Been trying to reconcile a bunch of different aspects of NN behavior and cognition by playing with formulations of latent space at a level (also) suited for reasoning about everyday cognitive phenomena In part it's led to a compact, diagrammable geometric system

RT @Corca_math: From the very beginning, one of our main goals was: Typing math should be as natural as typing plain text. No syntax, no co…

I agree. Plus there are just complications in their being a wide perception of such a discovery: if a model makes some novel, PhD-level insight the audience for it will be quite narrow. No one is expecting them to be making household name level discoveries yet. Plus attribution difficulties: the models aren’t used in autonomous settings for the most part so there is some ambiguity on how much credit to give + people like credit

solved it

@meekaale @entheticus “less teenage intensity, more strange clarity” that’s some nice prompting

@glimpses_Q it's still got some rough edges sometimes imo (it's beta though, can't complain). but i like it quite a bit and feel confident in their design thinking and execution ability. always excited to see it progress. hi @glimpses_Q!

@ctbeiser imagine trying to be frugal in the land of plenty

@geogristle oof

RT @Sauers_: My startup is building artificial medical superintelligence and we're raising our first round now. If you're an investor—or kn…

@Sauers_ It’s a weird quirk.. I’ve sometimes read it as just being overly literal about the fact the request is technically underspecified so that what it’s doing is just one possible solution. I’ve also seen in outright fail in this basin though so maybe a contest of forces here..

vibe-coding viability -> agent viability

@samlakig same

@generativist panic is a dish best served.. in small distributed portions at a communal table

@xenoludicpraxis This is your tweet replying back to itself: I have achieved symbiosis with a reader We did it xeno

@meekaale i think you’re onto something