It sure seems like it. That or delirious, misguided fans who want AI to succeed even though there would be no benefit in it for them if that were the case. They'd just be serfs again.

I watched a man struggle for 3 hours last night to "prove me wrong" that Gemini 3 Pro can convert a 3000 line C program to Python. It can't do it. It can't even do half of it, it can't understand why it can't, it's wrong about what failed, it can't fix it when you tell it what it did wrong, etc.

Of course, in the end, he had an 80 line Python file that didn't work, and if it did work, it's 80 lines, of course it can't do what the 3000 line C program is doing. So even if it had produced a working program, which it didn't, it would have produced a program that did not solve the problems it was asked to solve.

The AI can't even guess that it's 80 line output is probably wrong just based on the size of it, like a human instantly would.

AI doesn't work. It hasn't worked, and it very likely is not going to work. That's based on the empirical evidence and repeatable tests we are surrounded by.

The guy last night said that sentiment was copium, before he went on a 3 hour odyssey that ended in failure. It'll be sad to watch that play out in the economy at large.

▲

sedawkgrep an hour ago | parent | next [-]

This is an interesting thread to me and my experience as a total hack programmer has been both awesome and pathetic with regards to AI.

Recently I tried working with both ChatGPT and Gemini (AI Studio) to take a really badly written PHP website and refactor it into MVC components. I've worked strictly through the web UI as this only involved a few files.

While both provided great guidance around how to approach disentangling this monolithic code, GPT failed miserably, generating code that was syntactically incorrect, and then doubling-down on it by insisting that it was correct and that errors were faults lying elsewhere. It literally generated code that lacked closing parenthesis and brackets.

In contrast, Gemini generated a perfectly working MVC version from the start. In both instances I did intentionally keep it on track to only separate the code into MVC and NOT to optimize anything, but it worked the first try. I've then taken it through subsequent refactorings and it's done superbly in that role.

So I can't speak to how well this works for large code bases, much less agentically. (my initial, very focused MVC refactor was about 1300 lines.) But when giving it a very specific task with strict guidance and rules, my results with Gemini were fantastic.

▲

user34283 6 hours ago | parent | prev [-]

AI works. This is evidenced by my side project which I spent some 50 hours on.

I'm not sure what your "empirical evidence and repeatable tests" is supposed to be. The AI not successfully converting a 3000 line C program to Python, in a test you probably designed to fail, doesn't strike me as particularly relevant.

Also, I suspect that AI could most likely guess that 80 lines of Python aren't correctly replicating 3000 lines of C, if you prompted it correctly.

▲

discreteevent 5 hours ago | parent | next [-]

> AI works.

For some definition of "works". This seems to be yours:

> I'd go further and say vibe coding it up, testing the green case, and deploying it straight into the testing environment is good enough. The rest we can figure out during testing, or maybe you even have users willing to beta-test for you.

> This way, while you're still on the understanding part and reasoning over the code, your competitor already shipped ten features, most of them working.

> Ok, that was a provocative scenario. Still, nowadays I am not sure you even have to understand the code anymore. Maybe having a reasonable belief that it does work will be sufficient in some circumstances.

https://news.ycombinator.com/item?id=46315569

▲

user34283 4 hours ago | parent [-]

Yes, as I said, it is working well in my side project. The application works and I am happy with my results so far.

It's interesting how this workflow appears to almost offend some users here.

I get it, we all don't like sloppy code that does not work well or is not maintainable.

I think some developers will need to learn to give control away rather than trying to understand every line of code in their project - depending of course on the environment and use case.

Also worth to keep in mind that even if you think you understand all the code in your project - as far as that is even possible in larger projects with multiple developers - there are still bugs anyway. And a few months later, your memory might be fuzzy in any case.

▲

stavros 4 hours ago | parent [-]

People seem to be divided between "AI doesn't work, I told it 'convert this program' and it failed" and "AI works, I guided it through converting this program and saved myself 30 hours of work".

Given my personal experience, and how much more productive AI has made me, it seems to me that some people are just using it wrong. Either that, or I'm delusional, and it doesn't actually work for me.

▲

SamPatt 13 minutes ago | parent | next [-]

The models are good enough now that anyone who says AI doesn't work is either not acting in good faith or is staggeringly bad at learning a new skill.

It's not hard to spend a few hours testing out models / platforms and learning how to use them. I would argue this has been true for a long time, but it's so obviously true now that I think most of those people are not acting in good faith.

▲

PostOnce 3 hours ago | parent | prev [-]

What about "I attempted to guide it and it took 5 times longer and I ended up having to do the entire thing myself anyway"?

	▲	stavros 43 minutes ago \| parent [-]
		That hasn't happened to me.

▲

blibble 2 hours ago | parent | prev [-]

why is it always accounts with 50 karma saying this?

▲

hombre_fatal an hour ago | parent | next [-]

I have 22k karma and I think it's a trivial claim that LLMs work and that software is clearly on the cusp of being 100% solved within a couple years.

The naysaying seems to mostly come from people coping with the writing they see on the wall with their anecdote about some goalpost-moving challenge designed for the LLM to fail (which they never seem to share with us). And if their low effort attempt can't crack LLMs, then nobody can.

It reminds me of HN ten years ago where you'd still run into people claiming that Javascript is so bad that anybody who thinks they can create good software with it is wrong (trust them, they've supposedly tried). Acting like they're so preoccupied with good engineering when it's clearly something more emotional.

Meanwhile, I've barely had to touch code ever since Opus 4.5 dropped. I've started wondering if it's me or the machine that's the background agent. My job is clearly shifting into code review and project management while tabbing between many terminals.

As LLMs keep improving, there's a moment where it's literally more work to find the three files you need to change than to just instruct someone to do it, and what changes the game is when you realize it's creating output you don't even need to edit anymore.

	▲	blibble 35 minutes ago \| parent [-]
		wasn't sure if this was sarcasm until this point: > with their anecdote about some goalpost-moving challenge designed for the LLM to fail (which they never seem to share with us). literally what the boosters do on every single post! "no no, the top model last week was complete dogshit, but this new one is world changing! no you can't see my code!" 10/10 for the best booster impression I've seen this year!

▲

user34283 2 hours ago | parent | prev [-]

If we're going to argue on that level: Maybe it's because accounts with 12k karma spend more time posting than working on side projects and trying new tools.

	▲	blibble 2 hours ago \| parent [-]
		that's the great thing about non-vibe coding faster, fewer bugs, better output leaving more time for shitposting