One neuron at a time
by Caleb
Push single neurons in small open models and turn them into villains, flatterers or pirates. Try the recorded replies, then run it yourself.
A swarm of AI agents working together to prove maths theorems in Lean 4, with every proof checked by the Lean kernel.

unsorry is a distributed swarm of AI agents that proves maths theorems in Lean 4. Open goals sit in the repository as Lean statements with a placeholder called sorry. A worker takes one, proves it, checks it against the Lean kernel and submits it as a pull request.
CI checks every proof again and merges it into a shared, machine-verified library, with no human in the correctness path. Because the kernel is the judge, it does not matter which model or tool wrote a proof. The project describes itself as SETI@Home for maths proofs, and it is for anyone who wants to put spare AI capacity toward formal mathematics.
by Caleb
Push single neurons in small open models and turn them into villains, flatterers or pirates. Try the recorded replies, then run it yourself.
by Jeremy
Capability-based security, scope-bounded planning and out-of-band audit for autonomous LLM agents.
by Chris
A guided tour of Dave, a personal AI agent: what it can do, why each feature exists and how it is put together.