Addy Osmani16 min readintermediate
Agentic Skill Decay
Summary
Addy Osmani warns that AI agents can short‑circuit the hands‑on practice (“reps”) that builds deep expertise and judgment. He recommends deliberately inserting hypothesis‑forming, “why” questioning, diff inspection, failure prediction, and occasional manual coding into the workflow, especially for junior engineers. A 2026 Anthropic study showed junior developers using AI scored 17 % lower on a fo…
- Agents accelerate code generation but skip the learning loops that create expertise; engineers must re‑introduce those loops deliberately.
- Two core abilities for effective agent‑assisted work: deep domain expertise and applied judgment (defining outcomes, constraints, tests).
- Practical habits: write a hypothesis before prompting, ask “why” throughout, read diffs and predict failures, verify results, and occasionally solve problems by hand.
- Empirical evidence: junior engineers using AI assistants on a Python library scored 50 % vs 67 % for manual work; conceptual questioning narrowed the gap.
As AI coding assistants become ubiquitous, the risk is that engineers—especially those early in their careers—will miss the iterative debugging and reasoning that historically built deep technical intuition. Without intentional practice, skill decay can reduce the quality of code reviews, system de…
5/10
