Bitcoin Principles Applied to AI

At a recent Bitcoin party I expected to talk money. Instead, the conversation drifted — almost on its own — to AI. And the more it moved, the more I noticed the destination. A lot of the principles that make Bitcoin principled are the same principles AI should be built on. Not by accident. Because Bitcoin is really the clearest working example of a system that functions without a trusted central authority, and that is the part that actually transfers.

Most of what makes Bitcoin sound good is one idea wearing different hats: remove the trusted middleman, make the system verifiable, and govern it with incentives instead of authority. Run that same logic through AI and you stop renting a brain from a datacenter and start owning one.

Here are the three that carry the weight for me.

One: Self-Sovereignty — You Control the Hardware, the Software, and the Costs

"Not your keys, not your coins." The AI version is "not your weights, not your model."

When you run AI locally you own the whole chain, end to end:

• The hardware. Your machine, your power, your uptime. Nobody pulls the plug, nobody rate-limits you, nobody "deprecates" your access.
• The software. Open weights and an open harness you can read, tune, and swap. You are not subject to a changelog you did not ask for.
• The costs. Inference costs electricity, not per-token API fees that a vendor can quietly change overnight. The cost curve is yours to manage, not someone's pricing page.

The API model is a bank: someone else holds the keys and can freeze the account or rewrite the terms. Local inference is self-custody. This is the pillar I practice most — my own local models, my own memory, my own tools.

Two: Privacy — The Way Bitcoin Protects Transactions

Bitcoin lets you transact without revealing who you are or what you are up to to a third party. A local AI does the same thing for your thinking.

Your prompts, your data, your questions never leave your network. No provider is logging what you asked, no model is quietly trained on your material, no middleman is monetizing your curiosity. Privacy here is not a feature you bolt on after the fact — it is the architecture. It is the AI twin of Bitcoin's pseudonymous ledger: privacy by design, not by policy.

Three: Decentralization — Like Napster

The model that finally made this click for me was Napster.

Before Napster, music flowed through a handful of gatekeepers. Napster let the people who had the thing share it with the people who wanted it — and the network itself became the product. Decentralized AI works the same way. The value lives in the network of people each running their own stack, not in one company's datacenter. The more of us run our own sovereign models, the harder it is for any single vendor to be the gatekeeper.

"Napster" is my name in F3 — and it speaks to who I am. I genuinely don't think the guys who named me knew the full impact of what they had done. Years later it lands exactly right. The person who runs his own sovereign AI stack, who shares open weights instead of hoarding them behind a paywall, is doing for computation what Napster did for music. The name wasn't a joke. It was a description.

The Supporting Cast

These three are the headline, but a few more principles ride along and reinforce them:

• Verifiability. Open weights are the AI twin of a public, re-checkable ledger. "Here is the proof," not "trust me."
• Spontaneous order. A single giant model in one datacenter has the same knowledge problem Hayek warned about in central planning — it cannot hold the dispersed, local context that a network of sovereign stacks can.
• Censorship resistance. There is no vendor standing between you and the inference.
• Open source. You cannot audit what you cannot see.

Where the Analogy Breaks (and Why That Matters)

I would not trust a writer who skipped this. The transfer is real, but it has edges:

• Bitcoin moves value; AI does computation. The overlap is at the level of governance and architecture, not function. The moment this curdles into "AI should live on Bitcoin," it has left principle behind for tribalism.
• Scarcity does not map. Model weights are copyable and non-rival — there is no 21-million cap. The honest analog is "don't let the vendor quietly retrain the model under you." A silent retrain is what a central bank changing the rules is, for a model.
• Bitcoin's security is economic — it is rational to be honest because you are paid to be. AI agents do not have that substrate built in. You have to design the incentives yourself, and that is the genuinely hard part of autonomous agents.

The Through-Line

Strip it to one line and it is the same sentence Bitcoin and a sovereign AI stack both say: no trusted middleman, everything verifiable, and incentives doing the governing instead of an authority. You can be half-built into that stack before you know it — the local models were the obvious part. Verifiability and incentive design are the frontier.

I have a handle in F3 that I used to think was just a funny name. I no longer think it was a joke. I think it was a forecast.