Chapter 11 - The Full Sensorimotor Loop
Last updated 2026-07-13. This page hosts the extended worked example for Chapter 11 - one complete task walkthrough from user intent to verified result - plus the source documents the chapter's sequence diagram condenses.
Companion activity
Trace one task end to end through the full loop, naming the handoff at each transition and the check that would catch a failure there.
What this chapter argues
Parts I and II treated the regions one at a time. Chapter 11 puts them together: a single task, traced through dialog interface to working memory to prefrontal cortex to motor / sensory to long-term memory and back. The chapter argues that the discipline that matters most is at the handoffs - the transitions between layers - not within any single layer. A well-disciplined transition makes a mediocre layer adequate. A sloppy transition makes an excellent layer useless.
Key sources
Chapter 11 cites in the print book: D23. Entries below are the sources this page draws on; some are related literature cited elsewhere in the book rather than in this chapter.
- D10 - Anthropic. (2024). Building effective agents. anthropic.com/engineering/building-effective-agents
- D11 - Willison, S. (2025). Designing agentic loops. simonwillison.net/2025/Sep/30/designing-agentic-loops
Primary-source links
- Anthropic, "Building effective agents" - for the multi-step agent pattern that informs the sequence diagram.
- Willison, "Designing agentic loops" - for the handoff-discipline framing the chapter draws on.
- The chapter's sequence diagram is Figure 11.1 in the print book; this page's worked example traces the same handoffs step by step.
Chapter contents
- The eight-step sequence diagram for one complete task.
- The handoff discipline at each transition: what each layer owes the next.
- A single worked example that runs the full loop end to end.
- What the failure looks like at each handoff (forward reference to Chapter 12).
Related resources
Use the worked examples to inspect end-to-end traces, including full dialog transcripts, multi-step agent action logs, and the per-layer state captured at each handoff.