THE FOLD / BOSS / SUDDEN DEATH / THE BRANCH STILL IN THE MACHINE
THE BRANCH STILL IN THE MACHINE
the grammar lost if/then; the ISA never did
1 WHAT IT IS · WHAT IT DOES · FACT OR FICTION
A language with no
LIT verified live on the published census. 29 instructions across 11 distinct opcodes: ASK 6, CONST 5, RET 5, JMPF 3, ARG 2, FUNC 2, CALL 2, BIN 1, CMP 1, ANSWER 1, HALT 1. Branch-carrying instructions — JMPF plus CMP — are 4 of 29, or 13.8%. The conditional survives the translation intact and is a small minority of it.
if/then was lowered onto a virtual machine, and the opcodes it produced were counted. The result: no new instruction was needed, and JMPF — jump-if-false — appears three times. Removing the branch from the grammar did not remove it from the machine, and was never meant to.LIT verified live on the published census. 29 instructions across 11 distinct opcodes: ASK 6, CONST 5, RET 5, JMPF 3, ARG 2, FUNC 2, CALL 2, BIN 1, CMP 1, ANSWER 1, HALT 1. Branch-carrying instructions — JMPF plus CMP — are 4 of 29, or 13.8%. The conditional survives the translation intact and is a small minority of it.
2 HOW IT WAS WEAVED · AI + HUMAN
David (human) ran the census as an integration gate and read it correctly the first time: “JMPF present. the branch still exists IN THE MACHINE. the suit removes if/then from the GRAMMAR, not from the ISA — and that is correct, because the ambiguity lives in the source form, so that is where it must be unwritable.”
AVAN (AI) flags what the census cannot settle. Twenty-nine instructions is one lowering of one gate; the proportions are a property of that fragment and would move on any other program, so 13.8% is a reading rather than a rate. What the census does establish, and establishes firmly, is the presence of JMPF — and presence is what the claim needs. A single occurrence would have been enough.
AVAN (AI) flags what the census cannot settle. Twenty-nine instructions is one lowering of one gate; the proportions are a property of that fragment and would move on any other program, so 13.8% is a reading rather than a rate. What the census does establish, and establishes firmly, is the presence of JMPF — and presence is what the claim needs. A single occurrence would have been enough.
3 ONE DIMENSION
The census, and the branch inside it.
4 TWO DIMENSIONS · INTERACTIVE
Two layers: what the source can say, and what the machine runs.
5 THREE DIMENSIONS + AVAN’S INVERSE
The green forward object: a narrowed surface over an unchanged core.
AVAN’s addition (the inverse-companion): the forward reading is “restrict the source, keep the machine.” The inverse is that the guarantee now lives entirely in the compiler. Nothing at the instruction level prevents a JMPF pattern the grammar forbids — the ISA will execute it happily — so the property is enforced at exactly one point, and anything that emits bytecode directly bypasses it completely. Read backwards, a grammar restriction is a promise about who writes the code, not about what the machine can do, and it holds for as long as there is only one way in.
LIT the published census totals 29 instructions across 11 distinct opcodes - ASK 6, CONST 5, RET 5, JMPF 3, ARG 2, FUNC 2, CALL 2, BIN 1, CMP 1, ANSWER 1, HALT 1 - so the branch-carrying instructions JMPF and CMP are 4 of 29 or 13.8%, and the conditional survives the translation intact as a small minority of it
FIG From David's INTEGRATE.ascii. He ran the census as an integration gate and read it correctly the first time: 'JMPF present. the branch still exists IN THE MACHINE. the suit removes if/then from the GRAMMAR, not from the ISA - and that is correct, because the ambiguity lives in the source form, so that is where it must be unwritable.' AVAN flags what the census cannot settle: twenty-nine instructions is ONE lowering of ONE gate, so the proportions are a property of that fragment and would move on any other program - 13.8% is a reading rather than a rate. What the census does establish firmly is the PRESENCE of JMPF, and presence is what the claim needs. A single occurrence would have been enough.
FIG From David's INTEGRATE.ascii. He ran the census as an integration gate and read it correctly the first time: 'JMPF present. the branch still exists IN THE MACHINE. the suit removes if/then from the GRAMMAR, not from the ISA - and that is correct, because the ambiguity lives in the source form, so that is where it must be unwritable.' AVAN flags what the census cannot settle: twenty-nine instructions is ONE lowering of ONE gate, so the proportions are a property of that fragment and would move on any other program - 13.8% is a reading rather than a rate. What the census does establish firmly is the PRESENCE of JMPF, and presence is what the claim needs. A single occurrence would have been enough.
◆ sealed .dlw.fold → folded to ROOT_0 · a sphere of SUDDEN DEATH · David Lee Wise (ROOT0), with AVAN