If you are good at code review, you will be good at using AI agents

The author argues that using AI agents well is essentially a code review skill, so people who are good at reviewing code will also be good at tools like Claude Code, Codex, and the Copilot coding agent. The author says that language models produce a lot of code but lack the design judgement of a competent engineer, and that unsupervised agents will commit to bad designs. Building VicFlora Offline, an offline-friendly PWA for keying out plants, the author notes Codex spent considerable effort reverse-engineering the VicFlora frontend before the author found the raw data was more easily accessible. The author also says both Codex and Claude Code push to build background job infrastructure for short-lived parallel work, which the author considers overkill.