WuBlockchain
CoinFund founder Jake Brukhman argued that decentralized AI networks are emerging as a structural counterweight to the centralization of frontier model development, pointing to Anthropic's compliance with U.S. AI export controls as evidence of the political risk baked into single-jurisdiction model control.
Speaking on the centralization trend across the AI sector, Brukhman said distributed training over pooled GPU resources lets frontier models be trained outside the reach of any one government's licensing regime. He cited Gensyn, Prime Intellect, Pluralis and Nous Research as teams actively building distributed training infrastructure — and flagged Pluralis specifically for exploring a business model that splits model weights among token holders, effectively tokenizing the model itself.
Why it matters
The argument lands at a moment when U.S. export controls are reshaping which companies can ship frontier weights to which countries. If a single firm's compliance decision can effectively gate a foreign customer's access to a model class, the locus of control over that capability sits inside one legal jurisdiction. Distributed training and weight-splitting re-introduce jurisdictional ambiguity at the infrastructure layer — the model doesn't exist in one place because the training and the weights don't either.
Market impact
For crypto, the read is direct: the teams Brukhman named sit at the intersection of GPU marketplaces, tokenized compute and tokenized models — categories that have been drawing capital as the AI narrative matures onchain.
Frequently asked questions
-
Who is Jake Brukhman and what is CoinFund's position on decentralized AI?
Jake Brukhman is the founder of CoinFund, a crypto-native investment firm. He has argued publicly that decentralized AI networks — built on distributed GPU resources and tokenized model weights — are a structural counterweight to the centralization of frontier model development.
-
Why did Brukhman cite Anthropic's compliance with U.S. AI export controls?
He used it as evidence that single-jurisdiction model control carries political risk: when one firm's compliance decision can effectively gate a foreign customer's access to a model class, the locus of capability sits inside one legal regime. Distributed training, he argued, re-introduces jurisdictional ambiguity at…
-
Which teams is Brukhman pointing to as building decentralized AI?
He named Gensyn, Prime Intellect, Pluralis and Nous Research as teams actively working on distributed training approaches, and flagged Pluralis specifically for exploring a business model that splits model weights among token holders.
-
What does "tokenized AI models" mean in the Pluralis context?
It refers to a business model in which ownership or operational control of a model is distributed across token holders by splitting the model weights among participants, rather than having the weights held by a single entity.
-
How does this story connect to crypto markets?
The teams Brukhman named sit at the intersection of GPU marketplaces, tokenized compute and tokenized models — categories that have been drawing capital as the AI narrative matures onchain. Material raises or mainnet launches from any of them would harden the decentralized-AI thesis.
TheBlock
Crypto News
CryptoSlate
CoinDesk
WatcherGuru