This is an interactive visualization that animates 695 API calls from a single Claude Code session as a physics-based simulation of chickens being summoned from a coop. Like the Assembly Line visualization, it is self-referential — the session it depicts is the session that built this visualization, version by version, across 47 conversational turns.
Click any chicken to see its details. The anatomy bar shows what percentage of the API call was guardrails ("Dogs"), previous AI output ("Chicken"), and user text ("You").
This accompanies a presentation called "Everything You Know About AI Is Wrong." AI models are chickens — specifically HeiHei from Moana. They're energetic, they'll peck at anything, but they have no idea what they're doing on their own. The software architecture is the sheepdog that keeps them moving toward the goal. The user is the chicken whisperer — setting direction, correcting course, and trusting the dogs to handle the rest.
A typical API call is roughly ~75% guardrails ("dogs"), ~24% previous AI output ("chickens"), and less than 1% user text ("you"). The straw chicken mechanic shows context compaction — when the chicken's brain gets full, everything gets squished into a summary, and some information is inevitably lost.