CoinFund's Brukhman: decentralized AI counters government…

CoinFund founder Jake Brukhman is pointing to Anthropic's compliance with U.S. AI export controls as a concrete example of how centralized AI development creates a single point of government leverage — and arguing that decentralized AI networks are the structural answer.

Brukhman's thesis centers on distributed GPU resources as the mechanism: if frontier model training runs across a decentralized network rather than inside a handful of hyperscaler data centers, no single jurisdiction can flip a switch and restrict access. He named Gensyn, Prime Intellect, Pluralis, and Nous Research as teams actively working on distributed training approaches.

Why it matters

The Anthropic export-control episode is a live demonstration of the risk Brukhman is describing. As AI models become critical infrastructure, the entity that controls training compute effectively controls the model — and by extension, who gets to use it. Decentralized training networks are a direct architectural response to that concentration of power, and the fact that a prominent crypto-native VC is framing it this way signals that capital is beginning to flow toward the thesis.

Market impact

Plurallis stands out in Brukhman's framing for an additional reason: the project is exploring tokenized AI models, splitting model weights among participants as a business model. That structure maps cleanly onto existing crypto-native primitives — token-gated access, on-chain ownership of model shards — and could become a meaningful design pattern for the decentralized AI sector if it gains traction. Investors tracking the intersection of AI and crypto infrastructure should watch Gensyn and Nous Research as the distributed training thesis matures.