2026-08-30 UTC
# 
[Mike] I probably know much more than your average person (although not as much as a professional AI expert or engineer) of the nuts and bolts about how LLM’s actually work under the hood, but I really question what they’re going to do about what they’re calling (I think disingenuously—but that’s a different side conversation) “misalignment”. The problem is, every means of control I’ve seen seems to rely on the same flawed approach: a