Type: url
Original: https://addyosmani.com/blog/agentic-skill-decay/
Status: generated
Language: en
Summary
This post argues that coding agents can short-circuit the trial, failure and debugging through which engineers used to build expertise, so the repetitions that produce judgment now have to be pursued deliberately — forming a hypothesis before prompting, asking why, reading diffs, predicting failures and occasionally working by hand. It frames good agent work as resting on deep domain expertise and applied judgment, calls verification the floor and imagination the ceiling of what can be built, and recommends codifying lessons as tests, lint rules or durable notes so that each correction improves both the engineer and the agent.