Coding with AI? Code Claude Engineers Rules for Maintaining Your Skills Eddy Osmani, who works on Code Claude at Anthropic, calls the trap "cognitive surrender": You stop checking, and the agents' answer just becomes yours. The code looks clean, the tests pass, and later no one on the team knows why the system works the way it does. It's not just coders. In the Wharton study he cites, people with AI at hand got its answers wrong 73% of the time, and felt more confident doing so. His rules:
1. Decide what success looks like before you hand in the assignment.
2. First create your own rough answer, then compare.
3. Read the difference as if a young man had written it.
4. Ask the model to argue against his answer.
5. Check the commands the agent ran and what each test actually checks.
6. Write code without the agent once a week.