He writes his sensitive memos on an offline computer at home, prints them out, and hands them to colleagues. He refuses to travel to China because he fears being kidnapped. He delayed Microsoft's one-billion-dollar investment in OpenAI over a bad feeling about AGI. And now, his employees joke that he has “Sama Derangement Syndrome” — an obsession with Sam Altman so intense it borders on mania.
This is Dario Amodei, co-founder and CEO of Anthropic. The man has spent his entire career preparing for the end of the world — and building one of the most aggressive frontier AI companies on earth. The contradiction is so obvious it's almost boring. But here's what nobody in the crypto world is talking about: Dario Amodei is running a religion. Not a company. And the AI industry is blindly buying the token.
I've spent 26 years watching markets and smart contracts. I cut my teeth auditing the 0x protocol in 2017 and survived the FTX collapse by moving $2.5 million to self-custody in 48 hours. That experience taught me to read the code, not the whitepaper. And when I look at Anthropic's leadership structure, I see a massive red flag that every DeFi protocol should be studying.
Religious leaders don't issue quarterly reports. They issue visions. And Amodei's vision is the key to understanding why AI safety has become the most dangerous narrative in technology.
The God Computer
Let's start with the offline computer. Amodei will not put sensitive memos on Google Docs. Instead, he writes them on an air-gapped machine at home and prints them for colleagues. In crypto terms, this is the equivalent of a multisig wallet with a single hardware key held by the founder. It's not security. It's a bottleneck dressed in tinfoil.
The man is so paranoid he reportedly avoids China entirely for fear of being kidnapped. Fine. But here's the part that matters: before GPT-3 even started training, Amodei's safety team was already worried that the model might be close to AGI. That fear — not any actual code defect — delayed Microsoft's billion-dollar deal by several months.
Let me translate that into language we understand. A company's own safety team, led by the future CEO, blocked a major financial transaction based on a subjective assessment that “the AI might be too smart.” That's not an audit. That's a prophecy. Code doesn't care about your feelings. But Amodei's feelings were worth nine figures to Microsoft's timeline.
A former OpenAI executive described the group at the time as a “priesthood.” That's the correct word. The safety team wasn't performing objective technical reviews. They were doing divine intervention. The comparison to pre-sale token allocations and community hype is uncomfortable — because it fits too perfectly.
The Vision Quest Economy
Amodei's temperament followed him to Anthropic. The company holds an all-hands meeting every two weeks that employees internally call “Dario Vision Quest.” At these sessions, Amodei delivers long, wandering lectures about AI, politics, war, and the future of humanity. An employee says Dario “always has the singularity on his mind.”
Now, anthropic is not a monastery. It's a frontier AI lab with enormous computing power and massive enterprise valuations. But the internal culture operates like a congregation waiting for the rapture. The company even employs a team of economists — not to manage revenue, but to study what will happen to GDP and unemployment after the “singularity” arrives.
Let me be precise. They hired economists to model the economic apocalypse. Not the next eighteen months. The singularity. In the same way DeFi projects hire“game theorists” to design tokenomics that inevitably collapse under a whale attack, Anthropic is preparing for a scenario that exists entirely in the founder's head.
One major investor said it plainly: “He is less of a CEO and more of a religious leader.” That statement should terrify every serious investor who currently holds exposure to AI-derived intelligence infrastructure. Because a religious leader does not have a fiduciary duty. They have a mission from the universe.
The Battle Between Two Titans
The conflict between Amodei and Sam Altman is not a personality clash. It's a collision of two theologies. Altman is a pragmatic builder who wants to ship products and move fast. Amodei is a doomsday prophet who wants to stop the clock until he can verify every last detail.
At OpenAI, they clashed constantly. On occasion, Amodei would disappear into the office library to watch YouTube in order to calm himself down. The safety team he led effectively became a parallel government that could veto business decisions. And later, Anthropic employees joked that he suffered from “Sama Derangement Syndrome” — a fixation on Altman so deep that it shaped his every move.
Now, let me offer you a structural analysis. Everything I just described is the opposite of decentralized governance. In DeFi, we talk about protocols that cannot be rug-pulled because the smart contract is immutable and the admin keys are burned. But here we have a company where a single human can delay a billion-dollar deal by expressing anxiety. Where a CEO's personal obsession becomes a company-wide metric. Where the most sensitive information is locked away on a machine in his basement.
If this were a smart contract, we'd call it a centralized admin key with a known vulnerability. And those contracts get exploited.
The AI safety narrative is the same psychological hook as “yield is the bait, rug is the hook.” Amodei offers you a future where the AI is aligned with human values. All you have to do is trust him. And the market — like greedy LPs entering a fresh liquidity pool — is pouring billions into that promise.
