There is a moment in every meaningful interview when the conversation stops being about skills and starts being about soul. For candidates walking into Anthropic's hiring process in late 2024, that moment arrives with a question that strips away all pretense of technical prowess: "If the company's stock went to zero, would you still support us?"
A candidate on Blind, the anonymous professional network, described the scene with the kind of visceral detail that only genuine discomfort produces. They answered honestly — "No, I wouldn't be happy about that" — and watched the interviewer's face harden into visible disapproval. The candidate had passed every technical hurdle, demonstrated fluency in alignment research, and articulated a thoughtful position on interpretability. None of it mattered. The question wasn't about the stock. It was about the faith.
I have spent the better part of a decade watching organizations try to codify their values into hiring pipelines. I've audited DAO governance frameworks where token alignment was supposed to guarantee behavioral alignment, and watched them fail because you cannot engineer conviction. But Anthropic's approach here is different. It's not trying to engineer conviction. It's trying to detect it — and the detection mechanism itself reveals more about the company's internal anxieties than any mission statement ever could.
This is not a story about one awkward interview question. It's a story about what happens when a research organization built on ideological purity tries to become a commercial enterprise without losing its soul. And it's a story that every founder, every governance architect, and every person who has ever signed an employment agreement should read carefully.
The Context: Safety as a Business Model
To understand why Anthropic asks this question, you have to understand what Anthropic believes it is. The company was founded in 2021 by former OpenAI researchers who left over disagreements about the direction of AI development. Their thesis was simple: AI safety cannot be an afterthought bolted onto a commercial product. It must be the foundation upon which everything else is built.
This philosophy manifests in concrete technical choices. Constitutional AI, Anthropic's signature approach, trains models to follow explicit principles rather than relying solely on human feedback. The company has invested heavily in interpretability research — understanding what models are actually doing internally, not just what they output. And its Responsible Scaling Policy commits the company to safety evaluations before deploying increasingly capable models.
These are not marketing talking points. They are expensive, difficult, and often commercially disadvantageous choices. A model that has been safety-tuned to refuse certain requests is less useful to some enterprise customers. An interpretability team that publishes papers about model internals doesn't directly generate revenue. A Responsible Scaling Policy that delays deployment to run safety evaluations means competitors ship first.
This is the context in which the interview question makes sense. Anthropic is not asking candidates to demonstrate technical competence — that's what the other five rounds of interviews are for. It's asking whether candidates understand that they are signing up for a mission that may require personal financial sacrifice. The question is a test of whether you understand what you're actually joining.
But here's what's interesting: the question also reveals that Anthropic itself is not entirely sure the mission is worth the sacrifice. If you were truly confident in your long-term value creation, would you need to ask candidates whether they'd stay if the stock went to zero? The question is a confession of uncertainty disguised as a test of conviction.
The Core: What the Question Actually Measures
Let me be precise about what this question does and doesn't measure. On the surface, it's a test of mission alignment. The candidate who says "I believe in AI safety so deeply that I would work here even if I received no financial upside" signals the kind of ideological commitment that Anthropic wants in its ranks. The candidate who says "I'm here for the compensation package" signals the kind of mercenary attitude that Anthropic fears will dilute its culture.
But the question is actually measuring something more subtle: the candidate's willingness to perform ideological purity under social pressure. The interviewer's visible disapproval of the honest answer sends a clear signal about what the "correct" response is. This is not a genuine exploration of values — it's a test of whether you can read the room and say what needs to be said.
I've seen this dynamic play out in DAO governance. When we design token-weighted voting systems, we assume that people will vote their genuine preferences. But in practice, social pressure and fear of ostracism often override authentic expression. The person who votes against the community consensus is punished not through formal mechanisms but through social exclusion. Over time, this creates a culture where everyone says the same thing, and the organization loses access to the diversity of thought that made it valuable in the first place.
Anthropic's interview question risks creating the same dynamic. The candidate who genuinely believes in the mission will answer correctly and pass. The candidate who is genuinely skeptical but wants the job will also answer correctly — they'll just lie. And the candidate who is genuinely honest will fail, as the Blind post demonstrates. The question doesn't filter for mission alignment. It filters for willingness to perform mission alignment.
This is a critical distinction. A company that truly values ideological diversity would ask candidates to articulate their doubts and engage with them constructively. A company that values ideological conformity asks candidates to demonstrate their purity. The former produces robust organizations that can adapt to changing circumstances. The latter produces echo chambers that collapse when reality fails to conform to their beliefs.
The Commercial Paradox: Paying $250,000 for People Who Don't Care About Money
Here's where the situation becomes genuinely paradoxical. Anthropic is simultaneously offering compensation packages exceeding $250,000 — placing it in the top tier of AI industry salaries alongside OpenAI and Google DeepMind — while asking candidates whether they'd stay if the stock went to zero.
