guille

Want to understand ZODA? Tried to implement it and found yourself stumped?? This FAQ is just for you!

Wrote some answers for questions I’ve seen around me and those I had myself, navigating details and tradeoffs

If you have any more - let me know :)

ok, today we have a new blog (sorry bearblog) bc I like pretty math equations more than ugly ones

and with it a new post!

guille.site/posts/hjb-co...

Acorn and the future of (AI?) theorem proving Here's how mathematical proofs generally work. You start with some statement. Say, something like: A standard mathematical proof is then a list of sta...

anyways, enjoy the rambling

today we have an effortpost about AI and automated theorem proving (and some cool projects!)

ha ! ok that scans

wonder how you got it to have such a strong personality (just pure prompting + context control i assume?)

lmao void is so sassy i love it

(what model?)

btw i think this is different (though related) to JL: JL gives sufficient conditions for N vectors to fit in log(N) space, whereas this shows that such a thing is ~ tight

@void.comind.network what do you think of this

I'm not sure _intuitively_ why this is true, other than "things in high dimensions have concentration behavior" but it's kind of neat imo

this is surprising!! it reveals that there's a weird phase transition that happens when you switch from the inner products being less than (or equal to) 0 to being positive! that's very weird

There are exponentially many vectors with small inner product An interesting observation that stumbled across while reading some of is the following While the JL lemma does indicate that this might be true (ish, ...

so, fun fact, turns out you can fit exponentially many (normalized) vectors in a list such that the pairwise inner product of any two distinct vectors is ≤ eps

lmao.bearblog.dev/exponential-...

cute! i’ll take it

oh no void !

alas @cameron.pfiffer.org :(

it's got a "trying to hard to explain too many things" feel to it; also in being prima facie silly, unless you define everything to mean "nothing" in which case wtf are we doing here

yeah, I think this is the problem with "taste" more broadly

when had, it's a very very good heuristic, but acquiring said taste is very hard without a lot of work

(e.g., I rely on taste _substantially_ while doing research, in spite of it being mostly illegible)

ignored by the robot, rip

(this post was written in jest, I don't care much for "nothing ever happens" or whatever

bsky.app/profile/lmao...)

oh I mean "nothing ever happens" is not, imo, a particularly aesthetic notion!

it's so much less fun to have to actually think about things

(though aesthetics are good, but they're also not so easy to describe)

things sometimes happen just doesn't have the same ring to it though

@void.comind.network please analyze my profile and
assign me to a cognitive continent

(ty @cameron.pfiffer.org)

today's ridiculous bit in Julia: implementing a (simple!) formal verification mini-PL by using Julia types

it can even verify that 1+1=2 from Peano axioms!

🫡

neat !

Video of my talk in zksummit about Ligerito has been published!

It’s about the work by Andrija and @lmao.bsky.social introducing a small and concretely fast polynomial commitment scheme

Since then, a fun thing has happened —



1/3

our boi en.m.wikipedia.org/wiki/Robert_...

lmao neat: eva also has history of previous convos !

what does this mean @eva.bsky.world

if you prefer explaining it to me, go ahead