Let’s lay out the big picture for LLM style AI.
It is is statistical prediction of what should be the next word or symbol. That is why it required so much data to train and why, even if we had the tech, we couldn’t have created it 20 years ago: not enough data in digital format. It is not intelligent. It is not conscious. It is just an algo trained with a TON of data and which used massive amounts of processing power (and thus electricity) to produce results. Hallucinations are part of the tech, they cannot be eliminated, which means that LLM “AI” will always make mistakes and many will almost certainly be the sort of mistakes trained humans rarely make.
The current build-out in the US involves only a few companies, all in a huge circle jerk, and they make up 40% of the entire public stock market’s value. Neither Open AI nor Anthropic actually make a profit, and it costs more to do a query than is made even from paying customers, let alone all the free ones. There is absolutely no question in my mind that they are in a bubble.
The maximalist claim for “AI” is that it will become so smart it can replace at least 40% of jobs. (Or smart enough.) The more realistic claim is that it’s good for some things and can replace some workers by making those who remain more efficient. Plus, after all, most tech companies don’t care if their products are shit as long as they make money. See Google for “who cares what you think, it’s us or no one. You’ll use our product no matter how shit it is.” (Ironically, Search is one of the few things AI is better at than incumbents.)
So here’s the thing: no matter whether AI is a real tech or not, it’s in a bubble. (The internet was real, it had a bubble.) No one actually knows who’s going to make money from AI. The big internet winners (Amazon, Facebook, Google) came after the dot-com bust. The Feds may backstop and/or bailout, if they do, it will hurt everyone not involved.
If I am wrong about AI and the maximalist claims are true then what will happen is a massive replacement of tens of millions of workers. Since those people will now have almost no income, that will lead to a classic demand depression. A great depression like the one in the 1930s. The only way out would be a massive guaranteed annual income. Given our rulers and ideology, we’d probably have food riots long before they realized they were risking their own throats.
If it is a real tech, but not that big a deal, it will lead to a shittier economy where even more mistakes are made, and it’s even harder to find a human being to fix anything. Which is what tech wants: they want everything automated and certainly they don’t want to have real customer service.
And, if it is a real tech, as I have noted before, China is actually going to win. Their models are 20 to 30 times cheaper to run, and are open source. If your business uses AI you will use open source if you have half a brain, because with open source one of two major providers (Anthropic/Open AI) can’t just raise prices or change the model. To use closed source would be so stupid that even most American CEOs will not do it. Certainly no one with sense outside American vassal swarm will be so stupid.
So:
- Maximal AI leads to a great depression.
- Moderate AI leads to a shittier economy and shittier projects.
- There’s a bubble either way
- At the end of it China’s AI models will be used far more than American ones anyway. The US has already “lost” the AI race and can’t even see that. (Why? Fundamentally because they’re greedy and want to become billionaires of trillionaires. Genuine open source AI won’t print nearly as many rich people.
America can’t win at anything that matters any more, because the people who lead America are stupid, liars and so greedy they can’t think of anything but money. (See Trump, who is the avatar of all these vices.)
This site is only viable due to reader donations. If you value it and can, please subscribe or donate.






As everyone’s probably aware, Mamdani, a brown muslim social-democrat who has promised, among other things, to open city grocery stories, make transit free, a rent freeze on stabilized units (about 40% of New York’s apartments), universal free child care and to build 200K new apartments. He’ll pay for it with tax hikes on rich New Yorkers and corporations.
This is, again, because Chinese models are at least 90% cheaper to run, and mostly open source. Only a complete and utter moron would run their business using proprietary models where OpenAI or Anthropic can jack up the price any time they want or depreciate the model you actually needed. Even US startups agree, 70 to 80% of them are using Chinese open models.