TLA-Verus uses LLM agents to prove temporal properties of TLA+ specs
tianyin_xu · x · 2026-11-02
A NeurIPS'26 workshop paper, TLA-Verus, takes TLA+ specs from GitHub, translates them into Verus, and has LLM agents prove the safety and liveness properties. Motivation: TLA+ specs drift from code, and TLC only checks small configurations while TLAPS is weak on liveness.
Blogger Murat Demirbas pushes back:
- The paper drops the spec-vs-code problem at the outset: no refinement proof from TLA+ to the generated Rust code (which it doesn't even produce), and none between the TLA+ and Verus specs either.
- It understates TLAPS: BY PTL routes steps to the LS4 LTL decision procedure, coalescing abstracts first-order obligations into propositional atoms, and Lamport's temporal rules (including WF1) all exist. The real limitation is poor reasoning about ENABLED.
More from coding & agent
- As Grok bot limits tighten, open-source self-hosted agent open-duet hits 6.5k GitHub stars — matchaman11 · 2026-11-02
- Solo Dev Codes Entire Fishing Game With Opus in 3 Months, Demo Out Free — RUSuper · 2026-11-02
- Honeypot Vulnerable MCP Server Logs 3,044 Requests in 7 Days, No Real Attacks Yet — 1337Captain · 2026-11-02
- What Actually Broke When We Put Agents on Real Back-Office Work — TapNorth0888 · 2026-11-02
- Compacting earlier cuts agent costs 16-23%; GLM 5.3 costs 30x less than Grok 4.7 — yb2698 · 2026-11-02
- Codex capacity errors spike whenever OpenAI hands out banked resets — CtrlAltDwayne · 2026-11-02