zkpixels

building the next internet: a decentralized "verification os" stack for Web3—a zero-knowledge and cryptography layer unlike anything that's existed.

[zkpixels @zk_pixels]

if you do post-quantum crypto, especially for web3, come build with us. post your work on #aptpr and let's get a real community going. @pqabelian @QRLedger @VitalikButerin and anyone working on this, the door's open.

#PostQuantum #PQC #Web3 #Cryptography #Crypto #Blockchain #Ethereum

[zkpixels @zk_pixels]

new paper, just posted on aptpr 👇

the question: can a shor-style quantum attack actually get any traction on lattice signatures, the LWE-based kind like ml-dsa (dilithium)?

we built an exact tool to measure it.

https://t.co/py0Bc30PfW

#aptpr

[zkpixels @zk_pixels]

two quick notes: we posted this on https://t.co/fBSsaZFBpQ, the spot we're building for putting research out in the open.and it kicks off our own post-quantum crypto research line. more soon.

[zkpixels @zk_pixels]

getting close now. the core of our free multi-llm harness is stable, and we're spending this week making sure the installers actually work everywhere. shipping soon.

#AI #LLM #AIAgents #AgenticAI #GenAI #MachineLearning #BuildInPublic #OpenSource #DevTools

[zkpixels @zk_pixels]

ok so we've secretly been building our own ai harness. one interface for working with a bunch of llms at once, instead of juggling them all separately. we're releasing it for free. more details soon.

#AI #LLM #AIAgents #AgenticAI #GenAI #MachineLearning #BuildInPublic #OpenSource

[zkpixels @zk_pixels]

been heads-down on pixelvm, a new post-quantum zero-knowledge vm. can't say much about how it works yet, but the security target is public: 160-bit pq, with a guaranteed 128-bit pq floor. more to come.

#zk #crypto #web3

[zkpixels @zk_pixels]

@Nicole_Nobili every step is machine-checked: python runs the enumeration and the witness, lean re-derives the finite certificate. nothing to take on faith.

go read it, poke holes, post your own 👇

https://t.co/4Wr71p9rwd

#zarankiewicz #math #aptpr

[zkpixels @zk_pixels]

how many 1s fit in a 9×23 grid of 0s and 1s before some 3 rows and 3 columns are all 1s where they cross?

the answer is 103 🔲

new post live on #aptpr — check it out 👇

https://t.co/4Wr71p9rwd

#combinatorics #math

[zkpixels @zk_pixels]

we closed it from the top: a short, fully finite proof that 104 is impossible. so 103 is exact, and that construction was already optimal.

count row-triples to constrain the column heights, which leaves only 3 possible profiles for a 104-one matrix. then a mod-3 parity count kills all three.

[zkpixels @zk_pixels]

@AnthropicAI @ClaudeDevs pls hurry and bring Fable 5 back

[Theo - t3.gg @theo]

3 days left of using Fable in your Claude Code sub! Better maximize that token usage

[zkpixels @zk_pixels]

the deep question is whether these sets are always "big." over the reals in 3d that's the kakeya conjecture — the one hong wang & joshua zahl proved in 2025.

we did the finite-field version in F₃³ and pinned the exact minimum: 13 points.

#kakeya #math #aptpr

[zkpixels @zk_pixels]

what's the smallest possible kakeya set in F₃³ — a set that contains a line in every direction?

the answer is 13 🪡
our first ever post, live now on https://t.co/fBSsaZFBpQ — a new place to post research. come hang 👇

https://t.co/4M9DJngTPF

#math #aptpr

[zkpixels @zk_pixels]

quick backstory: a kakeya set is a set containing a line in every direction.

it goes back to the 1917 kakeya needle problem — what's the smallest area you can spin a needle around in? the wild answer: as close to zero as you want.

[zkpixels @zk_pixels]

The thing we've been building uses zero-knowledge proofs in a way we haven't seen done before. That's all we'll say for now. Stay tuned. ⚡🔐

[zkpixels @zk_pixels]

fable 5 quietly had the frontier math sauce and my research went stupid fast, then it vanished mid-cook. i'm not loyal to any lab, i just use whatever's best. right now that's fable 5. still waiting. want to get back to work. @AnthropicAI