Everyone appears to be having roughly 99% of the same insights about AI use at the same time but it feels original to everyone rediscoverer since almost none of it is common knowledge yet. So we get a discourse thats mostly about agreeing furiously and noisily and a kind of credit-fomo race. The mutual-belief/private knowledge to common-knowledge gap is big. So there is massively parallel CK game playing. Right now it’s cheaper to reinvent wheels than to imitate.

There is meaningful competition to name things (which will mostly be won by Andrej Karpathy), but you should assume that every new insight is discovered by about 10,000 people roughly within days of each other, and that the person who names a new pattern or idea may not have the best version of it (though Karpathy often does)

Genuinely differentiated AI use, or differentiation in talk about use, is vanishingly rare, but because there is not much imitation learning and consensus common knowledge, there’s a weird sense of expecting variety and not finding it.

I think this will shift slowly. CK will gel. Differentiated use and talk will show up, and it will mostly look like domain-specific magic where the magic is non-trivial to generalize. Like those guys bypassing normal DOM manipulation and doing browser rendering in seemingly magical ways. I’m sure there’s an insight there that generalizes to other domains but it’s not obvious.