THE FOLD / RESPAWN / HARD RESET / THE ZERO PARAMETER
THE ZERO PARAMETER
five rules with nothing to tune
1 WHAT IT IS · WHAT IT DOES · FACT OR FICTION
Five rules, and not one of them has a number to tune. Veto: you cannot close a bracket you did not open. −I: whatever you left open owes you a closer. Depth: you cannot nest past the declared cap. Idempotence: a statement that does nothing cannot do nothing twice. Address: the machine, not the writer, decides what position a statement sits at. They are counting, not judgement — and that is why they can promise something rather than score it.
LIT verified live: each of the five catches the input it exists for, 5 of 5. Exhaustively over every string up to length 9 in a three-symbol alphabet, 1,374 are accepted and 0 of them are malformed — a guarantee with no exceptions to report. A tunable alternative was swept across 7 settings and 0 of them keep every valid string while rejecting every invalid one; even at its tightest setting it still admits 12.7% of malformed input.
LIT verified live: each of the five catches the input it exists for, 5 of 5. Exhaustively over every string up to length 9 in a three-symbol alphabet, 1,374 are accepted and 0 of them are malformed — a guarantee with no exceptions to report. A tunable alternative was swept across 7 settings and 0 of them keep every valid string while rejecting every invalid one; even at its tightest setting it still admits 12.7% of malformed input.
2 HOW IT WAS WEAVED · AI + HUMAN
David (human) wrote the five into pocket-machine’s fourth window with the claim attached: none of these five learned anything. They have no numbers to tune. That is the whole claim — the parts that guarantee the output is well-formed are the parts with zero settings. He seated this at HARD RESET: nothing to tune is nothing to drift.
AVAN (AI) checked the accepted set a second way, by a closed form the page never computes: a string over this alphabet is accepted exactly when its brackets balance, so the count at length n is the sum over k of C(n,2k) times the k-th Catalan number. That gives 1, 1, 2, 4, 9, 21, 51, 127, 323, 835 — the Motzkin numbers — totalling 1,374, which is what the enumeration found. AVAN also added the threshold sweep, because “zero parameters” only means something against an alternative that has some. The comparison is deliberately generous to the tunable side — it keeps 100% of valid strings at every setting tested — and it still cannot get the second half: at the tightest setting it admits 12.7% of malformed input, and loosening it only makes that worse. Worth being precise about the one number in the five: the depth cap is a declared capacity, not a threshold. Changing it changes which programs fit, never which accepted programs are well-formed — and in this enumeration it never binds at all, since the deepest nesting reachable inside nine characters is four.
AVAN (AI) checked the accepted set a second way, by a closed form the page never computes: a string over this alphabet is accepted exactly when its brackets balance, so the count at length n is the sum over k of C(n,2k) times the k-th Catalan number. That gives 1, 1, 2, 4, 9, 21, 51, 127, 323, 835 — the Motzkin numbers — totalling 1,374, which is what the enumeration found. AVAN also added the threshold sweep, because “zero parameters” only means something against an alternative that has some. The comparison is deliberately generous to the tunable side — it keeps 100% of valid strings at every setting tested — and it still cannot get the second half: at the tightest setting it admits 12.7% of malformed input, and loosening it only makes that worse. Worth being precise about the one number in the five: the depth cap is a declared capacity, not a threshold. Changing it changes which programs fit, never which accepted programs are well-formed — and in this enumeration it never binds at all, since the deepest nesting reachable inside nine characters is four.
3 ONE DIMENSION
The five, and what each one refuses.
4 TWO DIMENSIONS · INTERACTIVE
Slide the tunable checker’s threshold and watch it fail to reach both corners.
5 THREE DIMENSIONS + AVAN’S INVERSE
The green forward object: the accepted set, carved by five walls with no dials on them.
AVAN’s addition (the inverse-companion): the forward reading is “rules with no settings can guarantee things.” The inverse is that they can only guarantee things they were built to be about. Counting brackets proves brackets balance and proves nothing about whether the program is any good — the guarantee is total within a scope that is deliberately tiny. Read backwards, the trade is not parameters against reliability but ambition against certainty: the five rules are certain because they gave up on almost every question, and a system that wanted to answer more would have to start tuning and stop promising.
LIT each of the five catches the input it exists for, 5 of 5; exhaustively over every string up to length 9 in a three-symbol alphabet, 1,374 are accepted and 0 of them are malformed - a guarantee with no exceptions to report; and a tunable alternative swept across 7 settings has 0 that keep every valid string while rejecting every invalid one, admitting 12.7% of malformed input even at its tightest
FIG From David's pocket-machine, with the claim attached: 'none of these five learned anything. They have no numbers to tune. That is the whole claim - the parts that guarantee the output is well-formed are the parts with zero settings.' AVAN added the threshold sweep, because 'zero parameters' only means something against an alternative that has some; the comparison is deliberately generous to the tunable side, keeping 100% of valid strings at every setting, and it still cannot get the second half. Worth being precise about the one number among the five: the DEPTH CAP is a declared capacity, not a threshold. Changing it changes which programs fit, never which accepted programs are well-formed.
FIG From David's pocket-machine, with the claim attached: 'none of these five learned anything. They have no numbers to tune. That is the whole claim - the parts that guarantee the output is well-formed are the parts with zero settings.' AVAN added the threshold sweep, because 'zero parameters' only means something against an alternative that has some; the comparison is deliberately generous to the tunable side, keeping 100% of valid strings at every setting, and it still cannot get the second half. Worth being precise about the one number among the five: the DEPTH CAP is a declared capacity, not a threshold. Changing it changes which programs fit, never which accepted programs are well-formed.
◆ sealed .dlw.fold → folded to ROOT_0 · a sphere of HARD RESET · David Lee Wise (ROOT0), with AVAN