Confession: I’ve been using Claude Code to write all my code for me. And I think it’s making me worse at the thing I’ve loved doing for twelve years. I can clearly see how AI coding is rewiring our brains – it makes developers crave instant gratification instead of deep understanding, and reduces us to gamblers who pull levers for the next hit of working code. If this is happening to me, someone who learned to code in the pre-AI era, what’s it doing to junior developers who’ve never known anything else?
I guess, one might argue that you don’t need to understand neither solution nor problem anymore, if the problem solving is done for you.
But yeah, I’m not convinced of that. Most of the time, the actual difficulty in problem solving is working out the requirements. The AI won’t magically know your requirements either. It’ll just pick a random guess and produce a result that’s subtly wrong.
I guess, I can also provide an anecdote for that. A few weeks ago, we were working on a time tracking software, so where the user would enter start and end time and we get to calculate how long that took.
That’s probably gonna sound easy to someone without programming experience, but it is absolutely not. We don’t know what timezone the times provided by the user are in, so if it’s on the switchover from winter time to summer time, or vice versa, you actually cannot implement this correctly.
So, I wanted to discuss with a colleague how we should handle that and their first reaction was to ask the AI. Which is fine, I did also do a web search to start out, to see if there’s a library that does this handling best-effort. The AI did reproduce the same non-applicable StackOverflow answer that I had found just before that.
But this would not have been fine, without understanding the problem. If the AI would have just copy-pastad that StackOverflow code, that would have resulted in a bug.
And I myself would not have realized that this is a problem, would not have realized the need to work out the requirements and decide how to implement it, if I would not have taken a crack at solving the larger problem myself.
calculator did calculations, not problem solving.
I guess, one might argue that you don’t need to understand neither solution nor problem anymore, if the problem solving is done for you.
But yeah, I’m not convinced of that. Most of the time, the actual difficulty in problem solving is working out the requirements. The AI won’t magically know your requirements either. It’ll just pick a random guess and produce a result that’s subtly wrong.
I guess, I can also provide an anecdote for that. A few weeks ago, we were working on a time tracking software, so where the user would enter start and end time and we get to calculate how long that took.
That’s probably gonna sound easy to someone without programming experience, but it is absolutely not. We don’t know what timezone the times provided by the user are in, so if it’s on the switchover from winter time to summer time, or vice versa, you actually cannot implement this correctly.
So, I wanted to discuss with a colleague how we should handle that and their first reaction was to ask the AI. Which is fine, I did also do a web search to start out, to see if there’s a library that does this handling best-effort. The AI did reproduce the same non-applicable StackOverflow answer that I had found just before that.
But this would not have been fine, without understanding the problem. If the AI would have just copy-pastad that StackOverflow code, that would have resulted in a bug.
And I myself would not have realized that this is a problem, would not have realized the need to work out the requirements and decide how to implement it, if I would not have taken a crack at solving the larger problem myself.