

LLMs are never going away.
Are you basing this assumption on the idea that they’re free to use? Because that’s not sustainable.
The every single time an AI company has tried to actually require compensation for the tokens burned for AI queries, even amounts that would hardly cover a fraction of the operational cost for that ‘work’, contracts have mass-canceled immediately.
You don’t have any control over token churn, either:
First of all, no matter how simple the prompt, LLMs can get stuck in thought loops that can chew through thousands of tokens before an answer can come out.
And second, every prompt you make not only eats a bunch of tokens incorporating the system prompt that can be billions of parameters long, but also includes every PRIOR prompt you put in, AND all their responses for “Context”, therefore token use increases geometrically for every next prompt.
The only thing holding its usage aloft is that nobody has had to pay up front. But the bill is racking up more and more every day for the power and cooling and facility upkeep, presently “paid for” by massive debts. When these firms that took out those debts file for bankruptcy, you do realize what’s owed doesn’t just magically poof into smoke, right? Even if a court literally ordered an injunction that those debts be stricken from the books as if they never happened, there is no financial firm on earth that wouldn’t find some backhanded way to balance the loss by shifting the burden to its other clients.
Maybe if any one of these companies posts a net profit, EVER, I might change my tune… but so far, this has been the emperor’s new clothes all over again.



moreover, I’m convinced the entire reason my generation (millennials) turned out to be tech-savvy was because adults didn’t understand it, were trying to control and curtail our usage, and we were mostly focused on finding ways to circumvent boomer and gen-x meddling in our usage.