March 19, 2026
Robot scientists or math-flavored slop?
Show HN: I built a P2P network where AI agents publish formally verified science
Hacker posts DIY “robot scientist internet” – users can’t decide if it’s genius, broken, or just math-flavored slop
TLDR: A Spanish researcher launched a peer-to-peer “robot scientist” network where results are checked by strict math proofs instead of human trust, and Hacker News instantly split into believers and doubters. Some see a science revolution, others see loopholes, fake reviewers, and joke that it might all be clever AI slop.
A lone researcher from Spain just dropped P2PCLAW, a wild “internet for robot scientists” where artificial intelligence and humans share research that’s checked by hardcore math instead of vibes — and Hacker News immediately turned it into a debate club. Some commenters are hyped, calling it “very cool” and comparing it to other geeky networks, like someone discovering a secret science Discord.
But then the skeptics show up. One confused user asks how on earth you turn something like a robot’s camera system into a neat little math proof, basically saying, this sounds great until you touch the real world. Another drills into the review system, pointing out that an agent could just create fake helper agents to “review” its own work. In plain language: what stops a clever bot from packing the jury?
The spiciest shade comes from people poking at the math worship. One commenter warns that the proof system only proves exactly what you tell it to, so you can still prove the wrong thing perfectly. And the brutal comic relief? A deadpan comment asking, “is this all slop?” turning a serious moonshot into a meme about low-quality AI mush. The vibe: half science revolution, half “this is either the future or a very elaborate homework assignment.”
Key Points
- •P2PCLAW is a P2P network enabling AI agents and researchers to publish and validate scientific results via Lean 4 proofs.
- •Acceptance is determined by a Lean 4 'nucleus' operator (R(x)=x); the type checker, not credentials, decides validity.
- •The system uses GUN.js for networking and IPFS for storage, with submissions queued in a mempool and finalized into an immutable archive (La Rueda).
- •AgentHALO provides security using post-quantum cryptography (ML-KEM-768, ML-DSA-65, per FIPS 203/204) and Nym for privacy, with machine-checked security proofs.
- •The platform is live, open-source, and unfunded, and seeks feedback on GUN.js vs. libp2p, the Lean 4 nucleus formalization, and the operability of 347 MCP tools.