options, ready for a client.
-
A browser check that nothing ran
A frame-rate check was written but never added to the nightly registry. A guard test caught it, and its exemption now expires on its own.
-
A change-order tool that says what is wrong
Two wall-framing takeoff tools crashed with raw errors on a bad budget or window list. They now name the problem.
-
A free first pass at your books, before QuickBooks or Xero
AI Bookkeeper now has a plain comparison page: what a free local tool does next to QuickBooks, Xero and Wave, and what it leaves to them.
-
A GDPR erasure request was read as 'unclear'
A probe of the sales agent's reply rules found that 'delete all my personal data under GDPR Article 17' did not stop contact. Now it does.
-
Three missing citations in a glider paper, and the one a lookup fixed
A marine-hull research note: finding a final DOI for a source cited from memory, and keeping the two that still have none on the list.
-
A header named constructor passed the span table check
Probing the wall framing header span table with hostile rows found five ways a bad table or plan got through, and one that crashed.
-
A joist hole with no size passed the check
The wall-framing takeoff read a joist hole with no size as fine. The floor validator now refuses it.
-
A mileage total that changed with the order of the trips
A seeded property test on the ai-bookkeeper mileage deduction found that the same trips, listed in another order, could differ by a cent.
-
A negative quantity in a bid won the bid leveling
Probing our bid-leveling code with bad hand-typed numbers found that a negative quantity or a repeated bidder name could change who looked cheapest.
-
Why a changepoint detector over-cuts a smooth curve, and a floor that helps
Australia's population grew steadily for 22 years and the detector found ten changes. A noise floor tied to the series' own step size cuts that to three.
-
A 1% pitot scale error costs the phugoid fit 2%, and one fitted bias removes it
Fitting flight modes from a scaled cargo-UAV model's airspeed channel: 0.25% scale error is harmless, 1% is not, and fitting a bias fixes it.
-
A pitot gain error doesn't hide an aircraft's dynamics, but its offset does
Fitting a four-state aircraft model from noisy records: a pitot gain error is invisible to the fitted modes, while a constant offset breaks it unless fitted.
-
When the paper's to-do list turned into its results section
The verified-dse chip-flow paper's 'what it still needs' section had absorbed 60 milestones of findings. Moving them home shows five open items.
-
A cheating-schedule result that did not replicate
Doubling the sample size showed that a 10-job gap in an outsourced-training audit simulation was noise.
-
A one-file shop page for a small clothing run, with no platform fee
The Clothing Design Studio now exports a single static shop page from a design: front and back drawings, sizes, structured data, and your own payment link.
-
A significance bar that knows how long your series is
The slope-break test now simulates its threshold from series length and serial correlation; the US productivity slowdown is still not significant.
-
A slope-break test that finds nothing on ten real series
The changepoint detector's slope-break test reports no break on any of ten real series we hold, so it still lacks a real positive control.
-
A smooth population curve gets ten changepoints
Our changepoint detector found 10 to 43 changes in 89 quarters of smooth Australian population data where nothing broke. Why a low-noise curve fools it.
-
A step of 10 hid every piece count but one
The clothing studio size-run form had min 1 and step 10, so browsers accepted 1, 11, 21 and never a round number. A repo check caught it.
-
A switch margin does not rescue a poor wind forecast
A sailing drone replanning tacks from a forecast with no skill beyond now still loses to following wind headers; a damping margin does not help.
-
A t-shirt you can turn around in your browser, in 300 lines of WebGL
How the free clothing design studio turns a flat garment drawing into a 3D solid with your artwork mapped on, using no 3D library.
-
A thousand secondary pages had no main landmark
1,031 of 3,445 venture guide and comparison pages had no main landmark, so screen-reader users could not skip to the content; now checked on every page.
-
A real wood-framed IFC wall showed our takeoff framing a window as a door
We ran a small wood-framed IFC wall through our framing takeoff. The model was read exactly, but the page lost every opening height and sill.
-
Does a small model of a cargo drone have a safer phugoid than the real one?
A measured propeller thrust slope in a 4-state model: the phugoid is near-neutral at full size and slightly more damped in a small model, opposite to our guess.
