Agentic formal verification tool vstack finds 17 bugs in vLLM, fixes underway
AccBalanced · x · 2026-11-14
Hao AI Lab released vstack, a formal verification workflow and skill stack (TLA+, Z3, Apalache, TLAPS) for coding agents, and tested it on vLLM together with the vLLM team. Even in a mature codebase, bugs that only surface in rare combinations are hard for human reviewers to catch — vstack's agentic formal verification has found 17 (and growing) bugs, ranging from silent wrong outputs to server lockout. Five are published so far, each with a fix PR. vstack is open source (MIT) with Claude/Cursor plugins and a public bug tracker.
More from coding & agent
- Hermes AI Agent Night at AWS Singapore on Dec 16 to preview Hermes mobile app — Teknium · 2026-11-14
- Running AI Agents in Production: Reddit Compares 6 Platforms' Pricing Models — AcanthaceaeLatter684 · 2026-11-14
- Openship: self-hosted deployment platform with built-in CI/CD hits 15.3k GitHub stars — jackedAJ · 2026-11-14
- MCP-Guard 0.14.0 boosts tool-poisoning detection from 10/15 to 15/15, blind spots remain — ErgoniaWorks · 2026-11-14
- Money-saving trick: let a cheap model do the work, then ask it to summon a pricier model to review — vista8 · 2026-11-14
- Open-source personal agent The Machine drafts everything but never sends without your OK — markloperman · 2026-11-14