Jake Brukhman

Founder & CEO @coinfund. Engineer. Investor. Digital network expert (blockchain, AI). Champion of the builders of frontier tech. (Not investment advice.)

[Jake Brukhman @jbrukh]

My prediction is that we’re about to enter a decentralized AI training race.

[Jake Brukhman @jbrukh]

Fortune: compute will eclipse oil, $100T futures market.

Also Fortune, three paragraphs down: contracts miss the settlement price by 10% two days before expiry.

That might be the opportunity for those who can build it.

[FORTUNE @FortuneMagazine]

According to Kalshi CEO Tarek Mansour, compute—a term that describes the chips and electricity powering the AI revolution—will eclipse oil as the world’s most valuable commodity, and spur a futures market for hedging it. https://t.co/0FzzNOTmuY

[Julian @_julianma]

EtherFi team is killing it.

It's been very hard to convince my friends and family to use crypto apps because of crypto's bad image, even if apps have great features.

Account protection and the investment and borrow options EtherFi has move the needle.

[Jake Brukhman @jbrukh]

NVIDIA releasing Nemotron 3.5 Lightning, a 30B MoE with only 3B active parameters that can run locally on a single GPU.

That is, AI is becoming dramatically more deployable.

Open models + sparse architectures + commodity/local compute will push useful intelligence out of hyperscale datacenters and into enterprises, edge devices, sovereign infrastructure, and decentralized networks.

The long-term AI stack will be much more heterogeneous than “a few giant models behind a few giant APIs.”

https://t.co/2f1jRiKy9o

[Jake Brukhman @jbrukh]

It’s truly the greatest era of blockchain ever.

[Bull Theory @BullTheoryio]

BREAKING: 🇺🇸 The SEC is preparing to allow US stocks to trade 24/7 on the blockchain.

[Jake Brukhman @jbrukh]

I am publishing my GPT-5.6-assisted proof of a dense-case theorem for Seymour's Second Neighborhood Conjecture.

Intuitively, the Conjecture says that it in every finite oriented graph, there is always at least one vertex whose "friends-of-friends group" is at least as big as its "friends group".

While this Conjecture is quite hard to prove in general, it is known for special classes of graphs. In particular, Fisher proved it for tournaments (oriented graphs which are maximally dense with edges), and others like Fidler, Yuster, Ghazal have proved it for slightly less dense graphs that are "tournaments minus some structure".

This proof is a clean, simple density condition which is less dense than a tournament but also doesn't require any structure on the missing edges.

This proof is currently in the process of being verified.

Finally, a note about how it was found. In the course of trying to direct GPT-5.6 to find a counterexample to SNC, we were partitioning the edges of the graph into a source vertex, its first and second out-neighborhoods and the set of edges unreachable from the source vertex. This set off a lightbulb that we could count mutually unreachable vertices two different ways. GPT-5.6 then found enough boundary conditions to cause a contradiction using these two counts if we assumed that our graph was a counterexample.

Overall, this exploration was very hands-on and I directed the AI through many different strategies before we found this angle. I am very bullish on the future abilities of AI to explore mathematics!

(Shoutout to @johnennis for help with getting this out.)

https://t.co/jV3LB61ceZ

[Jake Brukhman @jbrukh]

That twist is what use to be called by decentralization maximalists a centralization bottleneck.

[Laura Shin @laurashin]

🌍 Coinbase just landed a license to run tokenized stock offerings out of Abu Dhabi, with a twist: every token can be frozen and seized at the wallet level under regulator oversight.

https://t.co/cXKx28n5TZ

[Jake Brukhman @jbrukh]

@chloeallegra228 Seymour-tight finite oriented graphs.

[Abril Zucchi @abrilzucchi]

you can indeed dm your way into changing the trajectory of your life

you’re always just one dm away from a new job, a future partner, a new friend, or an opportunity you never saw coming

[Jake Brukhman @jbrukh]

Training as an activity is proliferating.

[Naval @naval]

People who are serious about software train their own models.

[Jake Brukhman @jbrukh]

@DennisonBertram @klimonchain Fun fact there is no Nobel Prize for math because math is useless.

[Sasha Barrie @Kapuster]

"Save the em-dash" is a real movement. I think it has a chance.

[Jake Brukhman @jbrukh]

@DennisonBertram Have you tried @grok 4.5?

[Jake Brukhman @jbrukh]

This really sucks for some people.

[Greg Brockman @gdb]

the bottleneck is increasingly knowing what you want

[Miss Sentient @0xsachi]

Has anyone thought abt implementing rem sleep for ai models?

[Jake Brukhman @jbrukh]

It's fascinating to watch AI video artists come up with some new original aesthetic only to have it be instantly "distilled" by an army of competitors flooding the attention markets and, in effect, causing a kind of eyeball fee erosion.

[Jake Brukhman @jbrukh]

@socrates1024 @shawmakesmagic Appreciate you trying, thanks

[Jake Brukhman @jbrukh]

Anyone here can endorse me for arXiv? (Trying to post https://t.co/9plgikZzOS.)

[Jake Brukhman @jbrukh]

@ldf_gm iykyk

[Jake Brukhman @jbrukh]

What does this mean for Coldcard?

[Jake Brukhman @jbrukh]

@joshephraim

[Jake Brukhman @jbrukh]

If you’re into writing air-gapped memos to your friends, you can buy a typewriter on Amazon for $269.