Judgment in the Age of AI · Article 49
Skills do not disappear merely because a tool is used. They weaken when you stop performing a critical step for a long time. The question is not whether to use AI, but which abilities must remain callable when you are offline, rushed or faced with a wrong output.
Keep a capability ledger
| Capability | AI may handle | Practice to retain |
|---|---|---|
| Writing | Organise material and flag ambiguity | Regularly draft core paragraphs from a blank page |
| Analysis | Generate hypotheses and repeat calculations | Define the problem and hand-check samples |
| Programming | Scaffold and explain errors | Build and debug a small feature unaided |
| Language | Give feedback and examples | Practise unsupported expression and real conversation |
Use an attempt–assist–repeat cycle
- Attempt independently to expose the real weakness;
- Use AI for a hint, feedback or an alternative method;
- Close the answer and perform the task again from the beginning;
- Test transfer in a new context several days later.
Each month, perform one critical task as a no-AI baseline and record speed, errors and confidence. If capability declines, reduce the stage delegated to AI rather than discovering the loss only when independent work becomes necessary.
References
Discover more from Geoffrey Chen
Subscribe to get the latest posts sent to your email.