Agentic Engineering Patterns - Linear walkthroughs

AI-assisted annotations

Cross-references from studying Linear Walkthroughs - Agentic Engineering Patterns by Simon Willison. Added with Claude Opus 4.6 via Claude Code.

Willison describes using coding agents to generate structured walkthroughs of codebases — turning vibe-coded projects into learning opportunities after the fact.

Related reading

• Build To Last — Chris Lattner warns that “delegating knowledge to an AI” while “just reviewing the code without thinking about what you want to achieve” is “very, very concerning.” Willison’s walkthrough pattern mitigates this by turning the review into active learning — but Lattner would likely ask whether the understanding should have been built during construction.
• Beyond agentic coding — Gabriella Gonzalez argues chat-based agents break flow state and proposes calm technology — AI that augments the editing experience peripherally (inlay hints, next-edit suggestions) rather than mediating through chat. Willison’s walkthrough pattern is arguably calm: passive, self-paced learning that doesn’t demand real-time attention.
• How to Solve It with Code — Jeremy Howard’s “Solve It” method would push back on the vibe-code-then-learn sequence: Howard argues you should stay in contact with the code as you build it, inspecting every output, because understanding compounds during construction, not after. The walkthrough closes the gap, but Howard would ask whether the gap should have existed at all.