AI / Coding Theory
Fragments: April 29
Martin Fowler's roundup pulling threads across Parsons on harnesses, Adam Tornhill on function length in the AI era, and Nilay Patel's "software brain" essay on why so many people resent the AI worldview.
AI / Coding Theory
How I Use AI to Code
Chris Parsons' April 2026 update to his canonical agentic-coding guide. The headline shift since his 2025 version: "verified" no longer means "read by you" — it means checked by tests, type checkers, and automated gates, with your eyes reserved for the diffs that actually need judgment. The senior engineer's job is now to train the AI via AGENTS.md, skill files, and a portable markdown vault, not to hand-review every output.
AI / Coding Theory
Harness engineering for coding agent users
Birgitta Böckeler's full treatment of harness engineering — guides (feedforward) and sensors (feedback), split across computational and inferential. The most rigorous mental model out there for the scaffolding around a coding agent.