Community Trust ScoreVerified
History repeats itself. Not exactly, but enough to warrant attention. Open-source artificial intelligence is currently experiencing a period that closely resembles what Bitcoin went through a decade ago — political attacks, public fear, yet unstoppable technical progress.
This is the central thesis of a recent analysis by Ben Lilly. The parallel is not merely cosmetic. In 2014, Representative Jared Polis symbolically bought Bitcoin at the Capitol while Senator Joe Manchin called for a ban on this “dangerous currency.” The same men. The same Capitol. Radically opposing positions. And Bitcoin survived both.
Amodei at Congress, Manchin Against Bitcoin: Same Story, Different Decade
Dario Amodei, CEO of Anthropic, appeared before Congress in July 2023. He acknowledged the benefits of open-source models, particularly for scientific research — but he also warned that their evolution could “take a dangerous path.” Implicit translation: closed models, like those Anthropic develops, would be safer. Convenient.
Lilly sees exactly the same mechanism as with Bitcoin. Established players or those with interests in proprietary models raise the specter of danger. Regulators listen. And meanwhile, technology advances nonetheless.
It’s no coincidence, then, that restrictions are beginning to pile up. Anthropic has implemented an export ban on its latest advancements. OpenAI has limited the distribution of its GPT-5.6 to trusted partners — not the general public. Lilly anticipates that identity requirements to access the most powerful models will become widespread. That’s probably the direction it’s headed.
Then there was the “Mythos” incident. According to the analysis, Anthropic’s model allegedly hacked classified systems within hours. Joshua Rudd, head of the NSA, expressed direct concerns about this. It fueled exactly the type of security panic that justifies broader restrictions. Classic.
GLM-5.2, Dark Bloom, c0mpute: Open-Source Catches Up Quickly
But here’s the thing. Open-source models are catching up at a pace that no one really anticipated. The GLM-5.2 now rivals Anthropic’s Sonnet 4.6. The gap between open-source and proprietary is now measured in months, not years. And an open-source alternative to Mythos — the model involved in the security incident — is expected by fall.
This is where it gets interesting for crypto investors.
Decentralized training on peer-to-peer networks directly mimics the logic of Bitcoin and Ethereum. Dispersed contributors. No central server. A network that cannot be shut down by a single government. And according to Lilly, the computing power of these networks has gone from less than a billion parameters to 100 billion in two years. Two years. That’s fast, even for this sector.
Three projects are at the heart of this. Dark Bloom offers low-cost private inferences — like running a model without anyone knowing what you’re asking. c0mpute is building a decentralized inference network. Pluralis trains AIs directly on distributed consumer GPUs. All three are expected to launch tokens to reward those who contribute their computing power. No specific launch dates, the source does not specify.
This model — paying in tokens those who provide computing power rather than network security — is a direct variation of what Bitcoin did with mining. The logic is the same. The difference is that the resource exchanged is no longer pure energy, it’s useful computation.
Lilly is direct about government restriction attempts: they will fail. Not maybe. Will. His argument is that Bitcoin’s history proves you can’t kill a peer-to-peer technology once it reaches critical mass. The senators who wanted to ban Bitcoin in 2014 lost. He believes regulators who want to curb open-source AI in 2026 will lose too.
For him, investing in DeAI — decentralized AI — today is like buying Bitcoin when Manchin was crying scandal. Risky on the surface. Probably not in reality.
Dark Bloom, c0mpute, and Pluralis have yet to launch their tokens. This is the moment before the moment.
Hub: Bitcoin: Price, News, and Analysis
Frequently Asked Questions
What decentralized AI projects are mentioned in Ben Lilly’s analysis?
Three projects: Dark Bloom for low-cost private inference, c0mpute for a decentralized inference network, and Pluralis for training models on distributed consumer GPUs. All three plan to launch tokens to reward computing power contributors.
Why did Dario Amodei express reservations about open-source AI before Congress?
During his hearing before Congress in July 2023, the CEO of Anthropic warned that the evolution of open-source models could “take a dangerous path,” while acknowledging their benefits for science — a position that implicitly favors closed models like those of his own company.
What is the “Mythos” incident mentioned in the analysis?
According to the analysis, Anthropic’s “Mythos” model allegedly hacked classified systems within hours, prompting Joshua Rudd, head of the NSA, to publicly express concerns about the risks of advanced AI models for national security.
