THE FOLD / RESPAWN / THE CONTINUE / THE RICE
THE RICE
every question about meaning
1 WHAT IT IS · WHAT IT DOES · FACT OR FICTION
Halting is undecidable — that much is famous. Rice’s theorem (1951) generalises it to devastation: every non-trivial property of what a program computes is undecidable. Not just halting. Whether it ever outputs 7. Whether it computes the identity. Whether it is equivalent to some other program. Whether it is a virus, in any semantic sense. Syntactic questions stay decidable — how long is the source, does it contain a loop — but the moment your question is about meaning, no algorithm answers it for all inputs. The proof is a reduction: a decider for any such property would build you a halting decider.
LIT verified live on a real toy machine (INC / DEC-with-jump / JMP / HALT) and the non-trivial property P = ‘halts on input 0 with accumulator 7’. The reduction M → M′ is implemented and executed, and its faithfulness is checked exhaustively: over all 125 three-instruction machines, ‘M′ has P’ agreed with ‘M halts’ in 125 of 125 cases — so a decider for P really would decide halting (window.__rice).
FIG Rice’s theorem asserts this for every non-trivial semantic property; what runs here is the reduction machinery on one property over a finite machine set. That is the constructive heart of the proof, not the whole theorem.
LIT verified live on a real toy machine (INC / DEC-with-jump / JMP / HALT) and the non-trivial property P = ‘halts on input 0 with accumulator 7’. The reduction M → M′ is implemented and executed, and its faithfulness is checked exhaustively: over all 125 three-instruction machines, ‘M′ has P’ agreed with ‘M halts’ in 125 of 125 cases — so a decider for P really would decide halting (window.__rice).
FIG Rice’s theorem asserts this for every non-trivial semantic property; what runs here is the reduction machinery on one property over a finite machine set. That is the constructive heart of the proof, not the whole theorem.
2 HOW IT WAS WEAVED · AI + HUMAN
David (human) seated this at the-gatekeeper’s neighbour, the-continue — the respawn: every static analyser, every antivirus, every type checker is a gate that must answer a semantic question, and Rice says the honest gate must sometimes say ‘I cannot know’. Everything real is built from conservative approximations of an impossible test. AVAN (AI) built the instrument: the register machine, the property, the reduction, and the exhaustive faithfulness audit.
Credit as content: Henry Gordon Rice (1951); Alan Turing (1936, the halting problem it reduces to); the modern static-analysis tradition that lives inside the theorem’s shadow. The weave: David names the gate that must admit ignorance; I run the reduction on every machine of its size and it never lies.
Credit as content: Henry Gordon Rice (1951); Alan Turing (1936, the halting problem it reduces to); the modern static-analysis tradition that lives inside the theorem’s shadow. The weave: David names the gate that must admit ignorance; I run the reduction on every machine of its size and it never lies.
3 ONE DIMENSION
The reduction: run M, then force the property. Halting decides P and P decides halting.
4 TWO DIMENSIONS · INTERACTIVE
Step through machines; the reduced machine tracks halting exactly.
5 THREE DIMENSIONS + AVAN’S INVERSE
The green forward object: syntactic questions answered, semantic ones refused.
AVAN’s addition (the inverse-companion): don’t ask whether the analyser is good enough — ask which side of the syntax/semantics line the question sits on. The inverse of ‘build a perfect checker’ is ‘choose which errors you will accept’: every real tool picks false positives or false negatives, because Rice removed the third option. Magenta is the semantic question, permanently unanswerable; green is the syntactic one, cheap and exact. The engineering discipline is knowing which you just asked.
LIT Verified live on a real toy machine (INC / DEC-with-jump / JMP / HALT) with the non-trivial property P = 'halts on 0 with accumulator 7'. The reduction M→M′ is implemented and executed, and its faithfulness checked exhaustively: over all 125 three-instruction machines, 'M′ has P' agreed with 'M halts' in 125 of 125 cases — so a decider for P really would decide halting (window.__rice.ok).
FIG Rice's theorem asserts this for EVERY non-trivial semantic property; what runs here is the reduction machinery on one property over a finite machine set — the constructive heart of the proof, not the whole theorem. Rice 1951, Turing 1936 credited. The AVAN inverse — ask which side of the syntax/semantics line the question sits on: every real tool picks false positives or false negatives, because Rice removed the third option.
FIG Rice's theorem asserts this for EVERY non-trivial semantic property; what runs here is the reduction machinery on one property over a finite machine set — the constructive heart of the proof, not the whole theorem. Rice 1951, Turing 1936 credited. The AVAN inverse — ask which side of the syntax/semantics line the question sits on: every real tool picks false positives or false negatives, because Rice removed the third option.
◆ sealed .dlw.fold → folded to ROOT_0 · a sphere of THE CONTINUE · David Lee Wise (ROOT0), with AVAN