New batch of Thought seriesBrownstone Research Newsletter Written by Ben Lilly, it argues that the battle is over open source artificial intelligence It follows the same path that Bitcoin did a decade ago, and investors who recognize this pattern stand to profit.
The memo begins with testimony given by Anthropic CEO Dario Amodei to Congress in July 2023. Amodei acknowledged that open source is a “good thing” in most scientific fields and that the risks of open models released so far have been “relatively limited,” but he warned that scaling up open source models is headed “down a very dangerous path.”
Lilly reads the implication clearly: If open models are dangerous, then the closed models sold by companies like Anthropic are the safe choice—and the policy that follows is to restrict what is open and elevate what is closed.
Early Bitcoin skeptics reflect what artificial intelligence is up against
This is a framework that digital asset investors know well.
It revisits early Bitcoin skeptics, from Rep. Jared Polis buying the first Bitcoin on Capitol Hill in 2014, to Sen. Joe Manchin’s call to ban the “dangerous currency,” through 2023 accusations that regulators tried to cut off cryptocurrencies from the banking system in what critics called “Operation Choke Point 2.0.”
He points out that the industry has survived, and that Washington is now moving toward clearer rules through… Pass The law of genius On hold The law of clarity.
Decentralized artificial intelligence, which Lilly calls “DeAI,” is now fighting the same battle. He points to recent developments as evidence that the fences are rising: a US export ban on the latest version of Anthropic, which he says will push the company toward permissioned access that verifies a user’s identity before granting a form, and OpenAI’s decision to limit the rollout of GPT-5.6 to trusted partners.
Identity requirements are expected to spread. “It’s for your protection, you see,” he writes. “It always is.”
The memo relies on a national security narrative to explain the fear driving these movements. Lilly cites NSA chief Joshua Rudd, via Senator Mark Warner, describing how the Anthropic “Methos” model “penetrated almost all of our secret system, not in weeks, but in hours.”
However, open source fills the gap, according to the article. Lilly says the latest GLM-5.2 scored at par with Anthropic’s Sonnet 4.6 as of February, leaving open models about three to four months behind the border, and it expects to have an open competitor for the Mythos and GPT-5.6 by the fall.
He sees the biggest solution as decentralized training on peer-to-peer networks that mirror Bitcoin and Ethereum — trading the security of network computation for training the computation for the model. He points out that distributed training has grown from less than a billion parameters to 100 billion in just two years.
He named three early projects — Dark Bloom, which enables low-cost private inference on idle Macs; c0mpute, a decentralized inference network; and Pluralis, which trains AI via distributed consumer GPUs — expects more token launches and rewards users for contributing to computing.
The note ends with the idea that governments will try to ban open forms and will fail. For him, investing in the space “would be like buying Bitcoin in 2014, when it was still risky.”
