Michelangelo's David took years to complete, many studies and iterations. The patrons that paid for it wanted the perfection such an artist could provide - that was not something to be rushed.
Software that is ergonomic, intuitive, fast, and maintainable takes time. To use the analogy - chipping away at granite is a slow and tedious process, mistakes in planning or approach are difficult if not impossible to recover from.
Taking time to do it right sometimes is the only way to achieve quality.
The problem is that, in the real world, people are neither competing for maximizing the code quality score, nor are they spending remotely the same amount of time with vs. without AI. They're spending a small fraction of the time with AI than they otherwise would have, and their goal is to ship the product. Hence why they commit bad code.
To put it bluntly, I've literally seen engineers who used to send excellent pre-AI code suddenly start sending trash for review when they use AI. And, when I push back and point out the problems and they actually put in the effort to improve it, their code quality improves to (or beyond) what they (or I, for that matter) would've written without AI. I don't know how you could get a more fair experiment than this.
So, to answer your question directly, no. (No for other reasons too, but this is the only relevant one.)
I simply don't believe you.
For personal projects where I have a deep understanding of what I want to build and genuinely care about the quality, even Fable can't touch me. Of course, AI would get a working system done 50x faster, but personal projects aren't about speed for me.
You don't care at all about (c) maintainability or fragility?
One of the biggest problems I see is LLMs duplicate sources of truth they have no business duplicating... all the time. E.g. if an internal constant in some code is X then they just hard-code X in fifty other places far away and code based on that. And it's not just constants, they often duplicate entire pieces of logic too. So everything is fine and dandy when the code is fresh, but the next time the original is updated, the rest of the code breaks, because nothing forces this update to update too. Of course the garbage tests they produce never test such hidden assumptions either.
And this is before I get to other things, like (d) test costs and (e) the limits of test matrices. That "inline assembly" you just inserted now requires you to test everything on older CPUs that may lack newer instructions. And on other architectures too. That's a huge cost and this kind of thing exponentially grows your test matrix. Or, more likely, you end up with a breakage on some CPU because you didn't care.
If you have unlimited tokens i guess this is pretty much true these days. I've noticed opus write write and remove the same piece of code in a session. They prefer to write a lot, a lot of defensive checks in python etc.
At my company I have around 4000 GitHub copilot credits per month. I cannot afford to just let the LLM finish a whole project by itself using just spec documents. I did that once, ran out of tokens and then found out i couldn't understand what was going on.
This will probably change in the future but right now I need to work with what's available to me
I'll be blunt: Claude Code using Fable High. But, it's not the only one, you can't hardly swing a stick without hitting an AI that is great at coding. If by coding you mean "produces working software". If your definition is "produces code that doesn't offend my sensibilities", then that's a different matter, and also likely something that a human expert developer likely couldn't accomplish either. ;-)
Or are you saying that the "mythical AI" is mythical because people are using it without putting in any effort, therefore producing poor results? If that's the core of your argument, it sounds like the mythical AI could be out there and you'd not know it because those around you are, to coin a phrase, using it wrong.
I'm kind of struck though that your experience seems to all be second hand (people are using AI and submitting bad code); do you not have any first-hand experience with using claude code and fable?
I'm saying that most code I see AI generate is frequently bad, in ways that nobody disputes, not even the engineers who sent me the code to review - they themselves previously always avoided those exact problems when they used to write the code by hand.
> Or are you saying that the "mythical AI" is mythical because people are using it without putting in any effort, therefore producing poor results?
This also happens.
> If that's the core of your argument, it sounds like the mythical AI could be out there and you'd not know it because those around you are, to coin a phrase, using it wrong.
If you hired a brilliant math PhD who always wrote bad code and you had to keep reviewing every single thing they did and getting them to rewrite their code until it met the quality bar every time, and they never learned from their mistakes from one project to the next so this was a persistent issue, but they were so fast that even micromanaging them let you achieve a better outcome faster than you would have alone, what would you call that? That's what AI feels like.
> do you not have any first-hand experience with using claude code and fable?
I certainly do. But for anything meaty I frequently end up rejecting most of the code they produce before they get to other reviewers, and even then, I frequently have to do more work on them by hand. In some cases I've quite literally had to restart a project half a dozen times from scratch - with different approaches each time - in order to reach results that I'm not disgusted with.
Note that I'm not claiming to be an expert in using AI by any stretch - I'm still very much learning - but that I'm not really seeing these mythical results from others either.