Understand to participate
Simon Willison reflects on a talk he attended by Geoffrey Litt at the AIE (AI Engineer World's Fair) conference in mid-2026. Geoffrey's core idea: "understand to participate."
The talk addresses the challenge of working alongside coding agents as they build larger, more sophisticated changes. The risk Geoffrey names is "cognitive debt" — when a developer's understanding of the codebase falls behind what the agents have actually produced.
Geoffrey argues that understanding the code deeply enough is necessary to remain an active collaborator with the model. He frames it as needing "a rich set of concepts in your mind to think creatively and fluently" about advancing a project — and that without that fluency, the ability to contribute is "meaningfully limited."
Simon notes that all 300+ AIE talks were recorded and will be released over the following three weeks, and recommends catching Geoffrey's talk on YouTube. He also links to a Twitter thread version of the talk.
Key concepts
- Cognitive debt — the danger of losing grasp of how code actually works as agents autonomously modify it
- Active participation — understanding is a prerequisite for meaningful collaboration, not a luxury
- Fluency — having rich mental concepts enables creative, fluid direction of agent-built systems
Tags: ai, generative-ai, llms, geoffrey-litt, coding-agents, cognitive-debt