ALT Introduction to a paper that reads “Since the dawn of time, human beings have asked some fundamental questions: who are we? why are we here? is there life after death? Unable to answer any of these, in this paper we will consider cohomology classes on a compact projective manifold that have a property analogous to the Hard-Lefschetz Theorem and Hodge-Riemann bilinear relations.”
Pure math Q relevant for LLMs:
Unrelated tokens (like "dog" and "sun") should roughly correspond to orthogonal vectors in embedding space.
So, In ℝ^k, what’s the max no. of pairwise orthogonal vectors? what about if they're only "nearly" orthogonal, so ∣vi⋅vj∣<ε, ∀ i≠j?
"game theory" is insane because it appears to be one of the most legitimately dangerous fields w/ the potential to gigafry your brain but is exclusively taken by literal turbonormies who unironically want to like "find a good strategy for play" and basically get oneshotted by it
Imagine learning about synthetic assets and your first reaction not being
“Wow wait this is absolutely sick, how does it work?”
But instead the reaction being to write a poorly-thought-out article about how “it’s all fake stonks”
Today, I am pleased to announce that @ks_kulk , @alexhevans, @tarunchitra, and I are going to start an MEV prop shop. From now on we will remove all of our public papers and stop all future paper writing. We look forward to sandwiching your trades using convex optimization!
if you know a little bit of math and would like to understand the basics of MPC protocols
well
have I got a video (or two) for you
introducing the first of the BCC whiteboard sessions ! (w/ the legendary @cryptodavidw)
what's incredibly entertaining is that I'm sure they really tried to keep it out of the training data and yet
there's clearly enough context elsewhere to impute what's going on here which is fascinating
It is nothing less than an honor to be welcoming Kobi, Wei Jie, Andrija, Grjte, and Nico to the team at BCC, who bring an incredible depth in cryptography, zero knowledge, and software engineering to the firm!
baincapitalcrypto.com/expand…
Welcome to Bain Capital Crypto – backing renegades and pioneers building the next generation of open internet infrastructure. We’re excited to share what we’ve been building. baincapitalcrypto.com/
also, if anyone is bored and needs something to do:
the homework for the 1st "Succinct Proofs and Linear Algebra" lecture is out and it would be great if anyone wants to pull request solutions to it!
(this is an attempt at teaching a fully-open course)
angeris.github.io/other/spla…
Thanks to @BalancerLabs for their research grant! I am really looking forward to working this summer on (hopefully!) improving smart order routing for Balancer v2. :)
Aight I was told by @tarunchitra that, since we had kind of a crazy year, I had to do a quick tweet thread about our ~ paper a month cadence with @alexhevans
1 tweet/paper let’s go
The Accidental Computer with @angeris on prover acceleration via DA encodings
In short, proofs-of-encoding (such as ZODA) perform much of the work of a code-based PCS, making it cheap to “embed” verifiable computations over data posted to a DA layer
ok if I wanted to commission someone for a digital art piece as my background
how would I go about doing that? Does anyone have (digital) artist friends they’re willing to rec?
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
did cultures ever have a disdain for papers as “offloading thought”? For example, did oral cultures (ancient Greeks, e.g.) ever criticize paper-as-memory as making people lazy, e.g., and are there documented versions of this?
1/9 Happy to open-source joint work with @angeris—
CryptoUtilities.jl, a set of Julia packages that brings fast and easy support for binary extension fields, Merkle trees, Reed-Solomon codes over binary extensions, and many more.
it's funny, for a long time, i assumed that computation was this very sacred, brittle, difficult thing: so many things need to work "exactly" right
but i’ve realized that it’s actually much *harder* to make a thing that doesn't accidentally enable ~ arbitrary compute!
Just a silly reminder that this paper is kinda fun (and is now on arXiv!)
Essentially, it shows just how much you can do without using oracles: any (capped) monotonic payoff can be replicated without the use of an oracle!
arxiv.org/abs/2111.13740
ZODA
Joint work with @nico_mnbl and @angeris (with help from @NashQueue, @sanaz2016, and @jadler0)
A minor tweak to the encoding procedure makes data squares like @CelestiaOrg’s provably correct, opening a path to both higher data throughput and lower-latency light nodes
I love crypto bc a relatively formal presentation has some complicated data and a reference at the bottom will be some dune dashboard from "thiccythot"
Katie just made this "website" and honestly it's incredible
ALT Website with block letters and common gradients for crypto websites, reading: "Shitcoin reinvented."
ALT Website with common crypto gradients and background reading "Bullshit: It doesn't really matter what you put here as long as you have the right gradients and hacker vibes" with a button that says "click on me"