We now live in an age where you can learn while asking an AI. But looking at the learning logs, the people who use one split into two groups. The difference is not how often they ask — it is how.

Ways of asking that stall

  • "Write this code for me" — you receive only the finished product
  • "Fix this error" — you never check why it was fixed
  • "Why does this not work?" — thrown out without a hypothesis of your own

All three move you forward in the short term, but none of them can be reproduced the next time something similar happens.

Ways of asking that make you better

  • "I think count is still a string — how would I check that?" — attach your hypothesis
  • "What is the difference between these two ways of writing it?" — ask by comparison
  • "I am going to summarise my understanding; please point out what is wrong" — ask for verification, not output

Whether you can use an AI as something to check your answers against is the single biggest fork in the road.

Using AI on lesson exercises

When you ask an AI about a lesson exercise, do not ask it for the answer. Ask what to check next and how to check it. It looks like a detour, but it turns out to be faster that way.

When to use which

For exercises while you are learning, treat it as something to verify against. For implementation work that is genuinely urgent, treat it as a way to generate a first draft. Different purposes, different correct usage.