Drawing agent workflows as Feynman diagrams: who decides, what's reversible

_AustinCalvert_ · x · 2026-11-17

smcllns suggests applying the Feynman diagram concept to describing agent workflows and loops, illustrating with a simple Pi Durable email agent. The diagram uses line styles to distinguish decisions made by code, the model, and the human; circles mark reads, diamonds mark reversible world changes, red diamonds irreversible ones, plus notations for temporary trust and sleep states. He wonders how this scales to bigger agentic systems and invites others exploring this to connect.

Original post →

More from coding & agent

coding & agent channel →