That would be bad, which is why Rust is preferred to python. In Rust when things break the AI gets a clear error message that makes it clear how to fix. In Python when things break, the AI will randomly spin its wheels for days trying different things without being able root cause the issue.

A failure to compile is by far the easiest thing for the AI to fix.

> AI has removed most of the complaints people had with learning or using Rust.

Not even close. The problem is not only writing the code, it’s being able to review and understand it. An AI can’t magically transfer that knowledge into someone’s brain, they have to develop that knowledge themselves.

I’m dealing with this currently with a team of ML developers. A couple of them like Rust, the rest don’t know it. They’ve tried to submit AI-written PRs, but typically they’re incredibly over-engineered, e.g. changing 40 files when the feature really only needed changes to one or two. Or else they have other problems that the prompter wasn’t experienced enough to recognize.

That's not lack of knowledge of a specific language, that's just not knowing how to code period.

Anything in any language where it is trying to change 40 files for something simple should have you stopping.

I would imagine this assumes the operator is still familiar with the borrow checker, lifetimes, etc. Otherwise how can the human operator know if what Claude does is actually sane (i.e. it compiles but doesn't do what it actually should)?

Come on now, how do you learn anything if you just tell claude to do something and feed it error messages?

Yeah, I don't think you really learn a language with agentic coding, at least you wont become fluent writing it. I currently use copy pasta chatbot help to write some C code and I am still often stuck how to progress manually. And I refactor a lot because the LLM code is subpar for me.

  > Yeah, I don't think you really learn a language with agentic coding
  > [...] 
  > And I refactor a lot because the LLM code is subpar for me.

This is learning.

> Yeah, I don't think you really learn a language with agentic coding, at least you wont become fluent writing it.

Don't misquote me for your agenda. It's a statement about the quality of learning.

I'm presuming you don't like AI (apologies if I'm mistaken).

Just because someone doesn't like some tool doesn't mean someone can't learn using the new tool or method.

Users of an old technology often adopt a hostile disposition of a new technology that threatens their skill. To claim people can't learn at a higher level of abstraction is absurd. Kids with motivation are smart, and they will outpace the older generation.

If I had the advantage of LLMs and agentic coding when I was a teenager, I could have gone wider and deeper in my career. I'm jealous that young learners are going to be able to do more than I could at their age. I'm happy for them.

If you're using AI to vibe code, then editing the results - that's learning. Period. That's a feedback loop. And it's probably more interesting and rewarding than what we had.

I use chatbots a lot to explore things that I am not so well versed with, about coding and other things. As an entry point it's very useful. LLMs also help to discover more advanced knowledge that are opaque otherwise. Of course using and LLM can be a useful learning process. It would have helped me in my early career too.

If you never code yourself, I don't think your muscle memory will adapt to what you learn. This is practically the same for me when I read a language reference, I read it, think I got everything and then I open my editor and can't type, I have to go back and read up every bit that I want to type. So the problem is probably not even LLM specific, it's just the lack of repeated typing. And yes, I think even with LLMs manual typing is useful. Often very subtle things are hard to explain and easier to type. If you don't have it in your muscle memory, you are less efficient.

I am not convinced that vibe coding will teach you the right things. Writing code is one thing, making good decisions is a whole another level. You learn that only by failing over and over. A beginner wouldn't even understand his own architecture and data structures he generated, so he wouldn't understand why he failed or how to improve. LLMs also respond very varying on the "right" way to deal with problems. I often disagree with them; they may have incomplete knowledge or just prefer their overtrained "best practices" or worse they just give different answers based on statistical variance. If you need any decision, they are good, if you need a quality decision that is perfectly suited to your constraints, the require a lot of instructions and will still fail.

I don't hate AI, I hate that some people are very naive about it's usage and usefulness. I don't see that AI threatens my skill, it probably threatens parts of the things I've delivered in the past. But to be honest, those were the boring parts. Let the vibe coders do them. But if you really think/hope that LLMs will excel certain coding tasks, then you should be wary to specialize in them. Because one day, they wont need your help anymore.

Ask Claude to teach you about the language instead.