I have a strange feeling that we'll see an increasing number of these to the point where we'll start to see legitimate issues in ecosystems as real talent exits (due to projects no longer evolving, being deprecated, etc as a response to AI).

Will be quite funny when there's a noticeable decline in stability and development—despite having AI—because it turns out you actually need talented, creative people to build the building blocks the AI spits out. LLMs can parrot but they're basically the Crazy Town [1] of software development.

[1] https://www.youtube.com/watch?v=NI6xffJn4-w

▲

tarkin2 5 hours ago | parent | next [-]

Well, people aren’t coding to keep their skills and knowledge fresh.

And they’re not even learning from their AI agents since they’re not closely reviewing the code.

The chance of code quality increasing seems remote.

And that’s not to mention the new generation—having rarely coded enough to properly review code.

Nor to mention the AI learning from AI coded slop: mad cow disease for AI.

What do we have, no programmers, no one knowing what’s going on inside the test passing slopy black box, except a caretaker guy who’s kept their skills sharp by not using AI—with their time taken up trying to understand constantly rewritten code?

	▲	cowlevel 4 hours ago \| parent \| next [-]
		I tried to write a little project the way an AI would, going gung-ho and not stopping to ask if this is a good design. It actually went okay and now I have a silly thing that wouldn't exist otherwise. It was still slower than an AI, but only took a few days and now I know how the thing works inside, which I wouldn't if I used AI. I don't think programmers can't match AI, I think we don't want the associated responsibility of knowing we might be building a pile of shit. I think we want to avoid writing a pile of shit so we slow way the heck down. This is also the way I wrote programs when I barely knew how to program btw and I got stuff done then. Think about doing carpentry by nailing planks together in roughly the pattern you want, rather than meticulously drawing and then cutting outlines - which I have also done. The first iteration (of that carpentry project) fell apart the moment I tried to use it and was a miserable experience because I still had to complete the task with the pieces, and didn't have the opportunity to go back to fix it. The second iteration, which still just took a couple of days to build, was much stronger and was awesome. It can be fun and get results that are obviously acceptable because they are what the AI would have output anyway. Actually, just for curiosity, I just asked ChatGPT how to make the thing I made, and ironically it's doing a big design up front process where it wants to know all sorts of parameters first before it will tell me anything. So my "AI mode" was more direct than an actual AI! It also wanted me to buy a welding machine even though the thing can be made just fine, better even, out of wood. Maybe that's that new product advertisement feature.
	▲	rglover 5 hours ago \| parent \| prev [-]
		> What do we have, no programmers, no one knowing what’s going on inside the test passing slopy black box, except a caretaker guy who’s kept their skills sharp by not using AI—with their time taken up trying to understand constantly rewritten code? That's where we're headed but I'd say the smart teams will have that 'caretaker guy' use judgment to know when and when not to apply AI. That describes one of my businesses perfectly [1]. [1] https://lattice.engineering

▲

bjacobel 5 hours ago | parent | prev | next [-]

> when there's a noticeable decline in stability and development

When?

▲

rglover 5 hours ago | parent [-]

It's already started imo. Spotify's desktop app on MacOS is becoming unusable (up to ~30s delays after clicking a button before the action runs—used to be snappy). I just saw the Epic Games launcher thing. Anecdotally, countless websites just break at random (more so than even when it was 100% human).

	▲	tancop 2 hours ago \| parent \| next [-]
		Spotify doesn't care about how fast their app is because for a lot of their users there is no alternative. Everything else is paid only (Apple), has a bad quality library with duplicates and trash metadata (YT) or is a lot less convenient (physical media and most pirate options). Epic is completely different, their app was awful before LLMs and the rewrite made it a lot better. It loads 10x faster, takes half the memory, every action is instant and it finally has a clear UI. If OpenAI wanted a success story to promote Astra this would be a good candidate.
	▲	wartywhoa23 3 hours ago \| parent \| prev [-]
		The current macOS itself would suffice.

▲

morkalork 3 hours ago | parent | prev [-]

Presumably, private companies and individuals with the money to pay for tokens continue forward rolling their own in-house solutions and slowly abandon OSS as it rots. What happens when each organization has its own entire stack though? Sure, FAANGs already were like that but most smaller companies were sort of "speaking the same language" and reusing software like apache, mongodb, nginx etc.

	▲	rglover 3 hours ago \| parent [-]
		> What happens when each organization has its own entire stack though? My personal opinion is that all hell breaks loose due to continuity issues, time and money waste (excessive token spend to simulate competency), and a general air of malaise to the point where problems are tolerated far longer than necessary because "the bot will get to it eventually."