Searching for stuff on the TV show The Way Home last night, and the word salad that Google's AI Overview provided was... something. I'm becoming more selective about publishing slop, even to laugh at it, but it feels like when we inadvertently uncover this stuff we need to note it, as a reminder.

I remember one of B.F. Skinner's books ending with "Now, let us see what man can do with man". If the continuum from A/B testing ads to LLMs and AI psychosis is the leverage of automated exploits of unconscious human behaviors, I think... ... I think we're discovering what that means.

When I type this out, it feels obvious: Multi-layered AI agents (OpenClaw, etc) are an attempt by an industry that's constantly flailing at trying to automate processes they don't understand to automate their own bad processes.