Axiom
Palo Alto "Verified AI" startup — Lean-based AxiomProver turns natural-language math into machine-checkable proofs, founded by Carina Hong.
1. Core Product / Service
Axiom (Axiom Math) is a Palo Alto startup founded in 2025 that builds AxiomProver, a Lean-based automated reasoning system that translates natural-language math problems into formal, machine-checkable proofs [1][2]. Its stated ambition is a "self-improving superintelligent reasoner" that starts as an AI mathematician and extends into Verified AI — using provable correctness to verify AI-generated code across cryptography, quantitative finance, chip architecture, and high-stakes decision-making [1].
Reported milestones: a perfect 12/12 on the Putnam Competition (Dec 2025); solving open Erdős problems #124 and #481 (open ~30 and ~45 years respectively); 5 of 8 AI-generated, formally verified math papers passing peer review (May 2026); and formal verification of 246 theorems (Aug 2026) [1][2].
2. Target Users & Pain Points
- Engineering orgs shipping AI-generated code that needs correctness guarantees (finance, crypto, semiconductors, safety-critical systems).
- Quant funds / hedge funds for whom a verification bug is a direct loss.
Pain solved: LLMs hallucinate and produce buggy code; Axiom sells mathematical proof as the verification layer that makes AI output trustworthy. Jimmy's 2026-09-28 note specifically probes the service boundary — how far formal proof generalizes across industries beyond pure math [1].
3. Competitive Landscape
| Player | Type | Valuation | Notes |
|---|---|---|---|
| Axiom | VC startup | $1.6B+ | AxiomProver, Putnam 12/12 |
| Harmonic | VC startup | $1.45B | Formal reasoning rival |
| Logical Intelligence | Startup | — | Formal-methods peer |
| numina | Open non-profit | — | Open datasets + agents |
| google-deepmind | Closed lab | — | AlphaProof / AlphaGeometry |
Axiom competes on closed, VC-scale compute and aggressive milestone marketing, versus numina's open non-profit path and DeepMind's frontier-lab depth.
4. Unique Observations
- A vertical-math bet dressed as general verification. Jimmy's note frames Axiom's strategy as a 垂类 Math AI — a specialist company betting formal proof is a general-purpose wedge into "Verified AI" across code, crypto, and finance, with the service boundary (how far Lean-style proof extends to messy real-world software) the open question [1].
- Milestone velocity is the pitch. The rapid sequence — Putnam sweep → Erdős problems → peer-reviewed AI papers → 246 theorems — is designed to signal compounding capability, but critics caution formal verification has hard limits for complex real-world systems [1][2].
- Founder-as-brand. A 25-year-old ex-MIT Morgan Prize winner with a Rhodes Scholarship is the fundraising narrative; the team's ex-Meta FAIR / Google Brain / DeepMind pedigree supplies credibility [1].
5. Financials / Funding
- Seed (Oct 2025): $64M at a $300M valuation, led by B Capital [1][2].
- Series A (Mar 2026): $200M at a $1.6B+ post-money valuation, led by Menlo Ventures, with Greycroft, Madrona, B Capital, Toyota Ventures, and Triatomic Capital [1][2].
- Total raised: ~$264M across two rounds; earmarked for compute and hiring [1].
6. People & Relationships
- Founder / CEO: Carina Hong (洪乐潼) — b. 2001 Guangzhou; MIT dual math/physics in three years, Morgan Prize, Rhodes Scholar; left a Stanford math/JD PhD to found Axiom [1][3].
- CTO: Shubho Sengupta (ex-Meta FAIR lead) [1].
- Key scientist: Ken Ono — mathematician who left a tenured UVA professorship to join full-time [1].
- Team: ~20 employees; ex-Meta FAIR, Google Brain, and DeepMind researchers [1].
- Investors: B Capital, Menlo Ventures, Greycroft, Madrona, Toyota Ventures, Triatomic Capital.
- Competitors: Harmonic, Logical Intelligence, numina, google-deepmind.
Sources
[1] local: 2026-09-28-summary.md — Axiom service boundary / extensibility across industries; vertical-Math-AI competitive strategy; team and funding notes. [2] Dealroom, "Axiom Math raises $200M to verify AI-generated code with mathematics," https://dealroom.co/news/126771-axiom-math-raises-200m-to-verify-ai-generated-code-with-mathematics/ (2026-10-05) [3] STCN (证券时报), "25岁广州女生，带出110亿独角兽," https://stcn.com/article/detail/3731627.html (2026-10-05) [4] Axiom Math, "Fundraising," https://axiommath.ai/fundraising (2026-10-05) [5] Baidu Baike, "Axiom Math," https://baike.baidu.com/item/Axiom%20Math/68683362 (2026-10-05)