From the Anthropic edition of November 1, 2026
Anthropic publishes Claude's claimed 2^o(n) 3-SAT algorithm refuting ETH with an AI-written Lean proof, noting no machine checks that the formal statement matches the paper
Anthropic · AI model provider; research · Global
- Publisher
- GitHub: anthropics/formal-math
- Original publication
- Undated repository README (claim announced late October 2026, per Papex)
- Source retrieved
- 2026-11-02
What happened
Anthropic's repository states that an internal research Claude model found a deterministic 2^o(n) algorithm for 3-SAT, refuting the Exponential Time Hypothesis, with a paper credited to Claude and a Lean 4 formalization written by AI models. It says the result is purely asymptotic with no practical effect, and that Lean checks the proofs but no machine checks that the formal statement matches the paper.
Why it matters
This is the primary vendor source for a research claim that universities, CS faculty and research offices are already discussing and citing (cross-tagged research). It is useful for showing how a vendor states verification limits alongside a headline claim. Why now: not archived; the independent examination reported Oct. 31 makes the vendor's own stated limits relevant to how campuses discuss the result.
Evidence and measured results
Vendor claim with a public formal proof artifact; independent kernel replays reported by third parties; no expert verdict.
Limitations and uncertainty
Vendor claim. The README is undated. The result has no practical algorithmic effect by Anthropic's own account.
Put this evidence to work
Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-11-02; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.
Sales
Role takeaway
- Customer problem
- Research customers hear AI discovery claims and need to separate vendor claims from verified results.
- Stakeholders
- CS faculty
- Research computing director
- Research-integrity officer
- Discovery questions
- How would your institution evaluate an AI-produced research claim?
- Who can review formal statements?
- Do authorship policies address AI-credited work?
- Value hypothesis
- Interpretation: positioning around verification practice is more durable than positioning around headline claims.
- Potential engagement
- A research-AI verification workflow review using this repository as a worked example.
- Unsupported claims
- That ETH is refuted
- That commercial Claude models produce comparable results routinely
- Any practical speedup
Pre-sales engineering
Role takeaway
- Fit
- Relevant to research computing and formal-methods groups.
- Architecture and integration
- Pinned Lean toolchain with Comparator and independent-kernel checks in CI.
- Prerequisites
- Lean toolchain
- About 24 vCPUs and 7 GB memory for builds per README
- Constraints
- No machine check of statement-to-paper correspondence
- Repository not maintained
- Security
- No direct security concern.
- Proof of value
- Reproduce the build and axiom check
- Have a faculty member review the statement file against the paper
Delivery
Role takeaway
- Work
- Use the artifact as a reference design for AI-assisted formal research workflows.
- Dependencies
- Research computing capacity
- Faculty reviewer
- Ownership
- Research computing director with CS faculty.
- Skills and adoption
- Workshop on formal verification limits.
- Governance checkpoints
- Faculty sign-off on statement-review checklist
- Acceptance criteria
- Proposal: replay documented
- Proposal: checklist adopted
- Risks
- Result later fails expert review
- Over-generalizing from one claim
Implementation considerations
Lighthouse Advisory interpretation across the operating dimensions a public-sector buyer must settle before this evidence becomes a design. Each note answers the question under its heading for this specific source.
Architecture and integration
What must connect, and where does the AI sit in the workflow?
Shows a reproducible pattern for AI-assisted formal work: pinned toolchain, multi-kernel checking and a short human-readable statement file.
Governance
Who approves, reviews and stays accountable for outcomes?
Authorship credited to an AI model raises questions for institutional authorship and integrity policies.
Security and privacy
What data, permissions and controls need testing?
No direct data-security effect.
Accessibility and workforce
Who is affected, and what skills or accommodations follow?
Limited direct effect; relevant to teaching about formal verification.
Procurement
What should contracts, pricing and exit terms secure?
Discovery claims should not drive research-AI purchasing without independent evidence.
Operating model
Which teams own the service once it runs?
Research offices may need a stance on AI-credited authorship.
Publication history
- 2026-11-01Anthropic · Issue 048 resources
Stable resource ID: anthropic-formal-math-eth-refutation-claude-lean-readme-202610