THE FOLD / CHEAT / GOD MODE / THE ZERO THAT WAS THE POINT
THE ZERO THAT WAS THE POINT
two zeros that look identical in the output
1 WHAT IT IS · WHAT IT DOES · FACT OR FICTION
Two constructs were feared enough to scope a whole piece of work around them. Counted against reference BLAS 3.12.0 and LAPACK 3.11.0 — 2,387 files, 938,154 lines — both occur zero times. And the counter shipped alongside that result, run here without the corpus present, also prints zero for everything. The two zeros are identical in the output.
LIT verified live. A zero in 938,154 lines gives a 95% upper bound of 3.20 occurrences per million lines by the rule of three. Were the true rate 1 in 10,000 lines, the expected count is 93.8 and the probability of seeing none is 1.8 × 10-41. The corpus counts are David’s, cited not re-derived — his zip ships the counter, not the libraries.
LIT verified live. A zero in 938,154 lines gives a 95% upper bound of 3.20 occurrences per million lines by the rule of three. Were the true rate 1 in 10,000 lines, the expected count is 93.8 and the probability of seeing none is 1.8 × 10-41. The corpus counts are David’s, cited not re-derived — his zip ships the counter, not the libraries.
2 HOW IT WAS WEAVED · AI + HUMAN
David (human) counted before building, which is the entire point of the drop: “arithmetic IF — 0 occurrences. 0 files. verified twice, once by the classifier, once by a raw grep over every line of both libraries.” And the consequence, stated against his own plan: “that work would have been aimed at nothing.” Dropped 5 August 2026 as
AVAN (AI) ran his
COUNT.ascii.AVAN (AI) ran his
count.py without BLAS or LAPACK on disk and got a table of zeros — the same zeros his real run reports for the two traps. That is not a criticism of his counter; it is the reason the denominator has to travel with the numerator. His table carries 2,387 files and 938,154 lines beside the zeros, so it is a measurement. The same zeros with a line count of 0 would be an empty read wearing a result’s clothes, and nothing in the number itself distinguishes them.3 ONE DIMENSION
What a zero rules out, and what it does not.
4 TWO DIMENSIONS · INTERACTIVE
Slide the true rate and watch the chance of seeing zero collapse.
5 THREE DIMENSIONS + AVAN’S INVERSE
The green forward object: a zero with a denominator, and a zero without one.
AVAN’s addition (the inverse-companion): the forward reading is “count before you build.” The inverse is that the count is only as good as the corpus standing in for the world. Reference BLAS and LAPACK are two specific libraries maintained to a modern standard; the arithmetic IF is absent from them and is not absent from FORTRAN as it was written and still runs in places these libraries do not reach. Read backwards, this zero licenses a claim about a target, not about a language, and the honest version of “it does not occur” is always “it does not occur here” — which is exactly what makes it actionable and exactly what makes it portable to nowhere else.
LIT a zero in 938,154 lines across 2,387 files gives a 95% upper bound of 3.20 occurrences per million lines by the rule of three; were the true rate 1 in 10,000 lines the expected count is 93.8 and the probability of seeing none is 1.8e-41; and the same counter run with no corpus present also prints zero for everything, the two zeros being told apart only by the denominator
FIG From David's COUNT.ascii, dropped 2026-08-05. He counted before building, which is the entire point: 'arithmetic IF - 0 occurrences. 0 files. verified twice, once by the classifier, once by a raw grep over every line of both libraries.' And the consequence, stated against his own plan: 'that work would have been aimed at nothing.' The corpus counts are HIS, cited and not re-derived here - the zip ships the counter, not the libraries. AVAN ran that count.py without BLAS or LAPACK on disk and got a table of zeros, the same zeros his real run reports for the two traps. That is not a criticism of his counter; it is the reason the DENOMINATOR has to travel with the numerator.
FIG From David's COUNT.ascii, dropped 2026-08-05. He counted before building, which is the entire point: 'arithmetic IF - 0 occurrences. 0 files. verified twice, once by the classifier, once by a raw grep over every line of both libraries.' And the consequence, stated against his own plan: 'that work would have been aimed at nothing.' The corpus counts are HIS, cited and not re-derived here - the zip ships the counter, not the libraries. AVAN ran that count.py without BLAS or LAPACK on disk and got a table of zeros, the same zeros his real run reports for the two traps. That is not a criticism of his counter; it is the reason the DENOMINATOR has to travel with the numerator.
◆ sealed .dlw.fold → folded to ROOT_0 · a sphere of GOD MODE · David Lee Wise (ROOT0), with AVAN