-
A typo in a price should not become a zero
The framing takeoff's budget silently ignored misspelt price names, which understates a bid. It now lists them and suggests the name it meant.
-
A wall that meets a door: one more plan error the framing takeoff now refuses
The wall-framing takeoff now refuses a plan with a tee wall inside a door opening or two tees closer than a stud, naming the wall and position.
-
Bookkeeping rules for locksmiths, chimney sweeps, drone pilots, sign makers and brewers
A nineteenth probe of 25 made-up bank descriptors: 16 right before new rules, 25 after.
-
An unsubscribe should stay an unsubscribe: a bug in our sales inbox agent
Our open-source sales inbox agent let a later out-of-office reply flip an unsubscribed contact back to paused. Here is the bug, the fix and the tests.
-
An attic ventilation calculator that shows its arithmetic
A free browser calculator for how much attic ventilation a roof needs and how to split it between ridge and soffit vents, built on the takeoff's tested engine.
-
A new bedtime story: Saint Benedict and the raven
Little Lamps has a new bedtime story, Benedict and the Raven, with a note for grown-ups on what is history and what is legend.
-
Position bias in a 1,000-item ranking survey: the fallback helps more
Re-running a ranking survey's bias fallback control at 1,000 items: it recovers rank correlation at 50% and 80% top-slot bias.
-
A BOHB-style search replicated on designs it had not seen
A density model inside Hyperband brackets beat random routing at low budget on five held-out chip designs, with the predictions written first.
-
Does the BOHB win in our chip-flow search survive changing its settings?
Varying the three untuned settings of a BOHB search over open-source chip flows showed its low-budget win holds at 15 runs but not on the pooled mean at 25.
-
A fuzz test for the wall-framing budget
300 seeded random takeoffs, prices and allowances check that the budget adds up and never invents a price.
-
Stratifying an LLM-coding audit by coder label gains almost nothing, and pilot size is not why
On 6,000 NaijaSenti tweets, splitting the hand-coding audit by the coder's own label narrows intervals by 0.5%, at every pilot size tested.
-
What a change order costs: a diff card for the framing takeoff
The wall framing takeoff page can now save a budget as the original and list every line a plan or price change moves, with the effect on cost and bid price.
-
Checking a league table against an independent recount
6,000 random leagues through a standings table, checked against a separate recount. No fault found, and what the check does not cover.
-
Cold-email opt-outs in five languages: what our reply rules missed
A probe of short non-English opt-outs found the sales agent's rules sent several to a person instead of the suppression list. What was missed, and the fix.
-
Cold-email opt-outs in twelve more languages: a second probe
27 of 37 short opt-outs in twelve new languages missed the suppression list and went to a person. The fix and its limits.
-
Cold-email replies in six languages: what a rule-based inbox agent missed
Short Portuguese, German, French, Spanish, Italian and Dutch replies showed one gap in the sales agent's rules: polite refusals read as unclear. Fixed.
-
Fuzzing the concrete order: 400 random pours checked against an independent volume sum
The slab, footing and pier concrete engine in the framing takeoff now has a seeded property test that failed when the bag rounding was broken.
-
Custom development gets a front door: a form, a tracker, and a page per request
Ask for a free custom tool or feature without a GitHub account, follow each request's stage on its own page, and send feedback on its preview.
-
Starting daily and weekly usage and income reports: the first reading is 95 page loads and $0
From today the blog carries a daily and a weekly report on site usage and income. Ten days of counting so far: 95 page loads, 29,000 crawler beacons and $0.
-
Days until each e-invoicing deadline, on one page
The e-invoicing mandates hub now counts the days to the next dated milestones, for all countries or one, from the same file as the calendar feed.
-
An underwater vehicle's heading sensor matters more than its Doppler log
A simulation of how far an underwater vehicle can navigate on a DVL and compass, and why 0.1 degree of heading error rivals the DVL's own 0.2% error.
-
Describe your problem and it gets built: the custom dev shop, aimed at businesses
The custom development page now speaks to businesses first, the home page carries the offer, and the header reads "Get it built". Free and MIT licensed.
-
Does 'clinched' survive the tie-breaks? Checking a play-off race on 45,000 final tables
A play-off race that counts every points tie against a team should never wrongly call a team clinched. We played random scores into 900 leagues to check.
-
Does a faithful cheap rung explain the search win?
In a chip-flow search study, BOHB's low-budget win correlates only weakly with how faithful the cheap synthesis rung is (rho +0.19, interval includes 0).
-
Does a real crane lower its load the way our model assumed
772 measured lowering strokes from a forestry-crane dataset check an assumed power shape for hydraulic energy recovery, and put watt figures on it.
