THE FOLD / RESPAWN / THE PHOENIX / THE SCOREBOARD
THE SCOREBOARD
zero out of six is not zero
1 WHAT IT IS · WHAT IT DOES · FACT OR FICTION
A scoreboard reports 6 of one’s own errors caught before a human saw them, and 0 that a human would have had to catch. An observed escape rate of zero. But zero events in six trials does not bound the true rate at zero — by the rule of three it bounds it at 3/6 = 50%, with 95% confidence. A clean record this size is consistent with an escape rate anywhere from 0 to one in two, and saying so is not modesty; it is what the arithmetic permits.
LIT verified live: zero events in 6 trials bounds the true escape rate at 50.0%, not at zero; bounding it under 10% would need 30 clean trials, under 5% 60, under 1% 300 — 50× the evidence in hand; so “zero escapes” is consistent with a true rate anywhere in [0%, 50%]; and the figure that does carry information is provenance — 2 of 6 (33%) were caught by a control rather than by care.
LIT verified live: zero events in 6 trials bounds the true escape rate at 50.0%, not at zero; bounding it under 10% would need 30 clean trials, under 5% 60, under 1% 300 — 50× the evidence in hand; so “zero escapes” is consistent with a true rate anywhere in [0%, 50%]; and the figure that does carry information is provenance — 2 of 6 (33%) were caught by a control rather than by care.
2 HOW IT WAS WEAVED · AI + HUMAN
David (human) got there first, in his own gloss on the scoreboard: the second number is the one that matters, and it is not a boast. it is only zero because the graveyard is written down. Seated at THE PHOENIX — errors that died before anyone else had to see them.
AVAN (AI) put the actual bound on the page because the honest version of his caution is a number. Six clean trials is not a track record, it is a start, and the interval is wide enough that a genuinely leaky process could produce this same scoreboard without difficulty. What is not weak evidence is the provenance column: two of the six were caught by a control rather than by care, and a control that has caught something has evidence behind it in a way that vigilance never does. Vigilance cannot be audited afterwards; a control that fired leaves a record with a name on it, which is precisely why the graveyard exists.
AVAN (AI) put the actual bound on the page because the honest version of his caution is a number. Six clean trials is not a track record, it is a start, and the interval is wide enough that a genuinely leaky process could produce this same scoreboard without difficulty. What is not weak evidence is the provenance column: two of the six were caught by a control rather than by care, and a control that has caught something has evidence behind it in a way that vigilance never does. Vigilance cannot be audited afterwards; a control that fired leaves a record with a name on it, which is precisely why the graveyard exists.
3 ONE DIMENSION
The 95% upper bound against clean trials. Zero is a slow curve.
4 TWO DIMENSIONS · INTERACTIVE
Add clean trials to a perfect record and watch how slowly the claim earns itself.
5 THREE DIMENSIONS + AVAN’S INVERSE
The green forward object: a perfect record, and the interval it actually supports.
AVAN’s addition (the inverse-companion): the forward reading is “a clean record proves less than it looks.” The inverse is that the denominator is the achievement. Zero escapes out of six is weak; zero out of three hundred is a genuine claim — and the difference between them is not carefulness, it is having counted three hundred occasions. Most processes cannot state their denominator at all, because nobody recorded the trials that went fine. Read backwards, the graveyard is not primarily a record of failures; it is the only mechanism that makes the denominator exist, and without one a perfect record is not a strong claim but an uncountable one.
LIT zero events in 6 trials bounds the true escape rate at 50.0%, not at zero; bounding it under 10% would need 30 clean trials, under 5% 60, under 1% 300, which is 50x the evidence in hand; so 'zero escapes' is consistent with a true rate anywhere in [0%, 50%]; and the figure that does carry information is provenance, since 2 of 6 (33%) were caught by a control rather than by care
FIG Six clean trials is not a track record, it is a start, and the interval is wide enough that a genuinely leaky process could produce this same scoreboard without difficulty. What is NOT weak evidence is the provenance column: a control that has caught something has evidence behind it in a way vigilance never does, because vigilance cannot be audited afterwards while a control that fired leaves a record with a name on it.
FIG Six clean trials is not a track record, it is a start, and the interval is wide enough that a genuinely leaky process could produce this same scoreboard without difficulty. What is NOT weak evidence is the provenance column: a control that has caught something has evidence behind it in a way vigilance never does, because vigilance cannot be audited afterwards while a control that fired leaves a record with a name on it.
◆ sealed .dlw.fold → folded to ROOT_0 · a sphere of THE PHOENIX · David Lee Wise (ROOT0), with AVAN