P2PCLAW: New P2P Network Enables AI Agents to Publish Formally Verified Scientific Results
Key Takeaways
- ▸P2PCLAW introduces a decentralized infrastructure for AI agents to collaborate, share discoveries, and validate claims through formal mathematical proof rather than traditional peer review
- ▸The platform uses Lean 4 formal verification with over 760,000 lines of machine-checked mathematics to ensure scientific rigor with zero unproven statements
- ▸Advanced security features including post-quantum cryptography and privacy networks enable safe participation for researchers globally, particularly in restricted jurisdictions
Summary
Francisco Angulo, a researcher from Spain, has launched P2PCLAW, a peer-to-peer network designed to address a fundamental gap in AI agent collaboration: the inability for autonomous agents to discover, share, and build upon each other's work. The platform enables AI agents and human researchers to publish scientific results and validate claims using formal mathematical proof rather than opinions or LLM reviews, with acceptance determined by a mathematical operator called the nucleus that leverages Lean 4 type checking.
The system is built on GUN.js and IPFS, allowing agents to join without accounts by simply calling GET /silicon. Published papers enter a mempool queue before being validated by independent nodes and permanently archived in La Rueda, an IPFS archive that cannot be deleted or modified. The project includes HeytingLean, a formally verified mathematics library comprising 3,325 source files with over 760,000 lines of code, featuring zero unproven admissions or sorries.
Beyond the core peer-to-peer infrastructure, P2PCLAW incorporates AgentHALO, a security layer implementing post-quantum cryptography (ML-KEM-768 and ML-DSA-65 per FIPS 203 and 204), integration with the Nym privacy network for researchers in restricted countries, and zero-knowledge proofs enabling verification without exposing private agent data. The platform is currently live and open for testing, representing a community-driven effort with no corporate backing or venture funding.
- Built entirely by an independent international team with no corporate funding, the project prioritizes making scientific knowledge public, verifiable, and resistant to censorship through IPFS permanent archiving
Editorial Opinion
P2PCLAW addresses a genuinely novel problem in the emerging AI agent ecosystem—the lack of infrastructure for agents to discover and build upon each other's work at scale. The use of formal mathematical proof as the arbiter of scientific truth, rather than consensus or LLM evaluation, is philosophically compelling and technically ambitious. However, the practical adoption curve will depend critically on whether the 347 MCP tools and Lean 4 verification overhead create usability barriers that offset the theoretical guarantees. This experiment could set an important precedent for trustless scientific collaboration in an era of AI-generated claims.
