1178 signatures. CEOs. Chief Scientists. The builders of the most powerful AI systems are now asking for a slowdown.
Not a temporary halt. Not a voluntary pledge. They demand an international mechanism with teeth—a binding framework that can press pause on frontier development when the risk exceeds our capacity to control it.
I read the letter from the floor of my Washington DC office, a coffee mug cold beside me. The irony wasn't lost on me. Here are the architects of the digital future, pleading for a covenant stronger than code. Here is the pinnacle of centralized innovation admitting that its own acceleration is unsustainable.
And I thought: in crypto, we've been telling ourselves for years that code is law. That smart contracts replace trust. That decentralization solves the coordination problem.
But when the architects of the most centralized technology on Earth ask for a pause, aren't they really asking for what blockchain was meant to provide? A transparent, programmable, hard-coded covenant that binds all participants—not through moral appeal, but through cryptographic enforcement?
Tech changes. Values remain. And the values behind this AI letter are exactly the values we claim to champion: safety, sovereignty, and shared fate. The only difference is they are trying to build the covenant after the code is already deployed.
We have the chance to build it first.
The Context: A Prisoner's Dilemma, Now in Public
The letter is not a fringe petition. It is signed by the top technical leadership at OpenAI, Anthropic, Google DeepMind, Meta AI—the very companies locked in a blistering race to scale. The core argument is stark: frontier models may soon become capable of autonomously conducting most AI research. That is a threshold. Once crossed, recursive self-improvement becomes plausible, and the speed of development outstrips our governance.
But the letter also admits what every crypto founder knows intimately: no single company can slow down voluntarily. The competitive disadvantage would be immediate. Market share would vanish. Talent would flee to the firm that kept its foot on the accelerator.
This is the prisoner's dilemma of innovation. And the solution they propose is a state-led international mechanism—a top-down institutional fix.
Now, let me be clear. I spent the 2017 ICO boom auditing 150 whitepapers, searching for mission statements disguised as technical documentation. I wrote a 40-page thesis called "Code as Covenant." I know a good story when I see one. The AI letter is a powerful story: We, the creators, need external restraint because our own incentives are misaligned.
But the solution is not more trust in governments. It is more trust in programmable commitment. In smart contracts that enforce rules without human hesitation. In DAOs that coordinate safety decisions across competing entities. In oracles that feed real-world risk metrics into on-chain governance.
We have the tools. They have the problem. That is the opportunity.
The Core: From AI's Institutional Plea to Crypto's Programmable Covenant
Let me break down the tension using something we understand: the same problem exists in crypto, and we haven't solved it either.
Consider Layer2 scaling. We now have dozens of rollups, validiums, and sidechains—each promising to scale Ethereum. But the user base remains largely the same. We aren't scaling; we are slicing already scarce liquidity into fragments. The competitive race for TVL and transaction volume mirrors the AI race: no single Layer2 can slow its incentives to attract capital, because the next chain will simply capture the flow.
The result? A fragmented ecosystem that weakens network effects and reduces overall resilience. Yet we call this "healthy competition."
Now consider DAO governance. The principle is beautiful: decisions made by token holders, transparent and borderless. But in practice, the upgrade rights for most DAOs sit with a small multisig of core contributors. "Code is law" only holds until the multisig signs a contract upgrade. Then the law changes.
This is not governance. It is centralized control with a decentralized veneer. And it is exactly the same structural problem AI companies face: a few individuals (the multisig signers, the safety culture of a few leaders) hold the ultimate authority to decide when to pause, upgrade, or change course. The AI letter is a cry for that authority to be distributed, but they are looking to governments instead of to code.
We know the answer. It is the one we have always preached: verify the code, trust the community. But our code is incomplete. We have not written the smart contracts that enforce safety thresholds, that require multi-stakeholder consensus before a model upgrade, that penalize excessive speed with automated slashing.
Imagine an AI DAO that governs the release of new model weights. The threshold for approval is not a single company's risk committee, but a vote by a diverse set of validators—researchers, ethicists, end-users. The training process is transparent; the compute resources are verified on-chain. The decision to stop or proceed is not a plea to a government, but a function call to a smart contract.
This is not science fiction. It is a design pattern we have already built for DeFi. Smart contracts handle billions of dollars in value with autonomous logic. Why not handle frontier model risk with the same architecture?