This is not a coherent position. If you're paying top-of-market salaries, you're competing in the market for top talent. That market is brutally efficient: the best AI researchers can command $500,000 to $1 million or more in total compensation. Anthropic's $250,000+ packages are competitive, but they're not extraordinary. The company needs to offer them just to be in the game.
But then the interview question sends a different signal. It says: "We're paying you well, but we want you to pretend that the money doesn't matter." This is the corporate equivalent of asking your partner to sign a prenuptial agreement while also demanding they swear they'd love you even if you were broke. The two messages are in direct contradiction.
I've seen this contradiction play out in the crypto industry repeatedly. Projects that raise massive valuations and pay their teams handsomely while simultaneously demanding ideological purity from employees. The result is almost always the same: the most talented people — the ones who have real options in the market — quietly leave for organizations that are honest about the compensation-for-work exchange. The people who stay are either true believers or people who lack better options. Neither group produces optimal outcomes.
Anthropic's situation is more complex because the mission is real. AI safety is genuinely important work, and the company's technical contributions in this area are substantial. But the interview question suggests that Anthropic is trying to have it both ways: compete for top talent on compensation while also filtering for people who would work for free. This is a recipe for attracting either true believers who may lack the technical edge or performers who say what they need to say to get the job.
The Industry Signal: What This Means for AI Talent Markets
Anthropic is not operating in a vacuum. As one of the most prominent AI safety companies in the world, its hiring practices send signals throughout the industry. And those signals are already being received.
The first signal is about the value of safety expertise. By making mission alignment a central hiring criterion, Anthropic is implicitly arguing that safety work requires a specific kind of person — someone who prioritizes the mission over personal gain. This could increase the market value of researchers who genuinely combine technical excellence with safety commitment. If safety becomes a scarce combination of skills and values, the people who possess both will command premium compensation.
The second signal is about the division of the AI talent market. Anthropic's approach effectively segments the market into "safety-oriented" talent and "capability-oriented" talent. The former flows to Anthropic and similar organizations. The latter flows to OpenAI, Google DeepMind, and other companies that are more explicit about the commercial nature of their work. This segmentation could have significant consequences for the industry's trajectory.
If safety talent becomes concentrated in a few organizations, those organizations become the de facto gatekeepers of AI safety research. This is both a strength and a vulnerability. It's a strength because it creates centers of excellence where safety research can flourish. It's a vulnerability because it means the industry's safety capacity is dependent on the health and stability of a few organizations. If Anthropic were to fail — if the stock actually did go to zero — the industry would lose a disproportionate share of its safety expertise.
The third signal is about the nature of values in organizations. Anthropic's interview question is a concrete example of how abstract values get translated into operational practices. "Mission first" is a nice phrase on a careers page. Asking candidates whether they'd stay if the stock went to zero is what that phrase looks like in practice. Other companies will watch this experiment and decide whether to adopt similar practices.
Some will. Some already have. But the question is whether this approach actually produces better outcomes. Does filtering for mission alignment produce safer AI? Or does it produce organizations that are so ideologically homogeneous that they miss critical risks?
The Contrarian View: Maybe the Question Is Right
Let me steelman Anthropic's position, because I think there's more to this than I've given credit for.
The AI industry has a serious problem with talent retention. The demand for AI expertise vastly exceeds the supply, and companies are constantly poaching each other's best people. In this environment, compensation packages become the primary differentiator, and organizations that can't match the highest offers lose their best people.
Anthropic's interview question is, in this context, a defensive measure. It's an attempt to identify people who are less likely to leave for a higher offer because they're genuinely committed to the mission. If you can find people who would stay even if the stock went to zero, you've found people who won't leave when OpenAI offers them $2 million.
This is not irrational. In fact, it's a sophisticated understanding of the talent market. The cost of hiring is high — one candidate reportedly spent $4,600 on interview preparation specifically for Anthropic's process. The cost of losing a key researcher is much higher. If the interview question helps identify people who are more likely to stay, it's a good investment.
There's also an argument that the question is a legitimate test of values alignment. Anthropic's entire business model depends on its reputation for safety. If the company hired people who didn't genuinely believe in safety, the culture would erode, and the reputation would follow. The interview question is a way of protecting the company's most valuable asset: its credibility.
And there's a deeper point. The question forces candidates to confront something they might not have considered: what does it actually mean to work on AI safety? It's not just a technical challenge. It's a commitment to a particular vision of how technology should be developed. If you're not willing to make sacrifices for that vision, you're probably not the right person for the job.
