Formal methods veteran: without formal semantics, a Lean proof can't make code 'correct'
JiaweiLiu_ · x · 2026-09-20
Formal verification expert RosuGrigore warns the new wave of AI-era computer scientists who believe "I proved it in Lean, so it's correct": you cannot even define code correctness without a formal semantics of the programming language.
His simple demonstration: run the same program under gcc and clang at various optimization levels and watch the results diverge before attempting a Lean proof. The formal semantics is the hard part, requiring both human judgement and an appropriate language framework — orthogonal to Lean, which only works after the semantics exists.
A pointed technical corrective to the current hype around AI-assisted Lean proofs.
More from Research
- Fields Medalist Villani on OpenAI's Millennium Problem: 'A Cataclysm Like Math Has Never Known' — GregCook2011 · 2026-09-20
- lateinteraction: Papers are just timestamped formats — the real projects live elsewhere — lateinteraction · 2026-09-20
- Two high school students, aided by AI, report progress on a problem studied by Fields medalist June Huh — IgorCarron · 2026-09-20
- GPU Programming Diary: Revisiting the Classic CUDA Matmul Optimization Worklog and MIT's Sparsity Lecture — NandoDF · 2026-09-20
- New paper finds 'anti-grokking': test accuracy collapses back to chance after successful generalization — sytelus · 2026-09-20
- With 50k+ ICLR submissions, researchers propose hybrid human-LLM peer review orchestration — HamedSHassani · 2026-09-20