At the very least, it’s worth reflecting on whether any approaches are based on observing actual behavior, and whether they have incorporated a reality where may millions of people have made these tools a daily habit, even if we know that is one that is artificially subsidized right now.

I drew this silly chart early on in my own look at LLMs. I stand by it to this day. More so, even, as the bubble cast a stronger shadow over the AI ecosystem.

LLMs have some very good uses. I've found them helpful for gathering my own thoughts and brainstorming the book I've been working on. But the metacognition to understand what they can and can't do is uncommon. I wonder a bit if it's similar to systems metacognition, techies seem to "get" it more.

she's also unhappy with me for pointing it out.

1/10th of the original shot, roughly.