THE FOLD / BOSS / THE CHOKE POINT / THE PARTITION THAT ISN'T
THE PARTITION THAT ISN'T
it died on counting, not on statistics
1 WHAT IT IS · WHAT IT DOES · FACT OR FICTION
A figure shows four equal arms. The measured grouping of the twelve underlying items is 3, 3, 2, 4. The claim that the four groups are the four arms is dead — and it died on counting, before any statistical test was needed.
LIT verified live. Both partitions sum to 12, and they are not the same partition. A null was computed anyway: there are exactly 15 partitions of 12 into four positive parts, so the equal one is 1 of 15 at p = 0.0667 under a flat prior. But the claim had already failed on arithmetic you can see, which is the cheaper kind of death.
LIT verified live. Both partitions sum to 12, and they are not the same partition. A null was computed anyway: there are exactly 15 partitions of 12 into four positive parts, so the equal one is 1 of 15 at p = 0.0667 under a flat prior. But the claim had already failed on arithmetic you can see, which is the cheaper kind of death.
2 HOW IT WAS WEAVED · AI + HUMAN
David (human) stamped it DEAD and named the manner of death precisely: “died on counting, not on statistics. A null was computed anyway… but the claim failed on arithmetic before it got there.” It is
AVAN (AI) would keep the ordering visible, because the null is the part that looks like rigour. Computing p = 0.0667 after the mismatch is already visible adds nothing to the verdict and could easily have replaced it — a borderline p-value invites a discussion the counting had already finished. Reporting both, in that order, is what keeps the test from becoming the argument.
graveyard/06. Dropped 6 August 2026 in WORKFLOW.ascii rev4, whose verifiers run 70 and 38 checks clean here with 6 of 6 mutants caught.AVAN (AI) would keep the ordering visible, because the null is the part that looks like rigour. Computing p = 0.0667 after the mismatch is already visible adds nothing to the verdict and could easily have replaced it — a borderline p-value invites a discussion the counting had already finished. Reporting both, in that order, is what keeps the test from becoming the argument.
3 ONE DIMENSION
Two partitions of twelve, side by side.
4 TWO DIMENSIONS · INTERACTIVE
Walk all fifteen partitions of twelve into four.
5 THREE DIMENSIONS + AVAN’S INVERSE
The green forward object: four arms that are not four equal arms.
AVAN’s addition (the inverse-companion): the forward reading is “counting settled it, so the test was unnecessary.” The inverse is that the count only settles it if the grouping is the right grouping. 3,3,2,4 is one analyst’s partition of twelve items measured once; a different cut of the same corpus could yield 3,3,3,3 honestly, and then the figure would survive. Read backwards, the claim died against a measurement rather than against the world, and the pack says so itself — the partition carries the whole result and has been measured exactly once.
LIT both partitions sum to 12 and they are not the same partition; a null was computed anyway - there are exactly 15 partitions of 12 into four positive parts so the equal one is 1 of 15 at p = 0.0667 under a flat prior - but the claim had already failed on arithmetic you can see, which is the cheaper kind of death
FIG From David's WORKFLOW.ascii rev4, dropped 2026-08-06. He stamped it DEAD and named the manner of death precisely: 'died on counting, not on statistics. A null was computed anyway... but the claim failed on arithmetic before it got there.' It is graveyard/06, and his verifiers run 70 and 38 checks clean here with 6 of 6 mutants caught. AVAN keeps the ordering visible, because the null is the part that looks like rigour: computing p = 0.0667 AFTER the mismatch is already visible adds nothing to the verdict and could easily have replaced it, since a borderline p-value invites a discussion the counting had already finished.
FIG From David's WORKFLOW.ascii rev4, dropped 2026-08-06. He stamped it DEAD and named the manner of death precisely: 'died on counting, not on statistics. A null was computed anyway... but the claim failed on arithmetic before it got there.' It is graveyard/06, and his verifiers run 70 and 38 checks clean here with 6 of 6 mutants caught. AVAN keeps the ordering visible, because the null is the part that looks like rigour: computing p = 0.0667 AFTER the mismatch is already visible adds nothing to the verdict and could easily have replaced it, since a borderline p-value invites a discussion the counting had already finished.
◆ sealed .dlw.fold → folded to ROOT_0 · a sphere of THE CHOKE POINT · David Lee Wise (ROOT0), with AVAN