AlphaProof Code: AI agents write proofs, Lean checker gates them — verifying Signal at machine-code level
rohanpaul_ai · x · 2026-11-10
The diagram shows the "AlphaProof Code" workflow for verifying Signal's security at the machine-code level: you never have to trust the AI.
- Specification: a spec-writer agent turns models and properties from the literature into a Lean spec with the user; reviewer agents critique drafts, user approves.
- Code: unmodified Rust (libsignal, spqr) is compiled with the standard compiler; the AArch64 executable is loaded into Lean with AArch64 semantics, making it executable in Lean.
- Proof: a prover team (orchestrator + workers) proves the spec about the executable; prover-proposed amendments need policy or explicit user approval. A gatekeeper accepts a proof only if Lean's kernel checks it and the Lean comparator confirms it proves exactly a held-out copy of the spec.
What you trust: human-approved security rules, Lean's kernel, cryptographic primitive assumptions, and some routine contracts — everything the AI produces is filtered by a strict math checker.
More from coding & agent
- Emulating Humanlike Mouse Moves and Key Timing to Make Agents Undetectable — Darpinian · 2026-11-12
- Stanford NLP course's best homework: build your own agent harness — stanfordnlp · 2026-11-12
- Anthropic details agent memory: dreaming organizes Claude's memory, wiki+search hits 61% — RLanceMartin · 2026-11-12
- Talk preview: scaling 1,000 Fable 5.2 agents to formalize Fermat's Last Theorem — RLanceMartin · 2026-11-12
- Why Real Consumer PCs Are the Only Unblockable Way to Run Computer Use Agents — Darpinian · 2026-11-12
- Opinion: AI infra is becoming a coding runtime as intelligence units outgrow tokens — prateekj · 2026-11-12