> But most of these solutions were more crude than they let on, and you wouldn't really know unless you were working in AI already.

Same with LLMs. Despite having seen this play out before, and being aware of this, people are falling for it again.