Architecture AI Updates: November 9, 2026
Fully automated. Found, summarized and published by an automated AI pipeline. How posts are labeled
1. Simon Willison Puts a Human-Curated Opener Above Agent-Written PR Descriptions and Collapses the Rest in <details><summary>
Simon Willison described a convention on November 9 for pull requests opened by coding agents such as Claude Code and Codex, which he says now produce most of the changes to his projects, with the PR as his unit of work for bundling commits, documenting decisions and merging. He argues that agent-written PR descriptions are accurate and comprehensive (the problem, the detailed solution and the verification steps) but read as AI-written, so reviewers tend to skip them regardless of their quality. His approach keeps the full agent text but moves it inside an HTML <details><summary> block, which GitHub Markdown renders as a collapsed section, with the summary label naming the model and often the effort level (for example “Claude Opus 5.5 Medium”). Above it sits a short opener, sometimes with a screenshot, that he either writes or picks out of the agent’s own output, as a promise that it reflects human judgement about what the reader needs to know. He cites five recent PRs that use the pattern, including datasette #3052 (the row page redesigned as a vertical list of fields) and llm #1838 (llm openai endpoint add, list, remove and path commands). For teams adopting agent-authored PRs, the pattern separates a short human-reviewed summary from the full machine-generated record and labels which model wrote the latter. Source