My problem is that, for a surprising number of applications, it's taken me longer to have the conversation with chatgpt to get the code I want than just doing it myself.

Copilot and the likes are legit for boilerplate, some test code, and posix/power shell scripting. Anything that's very common it's great.

Anything novel though and it suffers. Did AWS just release some new functionality and only like 4 people have touched it so far on GitHub? Are you getting source docs incomplete or spread out amongst multiple pages with some implicit/betwen-the-lines spec? Eh, good luck, you're probably better off just reading the docs yourself or guess and checking.

Same goes for versioning, sometimes it'll fall back into an older version of the system (ex Kafka with kraft vs zookeeper)

Personally, the best general use case of LLMs for me is focus. I know how to break down a task, but sometimes I have an issue staying focused on doing it and having a reasonably competent partner to rubber duck with is super useful. It helps that the chat log then becomes an easy artifact to more or less copy paste, and chatgpt doesn't do a terrible job reformatting either. Like for 90% of the stuff it's easier than using vim commands.

It seems great for like straightforward linear code, elisp functions, python data massage scripts, that sort of thing. I had it take a shot at some new module for a high volume Go server with concurrency/locking concerns and nil pointer receivers. I got more panics from the little bit of code GPT wrote than all my own code, not because it was bad code but because when I use dangerous constructs like locking and pointers that can be nil, I have certain rigid rules for how to use them and the generated code did not follow those rules.

> Do you really think people are just blindly pushing code they can't make heads or tails of? That they haven't tested?

Yes, most definitely. I've recently been introduced to our CTOs little pet project that he's been building with copious help from ChatGPT, and it's genuinely some of the most horrid code I've ever seen in my professional career. He genuinely doesn't know what half of it even does when I quizzed him about some of the more egregious crap that was in there. The real fun part is that now that it's a "proven" PoC some poor soul is going to have to maintain that shit.

We also have a mandate from the same CTO to use more AI in our workflows, so I have 0 doubts in my mind that people are blindly pushing code without thinking about it, and people like myself are left dealing with this garbage. My time & energy is being wasted sifting through AI-generated garbage that doesn't pass the smell test if you spend a singular minute of effort reading through the trash it generates.

Yes that's exactly what they are doing.

I literally had someone with the balls to tell me that it was ChatGPT's fault.

Due diligence and intelligence has shit the fucking bed quite frankly.

Do you think ChatGPT has changed any of those answers from Yes to No? Because it hasn't.

People blindly copied stack overflow code, they blindly copied every example off of MSDN, they blindly copy from ChatGPT - your holier than thou statements are funny, and frankly most LLMs cannot leave a local maxima, so anyone who says they dont write any code anymore I frankly think they are not capable of telling the mistakes, both architecturally and specifically that they are making.

More and different prompting will not dig you out of the hole.

This. Most people I know that use LLMs to be super productive are like "make me a button, it's red" (hyperbolic statement but you know what I mean). I can do that faster and better myself.

When I'm deeply stuck on something and I think "let's see if an LLM could help here", I try (and actually tried many times) to recruit those prompting gurus around me that swear LLMs solve all their problems... and they consistently fail to help me at all. They cannot solve the problem at all and I'm just sitting there, watching the gurus spend hours prompting in circles until they give up and leave (still thinking LLMs are amazing, of course).

This experience is what makes me extremely suspicious of anyone on the internet claiming they don't write code anymore but refusing to show (don't tell!) -- when actually testing it in real life it has been nothing but disappointment.

> Do you really think people are just blindly pushing code they can't make heads or tails of? That they haven't tested? Do you seriously think people are just one shotting code and blasting it into prod?

Yes, and I see proof of it _literally every day_ in Code Reviews where I ask juniors to describe or justify their choices and they shrug and say "That's what Copilot told me to put".

That sounds more like poor hiring decisions.

That sounds more like moving the goalposts. The claim (via sarcastic comment) was that people do not simply push code that they do not understand - and I provided a counter-example. No-one in that conversation disagrees that that's a bad practice - but until and unless I have full mandate to hire and fire whoever I want to work with, or to change jobs at will, I'm going to have to work with people whose development practices I disagree with.