One big problem with AI, especially if you’re new to programming or software development, is that you risk learning how to use AI, not how to solve problems.
There’s plenty of research showing that to truly learn something, you need to struggle with it first. That struggle matters. When you try to solve a problem you don’t yet understand, you’re forced to build mental models. When you later learn the correct solution, you retain it better and understand it more deeply.
AI removes that phase.
Instead of struggling with the problem, you learn how to ask the tool. While it feels productive, what you’re really optimizing for is prompting, not reasoning.
This will eventually bite you.
At some point:
- the bug is weird and context-specific
- the incident is happening at 2am
- the problem is new, messy, or outside the model’s comfort zone
When the AI doesn’t know what to do, you’re stuck. Not because the problem is unsolvable, but because you never learned how to solve problems. You learned how to use a tool.
That doesn’t mean “don’t use AI”. It means use it correctly.
How to actually learn effectively with AI
Start with the problem. Alone.
No prompts. No autocomplete. Try to solve it yourself.
Yes, it’s slow.
Yes, it’s uncomfortable.
That discomfort is the point.
Write your own solution first, even if it’s wrong. Especially if it’s wrong. You’re not trying to be correct, you’re trying to think.
Only after that, bring in AI. But use it as:
- a reviewer, not an author
- a sparring partner, not a replacement
Ask things like:
- “Is my reasoning sound?”
- “What edge cases am I missing?”
- “What alternative approaches exist and what are the trade-offs?”
Use AI to:
- validate or challenge your thinking
- suggest different angles
- point you to blog posts, articles, books, and deeper resources
Then re-implement. Don’t just read. Apply.
That’s how you actually learn and improve.
One more thing AI is bad at teaching you: judgment
AI is great at telling you that:
- everything is possible
- many approaches are valid
- lots of patterns are commonly used
And technically, that’s often true.
But one of the most important skills in software development is discernment.
Not “can I do this?”
But “is this a good solution for this problem, with these constraints, right now?”
AI is bad at telling you:
- when something is overkill
- when an abstraction will hurt later
- when a solution doesn’t fit your context
It tends to flatten everything into “sounds reasonable”.
Judgment comes from:
- trying bad solutions
- maintaining systems over time
- feeling the cost of wrong decisions
If you skip the struggle, you skip developing taste.
The bottom line
AI is a force multiplier for problem solvers.
It’s a crutch for people who never became one.
Struggle first.
Use AI to reflect, validate, and expand.
That’s how it makes you stronger, not fragile.