The Stock-to-Zero Question: Anthropic's Radical Bet on Values as Infrastructure

There is a question being asked in Anthropic's interview rooms that most founders would consider career suicide. It is not about technical architecture, not about scaling laws, not about training efficiency. It is this: would you be willing to let your company stock go to zero if it meant the AI you built remained safe? In a market where talent churns for token packages and equity cliffs, this question cuts through the noise like a blade. It is not a negotiation tactic. It is a values filter, designed to separate those who build for mission from those who build for exit. And it tells us something profound about where the AI-crypto convergence is heading. Noise fades. Value remains. And Anthropic is betting its entire competitive strategy on that principle. Let me be clear about what this is not. This is not a PR stunt, though it will be dismissed as one. This is not a naive moral posture, though it invites that critique. This is a deliberate piece of infrastructure — the same kind of infrastructure we talk about in decentralized systems, but applied to human capital instead of code. Think about it through the lens of trust. In crypto, we spent years building consensus mechanisms to align incentives across anonymous actors. Proof of stake, slashing conditions, bonding curves — all of it designed to ensure that participants who deviate from protocol rules pay a price. Anthropic has essentially built a slashing condition for its own hiring process. The question "would you accept zero equity for safety?" is a bonding curve for human values. It forces candidates to stake their own financial future on their stated principles. Based on my years auditing both DeFi protocols and the teams that build them, I can tell you this: most projects fail not because the code breaks, but because the people running it were never aligned on what the code was for. I have seen governance attacks that had nothing to do with smart contract vulnerabilities and everything to do with founders who cashed out their conviction for a higher valuation. The stock-to-zero question is an attempt to prevent that failure mode at the hiring stage. The strategic signal here is unmistakable. Anthropic is not trying to beat OpenAI on raw capability — at least not in the short term. It is trying to win a different game entirely. The game of trust. In a market where model performance gaps are narrowing every quarter, the differentiator becomes who you can trust with the most sensitive deployments. Governments, healthcare systems, financial infrastructure — these are the customers that will pay a premium for alignment guarantees. And Anthropic is building the brand that says: we will burn our own equity before we burn your trust. This is where the crypto parallel becomes almost uncomfortable in its clarity. We spent 2021 through 2023 watching DeFi protocols collapse because their founders prioritized TVL over transparency. The ones that survived — the ones that built durable value — were the ones that treated user funds as sacred. The same logic applies to AI. The models that get adopted for critical infrastructure will not be the smartest ones. They will be the ones whose creators demonstrated, under stress, that they would not compromise safety for revenue. But here is the contrarian angle that most commentary misses. The stock-to-zero question is not a perfect filter. It is, in fact, a deeply flawed one. And understanding its flaws tells us more about the industry than the question itself reveals. First, it selects for a specific kind of personality — the true believer. But true believers are not always the best engineers. Some of the most important safety work requires skeptical, contrarian minds that question the mission itself. If Anthropic filters too aggressively for ideological purity, it may exclude the very researchers who would catch its blind spots. I have seen this dynamic play out in crypto communities where "maximalist" cultures drove out the pragmatists who could have prevented catastrophic design flaws. Second, the question is performative. Anyone can say yes to a hypothetical. The real test comes when the board is pressuring for a launch date and the safety team says not yet. A single interview question cannot predict behavior under that kind of pressure. It can only filter for people who are good at expressing commitment — which is not the same as having it. Third, and this is the one that keeps me up at night: the question may be a form of ethics washing. It signals virtue without requiring proof. Anthropic can point to this question as evidence of its safety commitment while its actual safety research remains opaque. In crypto, we call this "proof of authority" — claiming legitimacy through position rather than through verifiable action. The market should demand more than a hiring question. It should demand auditable safety protocols, published red-team results, and measurable alignment metrics. Code executes. Ethics sustain. But ethics without verification is just marketing. So what does this mean for the broader AI-crypto landscape? I think we are witnessing the emergence of a new kind of competitive moat — one built on demonstrated values rather than technical superiority. This has profound implications for how we think about decentralized AI, autonomous agents, and the governance of intelligent systems. If Anthropic succeeds in establishing "trust" as the primary differentiator, it will force every other AI lab to answer the same question: what are you willing to sacrifice? And that question, once asked, cannot be unasked. The market will begin to price in values the same way it prices in compute. We will see the emergence of "trust ratings" for AI companies, similar to how we now evaluate the security of DeFi protocols. This is the insight most analysts are missing. The stock-to-zero question is not about hiring. It is about creating a new asset class — trust itself — and positioning Anthropic as its largest holder. In a world where AI agents will manage billions of dollars, make medical decisions, and participate in governance, the entity that controls the trust layer will control the future. Silence speaks louder than pumps. And in the silence of that interview room, Anthropic is building something that no model benchmark can measure. The question for the rest of us is whether we will demand the same standard from the projects we build and the protocols we use. Will we ask our founders what they are willing to sacrifice? Will we build slashing conditions for values, not just for tokens? Or will we continue to fund teams that treat mission statements as marketing copy? The answer to that question will determine which systems survive the next decade. And it will not be the ones with the fastest throughput or the largest models. It will be the ones that proved, under pressure, that their code and their values were the same thing. That is the legacy worth building. That is the architecture of trust. And it starts with a single question that most companies are too afraid to ask.