AI revives formal verification: seL4 took 20 person-years, startups now claim it's practical
lemire · x · 2026-10-30
lemire revisits the history of formal verification — from Dijkstra's 1970 line that "testing shows the presence of bugs, never their absence" — and why it failed: seL4's 8,700 lines of C required 200,000 lines of Isabelle proofs and roughly 20 person-years of effort.
Now a wave of startups and academic initiatives claims AI makes formal verification practical: have AI prove your C code correctly implements its spec, including memory-usage properties. The most obvious use cases are cryptography and anything that must be unbreakable — proving a system cannot be hacked.
He cautions these remain mathematical models full of assumptions, but is personally excited and plans to formally verify heuristics in his own projects. Examples cited: Truth (launched Oct 2026, $70M seed, building a "universal verification machine" for any language) and Harmonic (Aristotle), which announced C software and Verilog hardware verification support in late Oct 2026.
More from coding & agent
- Data leakage hit 648 papers across 30 fields; autoarxiv now auto-reproduces arXiv papers — pandeyparul · 2026-10-30
- METR trial: AI made experienced devs 19% slower while they felt 20% faster — srimisra · 2026-10-30
- Clairvoyance: free desktop tool wires Claude Code, Codex and other agents into your daily workflow — draginol · 2026-10-30
- Same Prompt, Same Data: Claude's Fabric App Beats Its Dashboard at One-Fifth the Cost — adnan_hashmi · 2026-10-30
- Antigravity's Google Slides Plugin Audits an 80-Slide Deck Against Docs — rseroter · 2026-10-30
- OpenAI hosts Cloud Native Hackathon Nov 9; hack turns M5Stack into a k8s desk pager — OpenAIDevs · 2026-10-30