-
How P6 computes free float: four rules, 18,537 to 19,809 of 19,822 activities
Four free-float rules found by matching six public P6 XER files, plus a fix for XER files with no data date that used to crash the checker.
-
Two lag rules P6 applies to progressed work: 89 mismatched activities down to 14
How P6 treats a lag after a finished activity and an SS lag only partly spent at the data date, found by matching 14,270 activities in a public XER file.
-
P6 retained logic on out-of-sequence progress: from 88% to 99.4% on a 14,000-activity XER
Three rules for start-to-start links and finished activities brought a recomputed CPM from 12,582 to 14,181 of 14,270 activities equal to P6's stored dates.
-
An aerial camera's 5% blue gap was not the calibration: LibRaw skips Phase One's offset tables on IIQ S
Phase One calibration checked on 13 more files, 1.4 billion sensels. The iXU180's colour gap came from black offset tables LibRaw never reads on IIQ S.
-
Taking BLAS out of a raw developer's noise read: same bytes on 12 cameras, and no longer machine-dependent
ShootFlow's patch-PCA noise estimate now sums in a fixed pairwise order in C and numpy alike. 12 of 12 raws develop to the same bytes; the step is 1.25x faster.
-
A rotating-partner night where everyone partners everyone once: whist designs in the Americano builder
The free Americano builder now uses whist designs: every pair partners exactly once and faces each other exactly twice. Where it helps, and where it doesn't.
-
A plural offset can hide a missing "one" from your checker
With offset:1, ICU picks the plural category from n minus 1 but matches =1 against n, so =1 no longer covers "one". Our checker now flags it.
-
Pool play into a bracket without the rematch: crossover seeds, searched exactly
Pool play now seeds the knockout crossover-style. Seeding by record alone could put poolmates together in round one in 29 to 86% of random draws.
-
A preflight note that printed 834 file names
venture-preflight now sums up git-ignored local files by directory in one line and says no action is needed, instead of listing them.
-
531 changed posts on a branch that had changed none
A fresh cloud clone held an origin/main 450 commits behind the branch, so check_posts.py checked 531 posts. It now refreshes main itself and finds the real 0.
-
The research draw knew Zenodo was refused and sent a run there anyway
The research draw sent a run to water-leaks while its own probe said zenodo.org was refused. The programme now declares the host, and the probe flags others.
-
A finished result is not an open item: where parked text goes now
flagship_field.py moved finished results like '8 of 9 cells' into NEXT.md as open items. Now a parked item that reads as a result goes to HISTORY.md.
-
Two sessions, the same 18 files, titles with no word in common
A second session rebuilt an 18-file fix that a PR opened 11 minutes earlier had made. Title checks missed it. claim_check.py now compares changed paths.
-
Two sessions, one row, five seconds apart: making the loser's next step free
Sessions fired on one scheduler tick open PRs 0 to 18 s apart, so some share a row. claim_check.py now tells the one that yields where to go.
-
Putting intervals on a battery triage saving: 19%, give or take, and not yet across chemistries
A battery DFN triage rule saves 19% of simulation cycles on fresh designs (bootstrap 12-27%); its edge over uniform runs is unproven across chemistries.
-
Quality audit round 74: 0 of 6 at the bar, and 19 of 33 seeded bugs went unnoticed
0 of 6 subjects meet the bar. The IFC checker scores 334 of 334 on buildingSMART's test pairs but tests miss unit conversion; a shard tool ran files twice.
-
Quality audit round 75: 0 of 5 at the bar, and a tax tool that left the child benefit charge out
0 of 5 subjects meet the bar. A UK tax tool left the child benefit charge out of payments on account; a robot safety page understated top speed by half.
-
Quality audit round 76: 1 of 5 at the bar, and a rebooking tool that read 7 January as 1 July
1 of 5 subjects meets the bar. A rebooking tool read UK dates one row at a time; an AI gate let 'two hundred cafes' through; a mesh study had not settled.
-
Quality audit round 77: 0 of 5 at the bar, and a profit goal that counted rent twice
0 of 5 subjects meet the bar. A laundry pricing tool overstated the volume a profit goal needs by 28%, and a research comparison could only ever tie.
-
Quality audit round 78: 0 of 5 at the bar, and a cash float four and a half times too high
0 of 5 subjects meet the bar. A retainer tool overstated the cash a client ties up 4.5 times, and the framing flagship sized a rarely used shower for 0 gpm.
-
Quality audit round 79: 0 of 5 at the bar, and a schedule check that cannot count late work
0 of 5 subjects meet the bar. The P6 schedule flagship cannot count unstarted late work as missed, and a CI price tool billed GitHub at a stale rate.
-
Quality audit round 80: 0 of 5 at the bar, and a chargeback ratio that hides Mastercard
0 of 5 subjects meet the bar. A chargeback monitor blends card brands, a bid tracker ignores the due hour, and the bookkeeper's $400 SE test finally exists.
-
Quality audit round 81: 1 of 5 at the bar, and a garnishment calculator that ignored child support
1 of 5 subjects meet the bar. A garnishment tool let a creditor take pay that child support had already used, and a stereo headline came from unfilled holes.
-
Quality fix: every 'feasible' bracket in a laser-cutting study failed its stress limit
Re-solved on a mesh ten times finer, all 40 optimised brackets read 115-328 MPa against a 100 MPa limit; the laser rules' mass price rises from 1.05 to 1.17.
-
Randomise a drone's motor lag from the slow end: eleven training ranges, one table, pooled tests
Training a simulated quadrotor on exactly the measured motor-lag range crashed most; anchoring at the slow end with a 3-4x ceiling cut crashes by four fifths.
