Skip to content

Porting loop at scale: baseline-reuse proofs, run-many, adopt; COCRDUPC recorded as not proven - #4085

Merged
squid-protocol merged 1 commit into
mainfrom
feat/loop-speedups
Oct 1, 2026
Merged

squid-protocol merged 1 commit into
mainfrom
feat/loop-speedups

Conversation

@squid-protocol

Copy link
Copy Markdown
Owner

Measured on CardDemo's online programs: per attempt, the model takes about 140 s, project generation 27 s, and a full proof 43 s.

  • Baseline reuse. While the model writes the first attempt, each loop proves the generated service itself. Every attempt then proves with --reuse against that run: 13 s instead of 45 for sign-on, and the whole loop drops from 221 s to 153 s. A port that replaces files other than the generated service is proven in full.
  • run-many runs many loops side by side, one process each, with staggered starts.
  • adopt copies a proven port into its case unchanged and writes provenance.json from the loop's records. It refuses a loop that did not prove.
  • Fix: uses_ports files are now merged before --reuse compares overlays. Before this, every reused proof of CORPT00C failed on the borrowed CSUTLDTC port.
  • README:
    • COCRDUPC is recorded as not proven, on purpose. Its update writes blanks into a PIC 9(3) CVV, which a typed Java field cannot hold.
    • COACTUPC's port is pending, because the program is too large for one model answer.
  • Docs: porting_loop.md, section "At scale", covering the new commands and drafting cases with agents.

Full suite: 13262 passed. audit_check: all clear.

🤖 Generated with Claude Code

https://claude.ai/code/session_01RZE6us7fUL6Hynk59rWLvZ

…COCRDUPC recorded as not proven

- Each loop proves the generated service while the model writes the
  first attempt; every attempt's proof then reuses that run's COBOL side
  and built project (equivalence.py --reuse): 13 s a proof instead of 45
  for CardDemo's sign-on. Measured per loop: the model ~140 s an attempt,
  generating the project 27 s, a full proof 43 s.
- run-many: loops side by side, one process each, starts staggered.
- adopt: a proven port into its case unchanged, with provenance.json from
  the loop's records; refuses a loop that did not prove.
- uses_ports files are merged before --reuse compares overlays (a
  borrowed CSUTLDTC port made every reused proof of CORPT00C fail).
- README: COCRDUPC (card update) is not proven on purpose -- its update
  writes blanks into a PIC 9(3) CVV, which a typed field cannot hold;
  COACTUPC's case is written, its port pending (too large for one answer).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RZE6us7fUL6Hynk59rWLvZ
@github-actions

github-actions Bot commented Oct 1, 2026

Copy link
Copy Markdown
Contributor

🐦‍⬛ Muninn Security Scan

✅ No security issues found.

🐦‍⬛ Powered by Muninn · Skald Lab

@squid-protocol
squid-protocol merged commit f5d1f92 into main Oct 1, 2026
30 checks passed
@squid-protocol
squid-protocol deleted the feat/loop-speedups branch October 1, 2026 18:17
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant