srush: formal verification is the only plausible path to serious LLM-written code
srush_nlp · x · 2026-10-28
Princeton's Sasha Rush publishes the third post in his Lean for ML series, arguing formal verification is the only plausible way to write serious code with LLMs — while noting that in 2026 formal verification remains so poor that understanding proofs is beyond human ability.
Alongside it he released Lean Verified Tokenizers: training-inference mismatch kills RL runs, yet agent message templates are still assembled with Jinja. The project prototypes a Lean-verified Rust agent pipeline to make message construction provably correct.
Related event: srush pushes Lean-verified tokenizers and formal methods for LLMs(2 posts)→
More from coding & agent
- UCLA and Tsinghua paper probes how far coding agents can go in visual generation — ucla-cmllab · 2026-10-28
- 'Claude Duet has won': devs say use provider-agnostic harnesses as the top model keeps rotating — p_e_cooper · 2026-10-28
- ArkenOS deploys swarm of autonomous scientists; Sam Altman approves — _AustinCalvert_ · 2026-10-28
- Lucas Meijer demos cloud agent workspace running parallel Opus tasks at AI Tinkers meetup — lucasmeijer · 2026-10-28
- Stanford's 3-year medical AI benchmark review: GPT-5.5 drops to 28% when tasks must succeed 3 runs in a row — jonc101x · 2026-10-28
- Chatbots say things, agents do things: a clinician's case for medical AI agents — jonc101x · 2026-10-28