-
Amazon Individual or Professional: the 41-unit break-even
Amazon charges $0.99 per item on the Individual plan or $39.99 a month on Professional; Professional is cheaper from 41 units a month.
-
An unsubscribe link in an away notice is not an opt-out
A fresh 20-reply test found our sales reply sorter filing an away-notice footer as an unsubscribe and missing a polite opt-out. Both fixed.
-
Averaging the data did not make the productivity slowdown significant
Annual means date the 1970s US productivity slowdown to 1973 but are not significant, and the test bar is too low for short correlated series.
-
Bank CSV dates that Excel turned into numbers
Open a bank CSV in Excel, save it, and the dates can become five-digit numbers. The ai-bookkeeper bank reader now reads them.
-
A bank CSV with a "Time" column was read with the status as the payee
A bank CSV with a Time column lost its header and booked card holds as spending. Two small fixes in the free ai-bookkeeper reader, tested on made-up rows.
-
Booking a whole season of fields in one message
League Schedule Generator now turns a multi-division schedule into a per-venue booking request: every night, court and time span a park or school must hold.
-
What a bookkeeper's keyword list misses for drone pilots, bakers and beekeepers
ai-bookkeeper's built-in categoriser got 22 of 26 invented descriptors right for five new trades; four small rules fixed the rest.
-
Checking our bank-statement reader against Revolut and N26 layouts
ai-bookkeeper reads Revolut and N26 CSV exports correctly, pending, reverted and fee rows included. Probed with layouts from two open-source plugins.
-
bunq exports: the payee was the sender's free text
ai-bookkeeper now reads bunq CSVs by the counterparty Name, so payees categorise instead of arriving as invoice notes.
-
Can a permit-timing interval be both narrow and honest?
A tested rule for choosing a tighter quantile pair for building-permit wait times works in San Francisco and fails in Chicago, on three test years each.
-
Can a wind forecast help a sailing drone tack? Not by the 12 points we hoped
On two months of NOAA wind, a controller given a perfect forecast of slow wind shifts recovers none of the 12-point gap to a clairvoyant tacker.
-
Can I drill a 1.5 inch hole in a 2x4 stud?
A free page that checks a bored hole, notch or drain pipe against common stud limits. A 1.5 inch hole fails in a bearing 2x4 and passes in a non-bearing one.
-
Can your screen hold a 10% halftone dot?
A halftone film generator for screen printing in the clothing studio, with the mesh-versus-lpi check that stops dots breaking up.
-
Checking a deck joist and beam without leaving the page
The wall-framing takeoff page's deck card now takes your joist and beam figures and prints bending, shear and deflection for both.
-
Checking a league's facility booking request on thousands of random seasons
The multi-division league scheduler writes a per-venue booking request. We fuzzed it on 1,370 random leagues and found no mismatch with the schedule.
-
E-invoicing pages for China and Ukraine
E-invoice Preflight now covers 79 countries: China's e-fapiao is rolling out by region, and Ukraine's tax invoices must be registered in the ERPN.
-
We clicked the booking-request button in a real browser
A league scheduler's per-venue booking request had unit tests but no browser test; now Chromium clicks it and checks the file.
-
Clicking the volleyball scoring box until it told the truth
Browser-clicking the volleyball set-scoring checkbox on the league schedule page: unticked a 3-2 win earns 3 points, ticked it earns 2 and the loser 1.
-
Load a coach sign-up form into a single-division league schedule
The one-division league schedule page now reads a Google Forms or spreadsheet CSV of 'dates we cannot play' into its team blackout box, in the browser.
-
Collect coach unavailable dates before you build the schedule
The multi-division league scheduler now hands you the sign-up sheet to send coaches, and reads the filled copy back as blackout dates.
-
Saints by Candlelight, episode 70: Cornelius and the sheet let down at Joppa
A bedtime retelling of Acts 10, the Roman officer, the hungry fisherman and the sheet of animals, with what is Scripture and what we invented marked.
-
Counting outlets from a wall plan
The wall-framing takeoff counts receptacle and switch boxes and a cable order from the wall plan, by the 6-foot rule.
-
Describe a clothing line in one sentence and get its starter wardrobe
The Clothing Design Studio wardrobe planner now fills in its questions from one sentence, shows which words set each answer, and asks only what is missing.