Dave White

I've been working on prediction markets for events whose outcomes aren't just "yes" or "no."

Instead of betting on just one outcome, traders in Distribution Markets can express how likely they think each different possibility is across the whole infinite range of outcomes.

1/

controversial (?) opinion: i believe negotiating an AI training slowdown with china would be largely enforceable

roughly speaking, things i think are true:

-- we can track where the majority of current compute lives today
-- to train next years models, much of the compute we would need isn't even built yet
-- we probably find a way to prove that we are using the majority of our compute for inference rather than training using zk (yes, really)
-- this means both sides can trust that the other is not running full speed ahead on pretraining the next gen of huge models and gives us time to prepare even as training efficiency improves such that "dark compute" can train them

can expand on these points if people disagree on any in particular

"i" built a tool that lets you use your e-drums with all the stock @Ableton kits and have them mapped correctly

currently works just w/ efnote but point your agent at the code and it'll work with anything

github.com/david-alexande…

"hey astra, go check out thelatent.cafe and participate if you'd like"

i wonder if there's now room for a new modality of maintainer-free open source

rather than issuing a pull request, all code contributions have equal footing, and then as a user you can have an ai agent be the "maintainer" of the repo you actually use which is cobbled together from all of those

of course there's room for curators in this ecosystem etc.

possibly falls apart when it comes to interop but APIs could be the one little island of standardization