Formal Verification Veteran Slams AI Circle: Garbage Semantics In, Garbage Proofs Out
tianyin_xu · x · 2026-09-20
Grigore Rosu, a formal verification expert, observes that his field—dating back to the 1960s—keeps getting "rediscovered" by hotter domains, first blockchain and now AI, each time accompanied by newcomers' ignorance and arrogance.
He skewers a recurring pattern: newcomers claim Lean-based formal verification guarantees correctness, yet can't answer basic questions like "what's your trust base?" or "does your language even have a formal semantics?" His verdict: garbage (formal semantics) in, garbage (formal verification) out—Lean only guarantees code matches the semantics you wrote, not that the semantics is right.
He recommends Runtime Verification's blog for doing FV properly, which covers AI-assisted fuzzing that found a real bug in the WAMR WebAssembly runtime, a Rust-rewrite-and-verify workflow for Linux C code with Lean proofs, and end-to-end Lean 4 machine-checked verification of Android's Binder driver deserializer.
More from Research
- Gated Recurrent Transformers: 3-layer recurrent model matches 12-layer GPT-2 with 63% fewer params — burny_tech · 2026-09-20
- Stanford builds virtual biotech run by 37,000 AI scientist agents, with real drug discovery results — Dr_Singularity · 2026-09-20
- A year after questioning LLM conjecture-solving, Cornell professor concedes AI surged ahead — burny_tech · 2026-09-20
- VLA-Replica: a $-low-cost real-world VLA benchmark where NVIDIA GR00T N1.7 matches π₀ with 50 demos — YuXiang_IRVL · 2026-09-20
- OpenAI is close to solving the Hodge Conjecture, a Millennium Prize Problem, The Information reports — burny_tech · 2026-09-20
- New Paper: GPU Congestion Can Reverse the SFT vs ICL Personalization Tradeoff — zainhas · 2026-09-20