How about a $10 billion fine for OpenAI for every mistake? Make it hurt. Make them pull the plug on this travesty.

It goes beyond the problems introduced by the model router, though. I have to work with GPT 5.2 for my job (along with Claude, Gemini, and a few others), and we have enterprise API access to it. So when I select GPT 5.2 as the model to use, it’s spending tokens to actually use it.

And it’s pretty bad. It’s noticeably worse than the 4.x series. I find myself having to fix its mistakes far more often.

I’ve struggled to reason out an explanation, and model collapse really seems like a contender, especially if you follow information theory and why training these things is so hard.

As it happens, there’s a new talk about exactly this from George D. Montañez. You might find it interesting: https://youtu.be/ShusuVq32hc

Then why are newer versions of the major models performing so poorly? For instance, GPT 5.2 is definitely not an improvement over 4.5. What’s the root cause?

And that is why I no longer buy anything from them. I’m just embarrassed it took me as long as it did to realize what they were really doing.

I mean, we’re watching it happen. I don’t think it’s hypothetical anymore.

It’s already happening. GPT 5.2 is noticeably worse than previous versions.

It’s called model collapse.

Oh, the guy from Hermit Tech! He’s great, and his blog is hilarious and poignant in turns (though sometimes both at the same time).

You wanna know who really bags on LLMs? Actual AI developers. I work with some, and you’ve never heard someone shit all over this garbage like someone who works with neural networks for a living.

They’re also not providing a large language model, so they actually did have a path to profitability. It’s keeping LLMs updated and running that costs so much money that companies trying to do so are losing billions, and Midjourney doesn’t have that problem.

It’s just that their path to profitability was built on plagiarism on an astonishing scale. You’re spot on, they should have been utterly destroyed right at the start.

And it’s not even working. Not one of the AI companies is profitable. So they’re putting the hope for profits some time in the future over sanity and safety.

I have a work-supplied laptop with Windows on it. I use it maybe once or twice a month, just for the things requiring a VPN. The rest of the time it sits there gathering dust while I get real work done on my Linux laptop.

The specs on the work laptop say it should be a performance beast, but my Linux machine (with half the RAM) runs circles around it.