Field notes
The cycgraph blog.
Design notes, pattern deep-dives, and what we learn running multi-agent systems in production. Subscribe via RSS.
Will McMahan
We poisoned our agent's memory on purpose. It healed itself.
Compound learning lets agents get smarter every run — but it keeps bad lessons forever, too. Here's how we made workflows verify their own learning, and the five experiment iterations it took to get the gate right.
findingsmemoryagentsRead →
Will McMahan
Introducing cycgraph: orchestration that loops
Why we built an agentic orchestration engine on a Cyclic State Graph instead of yet another DAG — and what that unlocks for multi-agent systems.
announcementsarchitectureRead →