Ex-Moonshot AI researcher startup uses AI to automate chip formal verification
teortaxesTex · x · 2026-11-13
Per LeiFang report, Zhang Jujian (former head of formal verification at Axiom Math) and Liu Junqi (formerly AI math research at Moonshot AI) founded Beijing Rational Numbers Technology to automate formal chip verification proofs with AI, and completed a seed round.
- Problem: Chips with logic errors discovered post-fabrication are extremely costly to redo; traditional formal tools like Cadence JasperGold and Synopsys VC Formal require heavy manual setup and can stall on complex circuits.
- Team: Zhang contributed to Mathlib and AxiomProver; Liu is a core author of Numina-Lean-Agent, which proved all 12 problems of the formalized 2025 Putnam competition with Lean-checked proofs.
- Progress: The prototype verified rules in projects with hundreds of thousands of lines of RTL that existing EDA tools could not determine; the team is working with chip customers toward a commercial verification service.
More from Venture
- Polymarket prices 31% chance Jensen Huang is reinstated as Nvidia CEO by end of next year — Polymarket · 2026-11-13
- Mobile app sells for $1.3K at 0.6x multiple, closing TrustMRR's 218th acquisition in 102 days — marclou · 2026-11-13
- Nscale admits no binding commitments for $32B Anthropic megaproject as losses hit $4 per $1 revenue — GaryMarcus · 2026-11-13
- Building a product with Opus 5.5 cost $4,000 in API fees — and it's unprofitable — marclou · 2026-11-13
- Agent Cards demo shows AI agents spending money in Airwallex sandbox — kimmonismus · 2026-11-13
- Back-of-envelope: serving 1B AI Mode queries on Gemini costs only ~$4M — gaganghotra_ · 2026-11-13