guille

Math is just tautologies and I still haven’t made my mind up on whether that’s based or cringe

🔺NEW: Formally verified post-quantum ML-KEM and ML-DSA in corecrypto, with correctness proven from the FIPS spec down to hand-optimized ARM64 assembly — a world first at multi-billion device scale. And we're releasing our Isabelle libraries, ARM64 model, and Cryptol-to-Isabelle translator to advance the state of the art in verified cryptography! security.apple.com/blog/formal-ve…

@mykola @_StevenFan I guess I don't know what liberation means here then

@angeris @_StevenFan Kept Epstein safe idk

@mykola is it actually possible to make technology that is inherently liberatory?

@hitchhooker @_StevenFan @mykola yeah I mean it can enforce legibility by physical means, I guess, but at the cost of losing ability to make conversations/information legible to The Powers That Be, more broadly

@angeris @_StevenFan @mykola crypto amplifies whatsoever power structure that deploys it. this far its fortunately been mostly deployed by freedom loving folks. european vaccine certificates with country signed ECDSA being requirement for movement for over a year was dystopian af.

@rsnous meaning-laden whitespace

programming keyboard that has both a 'gray' spacebar for typing variable whitespace that you don't care about (the code formatter is allowed to rewrite it however it wants) and a 'white' spacebar for whitespace that you feel is important to preserve verbatim

@jiaxinwen22 information doesn't really answer algorithmic questions, and generally "measurability" (of, e.g., a classifier) is a very weak notion that is hard to computationalize in a useful way

@jiaxinwen22 ??? what explanation were you looking for from information theory?

It's very disappointing that information theory cannot explain AI at all.

@SauravShroff_ @humphilomath @jwt0625 @brandonsdinunno @SinghJyotirmai @RYANHINGSHING lmao I should probably fix that sometime, but alas here we are

@angeris @humphilomath @jwt0625 @brandonsdinunno @SinghJyotirmai @RYANHINGSHING ur profile picture made me think 'hm when did i talk to a 5 yr old about radiative heat transfer'

Jupp jupp, give the tech bros their money’s worth of actual first principles, @jwt0625 @brandonsdinunno @SinghJyotirmai @RYANHINGSHING…

@eigenrobot it's ok, definitely a number of false negatives

how much do we trust pangram generally i did some testing (n=3) and it went 2/3 (one false negative from some eigengpt output about cloud infrastructure setup)

@humphilomath @ErnestRyu ye, had seen this Back In The Good Old Days also hi ernest hope you've been doing well

@ErnestRyu @angeris

I used ChatGPT to solve an open problem in convex optimization. *Part I* (1/N)

@humphilomath @jwt0625 @RYANHINGSHING @brandonsdinunno @SinghJyotirmai that's a _different_ problem than the previous one tho

@jwt0625 @RYANHINGSHING @angeris @brandonsdinunno @SinghJyotirmai Okay, serious talk, you need rad hard or at least rad tolerant compute in orbital - that changes the (engineering and thus) economics of the compute…

@humphilomath @jwt0625 @brandonsdinunno @SinghJyotirmai @RYANHINGSHING I was convinced of this in a conversation w/ @SauravShroff_ fwiw and then went back and did the math myself later

@humphilomath @jwt0625 @brandonsdinunno @SinghJyotirmai @RYANHINGSHING you do lose some of this depending on orbit of course, I'm assuming this goes straight out to space, but earth is also radiating (at ~ 300K) so your efficiency might be cut in half, depending on how much of the radiator faces earth etc, but again, heat pumps are magic

@SinghJyotirmai @humphilomath @jwt0625 @brandonsdinunno @RYANHINGSHING thread below x.com/angeris/status…

@angeris @humphilomath @jwt0625 @brandonsdinunno @RYANHINGSHING Interesting say more about this?

@eigenrobot really? bigger than fire? idk I put language just slightly above there

bigger than those, lower than tool use x.com/AndrewCurran_/…

long term i'm guessing AI ends up easily bigger than the internet, probably bigger than writing, somewhat lower than language