Typed-entry baseline continuity — 2026-09-12

Actual native program hashes match for all84 unique source/target pairs in the 133 baseline fixtures when compiled with the retained original proof binary, FINAL3 installed binary and the current compiler-API3 Trisha binary. This proves compiler-output continuity for the compared input snapshots. Historical proof receipts have a separate input-binding limitation described below.

Compared implementations and input snapshots

• Original retained proof binary /tmp/trisha-v7-proof-binary/trisha, SHA-256 c73e8a58c848769abbc7f259c232ad1ae7d0bcc477872a3cffe55e06c568dede. This identity matches the actual remaining37 proof-run log and receipt.
• FINAL3 installed Darwin binary from /tmp/cyber-release-installed-v7-darwin-final3/bin/trisha, compiling the exact archived source tree under /tmp/cyber-release-extracted-v7-final3/cyber-source.
• Current /Users/master/cyber/trisha/target/debug/trisha, built successfully with compiler API3; compiling current source inputs with release source profile.

For each of84 source/target pairs, the temporary probe parses generated assembly with actual Triton Program::from_code and computes Program.hash(). Every native hash matches across all three implementations. FINAL3 versus current TASM is byte-identical in47 cases; the remaining37 have only whitespace differences, with identical token sequences. All266 files in baselines/triton (fixture manifests, source fixtures, hand assembly and witness files) are byte-identical between FINAL3's archive and the current tree. Actual fixture count133 and baseline count43 remain unchanged. This task generated no new proofs.

Machine-readable comparison, including per-pair source, target, native hash and assembly SHA-256: /tmp/final4-baseline-assembly-identity.json. Driver /tmp/final4-proof-binary-continuity.py, result /tmp/final4-proof-binary-continuity.log. Original and new assembly files are retained under /tmp/final4-baseline-assembly/.

Historical receipt limitation

The original remaining37 command used /tmp/trisha-v7-remaining-fixtures, whose fixture directories are symlinks into the live source tree. Its guard recorded the binary SHA, command, cycles, successful proof verification and log identity, but did not record hashes of source, fixture manifests, witnesses or native program claims. That selection is not an immutable proof-time input snapshot. The six separate recursive/native-currency/transaction full-proof logs invoked the then-live target/release/trisha and did not record its executable SHA.

All eight log hashes in audit/full-baseline-proof-coverage.json still match their retained files. The logs establish198 successful fresh proof verifications over99 positive fixtures and the documented negative executions. However, the preserved artifacts do not independently bind every proof-time input byte to today's fixture bytes. Recompiling today's inputs with the retained proof binary cannot retroactively add that missing binding. Therefore this report does not describe all198 historical statements as cryptographically bound to the new release.

Current FINAL3 fixture identity and current/original compiler output identity are verified facts; full historical statement continuity is the unsupported stronger claim. A new complete proof run from a frozen inventory with per-fixture statement, program and witness hashes would close that evidence gap. Existing full release smokes and typed-entry proof tests have their separate receipts and scope.