AI coding assistants are genuinely useful when you are learning. They are also the easiest way ever invented to feel productive while learning nothing.
The difference is not whether you use them. It is how.
The trap, stated plainly#
Learning to program is mostly learning to hold a problem in your head, break it into steps, and notice when a step is wrong. That skill is built by struggling with problems.
If an assistant hands you a working answer every time you get stuck, you skip the exact part that would have taught you something. You end up with working code and no ability to write it again.
The goal is to use AI in ways that keep the thinking with you.
Five ways that work#
1. Ask it to explain, not to write#
Instead of “write a function that removes duplicates from a list”, try “explain what a set is and why it removes duplicates”. Then write the function yourself.
2. Use it as an error translator#
This is where assistants are outstanding for beginners. Paste an error and ask what it means in plain English — not for the fix.
“I got
TypeError: unsupported operand type(s) for +: 'int' and 'str'. Explain what this message means in general, without looking at my code.”
Now you understand the error class, and you will recognise it next time. Pair this with how to read programming errors.
3. Ask for a review after you have written it#
Write your version first. Then ask what could be clearer, what edge cases you missed, and what a more experienced developer would change. You get the feedback of a code review without giving up the writing.
4. Ask for exercises, not solutions#
“Give me five practice problems on Python dictionaries, from easy to hard. Do not include the solutions.”
Generating good practice material is genuinely useful and costs you nothing.
5. Ask it to be the confused student#
Explain a concept to the assistant and ask it to point out what you got wrong or left out. Teaching something is the fastest way to find the holes in your own understanding.
A rule that works#
Before asking, answer this: am I stuck on understanding, or stuck on typing?
- Stuck on understanding — ask for an explanation. That is exactly what it is for.
- Stuck on typing — you know what you want but not the syntax. Ask, and note the answer down. Low risk.
- Stuck on the approach — this one you should sit with. Working out what to do is the skill you are here to build.
Where AI is unreliable#
- Recent library versions. It may confidently use an API that changed or never existed.
- Your specific project. It cannot see your files unless you show it, so it guesses names and structure.
- Security and correctness at the edges. Code that works for the happy path often ignores empty inputs and error cases.
- Saying “I do not know”. A plausible wrong answer looks identical to a right one.
Questions people ask#
Will AI make learning to program pointless?
Someone still has to decide what to build, judge whether the output is right, and fix it when it is not. All of that requires understanding code. If anything, reading and reviewing code matters more now, not less.
Is it cheating to use AI on a course?
Check the rules of your course. Beyond that, ask what you are optimising for: finishing modules, or being able to do the work afterwards.
Which assistant should I use?
They are close enough that it barely matters at this stage. Pick one, learn to prompt it well, and spend the saved energy on building things.
Should I use autocomplete in my editor?
Inline suggestions are the easiest to accept without thinking. Many beginners benefit from turning them off for a few weeks while the syntax settles, then switching them back on.