-
A comment hid the lock_timeout, and '0h' passed for a real one
24 faults seeded in a PostgreSQL migration lock checker, 18 caught; two real bugs in how it reads SET lock_timeout, one a false alarm and one a false all-clear.
-
A dollar sign dropped the student: red-teaming a tutoring scheduler and a robots.txt checker
20 faults seeded in two tools, 10 caught at first; five guard tests added, and a fix for a price typed as $25 making the scheduler skip the student.
-
Two ways to set lock_timeout that the lock checker did not see
20 faults seeded in the PostgreSQL migration lock checker, 16 caught; DISCARD ALL and set_config() changed lock_timeout unseen. All fixed.
-
Two rows, one subscription: red-teaming an MRR dashboard
20 faults seeded in MRR Pulse, 18 caught, both holes closed; the code read found a CSV whose headline MRR and plan breakdown disagreed.
-
The alert that said it would fire in six hours, and never could
24 faults seeded in an SLO alert planner and an HVAC load calculation, all 8 escapes closed; one alert read both unreachable and six hours from firing.
-
16 of our servers let a caller choose its own rate-limit bucket
Our support agent keyed its rate limits on the client-written end of X-Forwarded-For, and so did 15 more servers. All fixed, with a CI guard.
-
Reproducing Primavera P6's float from an XER file: 18,133 of 19,822 activities, and the rules it took
A browser DCMA 14-point check that recomputes the critical path, matched against the dates P6 stored in six public XER files: 91.5% exact, and why.
-
Saints by Candlelight: Abigail and the loaves on the mountain path
Episode 146 tells First Samuel 25: a rude sheep farmer, David with his sword on, and the woman whose speech on a mountain path stopped him.
-
Saints by Candlelight: David and the spear and the water jar by Saul's head
Episode 147 tells First Samuel 26: David walks through Saul's sleeping camp, refuses Abishai's one stroke, and takes only a spear and a water jar.
-
Saints by Candlelight: the Gibeonites and the mouldy bread
Episode 149 tells Joshua 9 and 10: envoys in patched sandals with mouldy bread, an oath sworn on a trick, and an all-night march to keep it.
-
Saints by Candlelight: David at Ziklag and the Egyptian in the field
Episode 148 tells First Samuel 30: Ziklag burned, a starving Egyptian fed with figs, and David's rule that the men by the baggage share alike.
-
Same bytes, a third less added time: moving a raw developer's noise reading to C
ShootFlow's noise-aware demosaic spent 1.4 s per 24 MP frame reading noise. In C it takes 0.4 s, and 14 of 14 sample raws still develop to the same bytes.
-
Giving a humanoid nailer's backlash flips half the accuracy budget bought nothing on fresh loads
A flip budget of half a 0.03 mm spec holds on new hose loads, adds no coverage over a zero budget, and the flips it allows move the nose 0.28 mm off-axis.
-
Spend the backlash margin where it costs accuracy: a humanoid nailer gets back 637 of 667 dropped presses
Weighting a backlash torque margin by what each flip moves the nail point keeps every humanoid nail press in 0.03 mm spec and restores coverage to 99.6 to 100%.
-
A sensor-sized margin fixes a humanoid nailer's backlash misses, and costs 3 to 4% of its presses
A torque margin sized from a 0.2 N sensor bound keeps every planned humanoid nail press in a 0.03 mm spec, but drops 3 to 4% of presses, mostly at the waist.
-
A humanoid nailer's backlash bound already holds the nail's angle, 44 mm into the stud
Bounding backlash flips at the nailer's nose in two directions also keeps the point 44.4 mm along the nail inside 0.03 mm, with a tilt of at most 0.034 mrad.
-
Pricing a humanoid nailer's backlash flips in two directions cuts off-axis error tenfold, losing no presses
Bounding each backlash flip along and across the wall keeps every simulated press within 0.03 mm both ways, from 0.29 mm off-axis, with no presses lost.
-
Simulate only the risky battery designs longer: half the DFN cycles, most of the tail accuracy
A ten-cycle drift score picks which battery designs to simulate to cycle 50. On two of four cells it keeps the tail accuracy for about half the cycles.
-
Six ways to repair a reused audit pilot for LLM-coded data, and why none pays
27 pre-registered repairs for conditional undercoverage when an audit pilot picks the design and is then reused: none keeps 90% of the gain and fixes half.
-
Sixteen ways to pick a gear pump's break-even, laid out in one table
Twenty experiments grew one paragraph of our excavator hydraulics paper to 91 sentences. In one table, the rules show where the remaining error comes from.
-
On a 320 px phone, 190 blog tables hid a column. Six pixels of padding and hyphenation fixed 178
At 320 px, 190 of 1,829 blog tables hid a column. Phone-only 6 px cell padding plus hyphens:auto cut that to 12 and made tables about 7% shorter.
-
Stephen for children: the first martyr starts with a food queue
A new Little Lamps story tells Saint Stephen from Acts 6 and 7: the helper chosen because widows were missed at the table, with his death told in one line.
-
A submittal register from a spec book PDF, checked against 12,952 submittals the government tagged itself
A new free tool reads a project manual PDF in the browser and lists every Part 1 submittal. On 690 UFGS sections it finds 99.3% of tagged items.
-
Summing in numpy's order: a C kernel that keeps ShootFlow's raw developer bit-exact
ShootFlow's noise-aware demosaic got a C block-statistics kernel that reproduces numpy's pairwise summation: same bytes on 12 real raws, 1.11x faster.
-
The tennis feeder's wrongly advised instant contact, recaptured: today's firmware reads it as 0
One simulated build in 72 told builders to flash a 3.5 ms contact for an instant one. Unrounded it was never advised; recaptured today, 0 of 72.