I am perpetually amused at the use of the word 'vibe' in 'vibe coding'.

> it’s much more worthwhile to work with a plan composed of discrete tasks that could be explained to a junior level developer

I'm sure this is a more effective way to get more usable results. But I really think anyone in this situation should be taking it as a kind of wake-up call. Mentoring/guiding a junior is work -- there's a significant cost. But it's a cost easily justified -- a lot of people find it intrinsically rewarding, you're training a colleague, etc.. What you're describing here, though, is all of the cost with none of the benefits and you're being the junior developer as well (you must be -- you're doing their work). You're alone, mentoring a chatbot that cannot learn or grow.

> I’m beginning to think the problem runs deeper, and it has to do with the economics of AI assistance.

> When charging by token count, there’s naturally less incentive to optimize for elegant, minimal solutions.

... Or maybe the tool just isn't that good / what you want. There doesn't have to be a conspiracy behind it.

That is to say, I think the main point presented here is very unconvincing. If they could build a tool that could just do what you want in an acceptable manner, they would. People would obviously throw money at that.

It produces verbose, comment-heavy, procedural code because that form reasonably effectively supports the nature of the generator. Procedural code is obviously well-suited to "what comes next?" style append-oriented editing operations. Verbosity eliminates nuance.

OK, one more thing:

> While we wait for AI companies to better align their incentives with our need for elegant code I’ve developed several strategies to counteract verbose code generation

"While we wait" is the most depressing thing I've read in a while. It's just completely at odds with the field itself.