Thoughtful Money with Adam Taggart · Tuesday, August 11, 2026
Ed Zitron argues that current large language models (LLMs) are not true AI but rather 'compute-based automation.' He highlights that they require constant 'nannying' and are not reliable, unlike traditional software designed to enhance human capabilities.
“When in fact, every story I hear about LLMs, even from the people that love them, is this talk of almost a nannying thing where it's like, well, it's really good. Well, you have to poke it, poke it every so often. You can set it on a loop, but you have to make sure the loops working.”
“It isn't AI. It's compute-based automation, writ large. It's the most vulgar version of script kitty stuff I've ever seen.”
“What is the end point product? What does it do? How can you actually, like, reputably say these are the effective things it's good at, these are the things it's bad at? Because it seems to, even on relatively small stuff, sometimes get them right, sometimes get them wrong.”