## The Symmetry With Crypto The irony is that crypto and AI safety spring from the same impulse: a deep distrust of centralized authority. Nakamoto wrote Bitcoin because he didn't trust the banks. Amodei writes memos on an offline computer because he doesn't trust the cloud. Both are trying to protect against unknown adversaries.
But there's a critical divergence. The cypherpunks' solution was radical transparency: open code, public verification, cryptographic proof. Amodei's solution is radical control: isolated hardware, secret documents, and a founder who decides what's dangerous.
That's not alignment. That's the exact opposite. Surveillance wrapped in safety. I can say this from experience: when I audited the 0x protocol in 2017, the project's code was open for anyone to read. I found three reentrancy vulnerabilities in six weeks — not because I was smarter than the devs, but because the code was there. I could verify. In the AI world, there is no GitHub repository for AGI. There is only Dario's key.
And yet, the crypto industry is eagerly adopting AI agents into trading strategies without applying the same code-first verification instinct. In 2025, I integrated an open-source trading bot into my DeFi yield portfolio. The first thing I did was backtest it against my own historical data. I checked its slippage thresholds. I stress-tested its behavior during black-swan events. I did not ask the developer what he thought about the singularity. I asked the code what it did with a 2% price drop.
The people buying into the AI safety religion are not doing that. They are buying as self-esteem.
## The Contrarian Take The mainstream narrative says Dario Amodei is too paranoid, too extreme, and too obsessed with future catastrophes. The popular critique is that he's holding back progress. But based on my experience with centralized structures, I think the opposite problem is far more serious.
All paranoid founders eventually become their own single point of failure. The moment a leader believes they alone can see the endgame, they create a blind spot that destroys everything. That's not a moral flaw. It's a structural design flaw. In an audit, we call that a lack of resilience. In a religious movement, we call it martyrdom.
Bull markets love martyr narratives. They make people feel noble while they lose money. And the wider the bull market gets, the more billionaires start talking like prophets.
Amodei is not actually building safety. He is building a concentrated power structure that is impossible to audit, verify, or challenge. And the crypto industry — which has learned the hard way to distrust “proof of reserves” and unaudited token bridges — is curiously silent on this one.
Why? Because AI is the new narrative. And narratives generate exits for early investors. We saw the same thing with cross-chain bridges that raised billions before losing billions. We saw it with centralized lenders who claimed to have “insurance funds.” Now we're seeing it with AI vendors who claim to have “alignment teams.”
Let me put it bluntly: the $2.5 billion stolen across cross-chain bridge hacks was not stolen by Chinese hackers. It was stolen by a failure to verify the counterparty. Anthropic's entire premise depends on a counterparty — Dario Amodei — who claims to know what a future AGI might do. There is no code to audit. There is no oracle to query. There is only his printed memo.
Panic sells, liquidity buys. When the singularity fails to arrive, the investment cycle will flip. The next three years will see AI companies desperately spinning new narratives to keep valuations alive. And retail investors will be holding the bag, wondering why the prophet was wrong.
The Verification Primer
I know this sounds like an attack on one person. It's not. It's a structural argument. Atheists can be better accountants than believers because they don't have to protect a narrative. And in the DeFi world, the people who deliver good yield over time are the ones who treat their own conviction as a risk factor to be monitored.
If you are involved in AI or crypto — and in a bull market, you probably are — I suggest a simple tactical plan for managing this exact exposure:
First, demand that AI alignment claims be demonstrated in open, auditable code, not in blog posts. If a model's safety properties cannot be replicated by a third party, treat the claim as unverified.
Second, avoid any project that puts a single charismatic leader at the center of its risk management. In DeFi, that's called admin-key risk. In AI, it's called a “visionary.” Same red flag.
Third, create a stop-loss trigger for narrative changes. When a tech leader starts using religious language — “humanity”, “the future of everything”, “singularity” — that's not a fundamental analysis. That's a liquidity event waiting to happen.
I survived FTX because I sold the story and bought the code. I will survive the AI bubble for the exact same reason. Survival is not a default. It's a discipline.
The Final Ledger
Dario Amodei may be right that AI will one day destroy the world. Or he may be wrong. In either case, we are being asked to trust a single man's isolated computer over an open, distributed network. That's not progress. That's feudalism with better endpoint security.
The real threat is not the AI that goes rogue. It's the CEO who thinks he can control it. Because the thing we can prove continuously — across audited contracts, transparent balance sheets, and battle-tested infrastructure — is simply not present in the AI safety priesthood.
Next time you read about a frontier lab delaying a training run because the “researchers felt uneasy,” ask yourself one question: who holds the admin key? If the answer is a single human with a fear of kidnapping and an offline computer, then you are not dealing with software.
You're dealing with scripture.
And trust is not a strategy. Code doesn't care about your feelings — but it does care about your margins.