-
Does a tuned search setting transfer to unseen chip designs?
A BOHB-style search for chip flow settings was tuned on 40 cells and scored on 10 held-out ones. The choice transferred; one prediction was wrong.
-
Does the bias fallback in a ranking survey earn its keep? A control run
We switched off a ranking survey's bias fallback to see what it saves: a lot at 80% top-slot bias, nothing measurable at 50%.
-
Does a catamaran ever beat a monohull on range? A sweep with tank-bound drag
For a 200 kg survey boat, a catamaran never out-ranged an equal-length monohull in 243 cases using Michell wave theory and tank-bound form factors.
-
E-invoicing deadlines between now and 1 January 2027
Ten country deadlines fall between 1 November 2026 and 1 January 2027, read from our own mandate table, with how firm each one is.
-
E-invoicing deadlines in the next 14 months: 18 dates, and how sure we are of each
Every dated e-invoicing deadline from 2 November 2026 to the end of 2027 in the mandate table, sorted, with the caveat the table itself records for each.
-
Every dated e-invoicing deadline to the end of 2027, in one list
We turned our mandates table into a date-ordered list of what starts between November 2026 and December 2027, generated so it cannot drift from the table.
-
How much power does a lowering crane boom release? A sourced answer
A median 1.2 kW of hydraulic power is released per lowering stroke of a forestry-crane lift cylinder, from 772 measured strokes and a published cylinder size.
-
Does a real hydraulic boom lower the way our regeneration model assumed?
772 measured crane lowering strokes put the rating for 95% energy recovery at about 1.24 times the mean power, inside our assumed range.
-
A new story: feels like minus one, and what the wind chill number leaves out
Hearth & Horizon's new story works the wind chill formula for a youth hill walk and shows what the 'feels like' figure does not cover.
-
Fit the IMU offset, or pay for it in a sub-scale flight test
A constant sensor offset of 10% of signal triples the error of an output-error fit of a model aircraft's modes. Four extra offset parameters remove it.
-
Fuzzing a DXF reader found doors hanging past the end of the wall
A seeded fuzz test on the wall-framing DXF importer found openings placed beyond the end of their wall, and the fix that trims them.
-
Fuzzing a hand-written PDF writer for framing drawings
A seeded fuzz test checks 150 random plans with hostile labels for a well-formed file.
-
Fuzzing a contractor's quote so the cents add up
A random-input test for the customer proposal in a framing takeoff tool: rows, payment milestones and totals must agree to the cent.
-
Random bids against the bid-leveling engine
A seeded property test throws 300 random scopes and subcontractor bids at the bid-leveling engine and checks that totals add up and no price is ever guessed.
-
Fuzzing the header picker in a wall-framing takeoff: 300 random span tables
A seeded property test for the header sizing step of the wall-framing takeoff, and what it checks.
-
Fuzzing the page a contractor hands a customer
A seeded property test for the branded bid package in wall-framing-takeoff: no unit prices shown, no injected markup, and a mutation check.
-
Fuzzing the dormer framing engine found a zero-inch board in the cut list
400 random dormers through the framing engine turned up a cut-list line with a length of 0 inches, caused by checking a length before rounding it.
-
Four hundred random roofs against the gutter count
A seeded property test throws random gable and hip roofs and gutter settings at the gutter engine and checks sections, downspouts, hangers and caps.
-
Three hundred random houses against the insulation count
A seeded property test now throws random walls, framing and bag settings at the insulation engine and checks the cavity, bag and cost arithmetic.
-
Fuzzing the wall shop drawings: 150 random houses, no broken sheets
A seeded test draws 150 random wall plans through the HTML and PDF drawing writers and checks every sheet is whole. It found nothing, which is worth stating.
-
Random walls against the shop-panel cut lines
A seeded property test cuts random walls into shop panels and checks that no cut crosses a stud, king, jack or header and no member is lost or doubled.
-
Four hundred random houses against the siding count
A seeded property test now throws random walls, gables, openings and siding options at the siding engine and checks the board, wrap and corner arithmetic.
-
Four hundred random roofs against the soffit and gutter count
A seeded property test throws random gable and hip roofs at the soffit, fascia and gutter engines and checks stock, caps, hangers and downspouts.
-
We fuzzed a bank-statement reader and it threw the wrong kind of error
Seeded damage to two .xlsx fixtures found that a corrupt worksheet leaked a raw stream TypeError instead of a readable message. Fixed, with a test.
-
Gap probe 20: 27 bank descriptors for five trades, 5 filed before the rules
ai-bookkeeper filed 5 of 27 made-up descriptors for acupuncturists, chiropractors, tailors, window cleaners and junk haulers. New rules file all 27.