I've seen this dynamic in the crypto world. The people who built the most meaningful projects — the ones that survived bear markets and regulatory crackdowns — were the ones who believed in the mission, not the ones who were in it for the money. The ones who were in it for the money left when the market turned. The ones who believed stayed and built.
So maybe Anthropic's question is not a test of ideological purity. Maybe it's a test of resilience. The AI industry is going to face serious challenges in the coming years — regulatory pressure, public scrutiny, technical setbacks. Anthropic wants to know that its employees will be there when things get hard.
The Governance Lesson: Values Are Not a Screening Tool
But here's where I have to push back, because I've seen this approach fail too many times to remain silent.
In DAO governance, we've learned that you cannot screen for values. You can screen for skills. You can screen for experience. You can even screen for behavioral patterns. But values are not a fixed trait that you can detect in an interview. They're a dynamic relationship between an individual and an organization that evolves over time.
The person who answers "yes, I'd stay if the stock went to zero" in an interview might feel very differently after two years of working on the same problem. The person who answers honestly that they'd be unhappy might become the most committed member of the team once they understand the work more deeply. Values are not static. They're shaped by experience.
This is why I believe the interview question is a governance failure, not a governance success. It treats values as a screening variable when they should be treated as a developmental outcome. The question doesn't identify people who will be committed to the mission. It identifies people who are willing to say they're committed to the mission.
And there's a more insidious problem. When you screen for ideological purity, you create an environment where people are afraid to express doubt. They learn to perform commitment rather than practice it. They learn to say what the organization wants to hear rather than what they actually think. Over time, this destroys the very thing the screening was designed to protect.
I've seen this in crypto projects that demanded absolute loyalty from their communities. The projects that survived were the ones that allowed dissent, encouraged debate, and welcomed criticism. The projects that demanded purity collapsed when their leaders made mistakes and no one felt safe to point them out.
Anthropic is not a cult. It's a serious organization doing important work. But the interview question is a step in the wrong direction. It's a step toward the kind of ideological conformity that produces echo chambers and groupthink. And in the AI safety field, groupthink is not just a cultural problem. It's an existential risk.
The Takeaway: Building Organizations That Can Handle the Truth
The question Anthropic is asking candidates is not really about the stock. It's about whether the candidate can handle the tension between idealism and pragmatism. And the answer Anthropic is looking for — "yes, I'd stay even if the stock went to zero" — is not the answer of someone who has thought deeply about the relationship between mission and sustainability. It's the answer of someone who has learned to say what the interviewer wants to hear.
The better question would be: "What would you do if the stock went to zero?" That's a question that invites genuine reflection. It acknowledges that the scenario is possible and asks the candidate to think about how they would respond. It doesn't demand a particular answer. It demands a thoughtful one.
I've spent my career building governance systems for decentralized organizations. I've learned that the best systems are not the ones that enforce compliance. They're the ones that create conditions for honest engagement. They're the ones that allow people to express doubt without fear of punishment. They're the ones that recognize that commitment is not a fixed trait but a dynamic relationship that needs to be nurtured.
Anthropic's interview question is a test of whether candidates will tell the truth. But the truth is that most people would be unhappy if their stock went to zero. The truth is that most people are motivated by a combination of mission and money. The truth is that the relationship between values and compensation is complex and context-dependent.
A company that can handle the truth is a company that can build a sustainable culture. A company that demands ideological purity is a company that will eventually lose touch with reality.
Code is law, but people are the soul. And the soul cannot be screened. It can only be cultivated.
The question Anthropic should be asking is not "would you stay if the stock went to zero?" but "how would you handle it if the stock went to zero?" The first question demands a performance. The second invites a conversation. And it's in conversations — honest, difficult, sometimes uncomfortable conversations — that real commitment is built.
I don't know if Anthropic will change its approach. I suspect it won't, at least not immediately. The question has become a symbol of the company's commitment to its mission, and symbols are hard to abandon. But I hope that the people at Anthropic — the ones who genuinely care about AI safety — will ask themselves a different question: are we building an organization that can handle the truth, or an organization that demands a performance?
The answer to that question will determine whether Anthropic's mission survives contact with reality. And it will determine whether the AI safety field as a whole can build the kind of honest, self-critical culture that the challenge demands.
In the end, the stock going to zero is not the worst thing that can happen to a company. The worst thing is when the people inside the company stop telling the truth because they're afraid of what will happen if they do. That's the real risk. And it's a risk that no interview question can screen for.
You don't govern the exit, you govern the entrance. And the entrance to Anthropic is currently guarded by a question that demands a performance rather than a conversation. That's a choice. And like all choices, it has consequences.
The question is whether those consequences will be the ones Anthropic intends.
