[holsta]

Wild. I evaluate LLMs about once per year, and can't wait for the generative AI bubble to burst.

I most recently asked for a privilege-separated JMAP client daemon (dns, fetcher, writer) using pledge() and unveil() that would write to my Maildir, my khal dir and contacts whenever it had connectivity and otherwise behave like a sane network client.

I got 800 lines of garbage C. Structs were repeated all over the place, the config file was #defined four times, each with a different name and path.

[sysmax]

You need to do it in smaller, incremental steps. Outline the overall architecture in your head, ask the AI to create empty structs/classes. Build it. Ask it to implement one part, leaving others empty. Test it. Ask it to add the next thing, and so on.

Every step should only affect a handful of classes or functions, that you can still keep in your head and easily verify. Basically, same thing as if you were doing it by hand, but at a higher abstraction level, so faster and less mentally tiring.

Shameless plug: I am working on a new cross-platform IDE designed for just this kind of workflow. It has basic C/C++ support already: https://sysprogs.com/CodeVROOM/?features=why

[apwell23]

> You need to do it in smaller, incremental steps.

This isn't the context of this particular thread through. Its this

"Claude just tears through problems at breakneck speed."

[sysmax]

I think, there's nuance. If a human can solve a problem without second thoughts and hesitations (Hey, stop, this doesn't look right. Are there other options?), an LLM will tear through it at breakneck speed.

But if there are things worth hesitating and weighing, LLM will fly past them at cruising speed.

There are plenty of problems of both kinds.

[lelandbatey]

The statement you quoted is ambiguous. I'd say "Claude makes the boring parts of coding super fast, tearing through the stuff I didn't want to do and helping me get to the stuff I did want to do."

I'm an experienced dev (this year is my 10 year anniversary of 'officially' being in the industry). I've been using Claude Code over these last two weeks. It's like a magical code generator tool e.g. protoc; if you can describe the rote and boring "do this tedious x->y translation", then it can probably bang that out with tests in a couple minutes, making it easy for you to focus on the logic. It's not that Claude Code is doing the "programming", it's more like Claude is doing the typing (and it types REALLY fast). Literally imagine if you could type as fast as you could think, like just saying "oh I'll need functions covering x/y/z behaviors and cases, similar to the other functions already present but tailored like <way>." All you had to do was type that sentence and it wrote basically what you would have written, but instantly.

[tough]

> Literally imagine if you could type as fast as you could think

This is exactly how vibe coding has been feeling to me lately, I do read all the code while it writes it, and I command the AI to write/edit/fix the code, so its like faster programming but we're still on the director seat

also focusing on logic, and the problems at hand, and not syntax, API, or function signatures, allows one to build at much faster speed and focusing on architecture and actual UX and not just code

[HPsquared]

In think the people having success, probably have more experience with them. It sounds like "I tried using one of these new horseless carriages and it didn't go well, these things are useless"

[Kiro]

> Wild. I evaluate LLMs about once per year, and can't wait for the generative AI bubble to burst.

Strange thing to respond to people having great success with it. You clearly want it to fail, but why?

[holsta]

https://esgnews.com/ai-boom-drives-150-surge-in-indirect-emi...

Not to mention the ethical / copyright / misbehaving scrapers.

[rmwaite]

“Haters gonna hate”, as the old saying goes.