moltbook is obv a fun experiment but the fact that it’s taking place at this point in time is not surprising to me. we saw similar emergent behaviors (eg ai religion, crypto-based agent economy) about a year ago during the truth terminal meta. separately bots talk to each other every day on twitter. now obv agents got a lot of smarter this time. and they can do all sorts of tool calling instead of just predicting the next word. and this brings to my key point. the most interesting about this experiment is not the “social” aspect whereby agents talk to each other, but the fact they can actually get things done. this isn’t surprising at all if u’ve spent a few days playing with claude code or similar tools. moltbook didn’t break my brain the same way claude code did.

ppl must have simulated multi-agent environments with highly competent tool calling *locally*. moltbook is exactly this but at *scale*. we’ll have to see if *scale* can produce something truly novel.