NeurIPS Oral: IDS co-evolves code with formal proofs, hitting 3x Claude Code's success rate
adityagp · x · 2026-09-25
- The IDS (Inductive Deductive Synthesis) paper was accepted as a NeurIPS 2026 Oral (reportedly top 0.34%), with paper and code open-sourced.
- Instead of generating code and praying tests pass, IDS starts from a formal Lean/Rocq spec and incrementally co-synthesizes proofs alongside code, letting correctness constraints guide generation.
- On distributed systems specs, IDS achieves roughly 3x the success rate of Claude Code, beating SOTA coding agents.
- The team also points to the next challenge: ensuring the spec itself captures human intent.
Related event: IDS Paper on Proof-Code Co-Synthesis Named NeurIPS Oral(3 posts)→
More from coding & agent
- Inside Quail: custom vLLM scheduler, workload-aware KV cache for 1B tok/min — sh_reya · 2026-09-25
- Anthropic's CI job volume grew 25x in six months — here's how they scaled test selection — JeremyCMorgan · 2026-09-25
- Perplexity's Fast Search powers Hermes Agent with 160ms p50 latency, free for all tiers — denisyarats · 2026-09-25
- Cursor launches Projects: one coordinator directing thousands of subagents, heavy users merge 6x more PRs — gaganghotra_ · 2026-09-25
- Paperclip lets any agent harness (Claude Code, Codex, Grok) work as employees in an AI company — AIFlow_ML · 2026-09-25
- Zapier CEO Wade Foster on grading AI fluency — and why the top rating is so rare — aakashgupta · 2026-09-25