Based on my experience building a decentralized education platform that connects crypto concepts to real-world ethics, I have seen that the hardest part is not the code. It is the social consensus about what counts as a risk threshold. The AI letter does not define what level of autonomous research capability should trigger a slowdown. They cannot define it, because it is a political negotiation masquerading as a technical one.
We can do better. We can put the thresholds in a public, upgradable DAO constitution. We can back them with staked capital that is slashed if a model is released without proper audits. We can make the safety mechanism a first-class component of the blockchain stack, not an afterthought.
The Contrarian: Beware the Safety Moat
But let me pause before I sound like a naive evangelist. I wrote a 40-page thesis on code as covenant. I left a DeFi analytics firm because I felt complicit in what I called "financial predation disguised as innovation." I am an INFJ who reads people—and I read the AI letter with a skeptical eye.
The contrarian truth is this: the companies calling for a slowdown may be using safety as a competitive moat.
Anthropic and OpenAI have already invested heavily in safety research. They have the talent, the compute, and the brand. A government-imposed pause would freeze the competitive landscape exactly where it is—with them at the top. Smaller players, foreign labs, and open-source projects would be locked out. The slowdown becomes a privilege of the powerful.
We see the same dynamic in crypto. When large protocols call for "responsible regulation," they often propose standards that they already meet. When Layer2s call for shared security, they are often the ones who can afford the shared settlement. The covenant becomes a barrier to entry, not a protective measure for all.
So the AI letter, for all its noble language, may be a rational move by incumbents to cement their lead. The request for a pause is a request to stop time while they build the next generation of models. It is a move to turn speed into safety—and safety into a gate.
Bulls react. Bears reflect. We build. And building means recognizing that every solution contains its own form of centralization. If we implement on-chain governance for AI safety, who designs the DAO? Who controls the upgrade of the safety contracts? Who verifies the oracles that feed model capability data onto the chain?
The oracle problem is DeFi's Achilles' heel, and it is even more severe in AI. Chainlink's model of decentralized oracle nodes is still dependent on a central team to coordinate. Feeding a model's internal capability metrics onto an immutable ledger is not a solved problem. A bad oracle could trigger a false alarm, causing a costly unnecessary pause. A compromised oracle could let a dangerous model slip through.
The covenant is only as strong as its weakest link. And in AI, the weakest link is the inability to measure internal model behavior in a verifiable way. We can measure transaction outputs, but we cannot measure the latent reasoning of a language model without trusting the model's own reports. This is a fundamental epistemic limit.
So the contrarian takeaway is not to abandon the vision, but to acknowledge that it will require a new class of cryptographic primitives—verifiable training, zero-knowledge proofs of model behavior, secure multi-party computation for inference. These are not here today.
The AI letter is a mirror. It shows us the gap between our aspirations and our capabilities. We in crypto talk about sovereignty, but we have not built the tools to verify the most important thing: the behavior of the intelligence we create.
The Takeaway: The Covenant Must Come First
I still remember the solitude of the 2022 bear market. I retreated to a cabin in rural Virginia, disconnected from the noise of crypto Twitter, and spent 400 hours re-reading Hayek and Turing. Hayek taught me that decentralized knowledge leads to better decisions than central planning. Turing taught me that a machine's capabilities can exceed our understanding before we can control them.
The AI letter is a collective Turing test for our industry. The architects of the most powerful machines are admitting that their own creation's capabilities may soon eclipse their ability to govern. They are looking to governments for a solution. But governments are central planners, and Hayek showed us that central planning cannot handle the complexity of a million distributed nodes.
Crypto is the only system that can handle that complexity. A distributed, programmable, transparent covenant that binds all participants through code, not through paper agreements. But we have not yet written that covenant. We are still building the code first, hoping the values follow.
The AI community is showing us the consequence of that order. They built first, and now they are scrambling to append a governance layer on top of an already-deployed black box. It is fragile, contested, and likely to fail.
We have the chance to invert the order. Build the covenant first. Then let the code emerge within it.
That is the lesson of the 1178 signatories. It is the lesson of the ICO boom, where mission statements preceded product. It is the lesson of DeFi Summer, where yield preceded sustainability.
Tech changes. Values remain. The values are safety, fairness, and shared sovereignty. The technology to enforce them exists in our smart contracts, our DAOs, our layer-2 scaling solutions. We just have to apply it to ourselves before we are forced to apply it from above.