Dario Amodei says the most important part of his working day is the part that does not appear on any product roadmap. In an interview published by Fortune on October 17, the Anthropic chief executive said he spends as much as 40 percent of his time on the company’s culture, and described his main worry as what happens to the way people talk when an organization gets large.
The same interview put a number on the other end of the job. Amodei said an initial public offering could value Anthropic at as much as 2 trillion dollars, a figure that would make it one of the largest companies ever to list.
Those two statements are usually made by different people. A chief executive who talks about valuation is talking to investors, and a chief executive who talks about culture is talking to employees. Amodei made both in the same conversation, and the tension between them is the interesting part.
The valuation figure needs context before it means anything. A 2 trillion dollar valuation on a company that sells subscriptions and API access implies revenue growth that continues at a rate few software businesses have sustained, and a market that expects AI spending to keep compounding after the current buildout. The number is a target, not a forecast, and the person setting it has an incentive to set it high: a private company’s valuation anchors the next round, the compensation of its employees, and the terms on which it can hire.
Public listing would change what the company is allowed to say. A private company can describe its safety work in the language of research. A public one has to disclose risk factors, litigation, customer concentration and the possibility that its own products cannibalize its revenue, in documents that lawyers review and short sellers read. The culture Amodei describes, in which people speak plainly and disagree in meetings, is easier to maintain in a company that does not publish quarterly guidance.
The mechanism he is worried about is familiar to anyone who has watched a small firm grow. Language in organizations drifts toward what is acceptable to say upward, and the drift is hard to see from the top because the people who report to the chief executive have already learned to translate. Amodei said his concern is that as the company grows, internal communication turns into corporate phrasing, which hides problems rather than surfacing them.
Anthropic’s own data suggests the organization is changing in ways that make that harder to manage. A report the company published on October 16 showed that the share of junior employees in its technical staff fell to 14 percent from 21 percent at the start of 2025, and that junior roles accounted for 9 percent of technical hiring, down from 27 percent. Technical staff fell to 56 percent of headcount from 61 percent, and employees reported spending more time in meetings.
Those numbers describe a company that is adding lawyers, policy staff, salespeople and managers faster than it is adding engineers, which is what happens when a research lab becomes a vendor. Each of those functions has its own vocabulary. The engineers who built the models talk in benchmarks and failure modes; the people who sell them talk in adoption and expansion. A chief executive trying to keep one culture across both groups is trying to keep two languages from diverging.
The forty percent figure is worth taking seriously as a management choice rather than as a sentiment. Time is the one resource a chief executive cannot delegate, and spending two days a week on hiring, performance conversations and the design of internal processes is a decision to be present in the parts of the company where culture is transmitted. The cost is visible: less time on product, on customers and on the capital conversations that a 2 trillion dollar valuation implies.
Analysts who follow the company said the culture question and the valuation question are connected by a practical thread. The thing investors are buying in a lab like Anthropic is the reliability of its output, and reliability depends on whether the people building the models will say when something is wrong. An organization that rewards agreement produces models that look good in a demo and fail in production. That is a commercial risk, not a moral one, and it is the argument Amodei is making to both audiences at once.
The comparison that comes up is to other founder-led technology companies that grew quickly and then listed. Some kept their internal candor and some did not, and the ones that lost it usually did so during a period of rapid hiring, when the number of people who had never seen the company small outnumbered the ones who had. Anthropic is in the middle of that period now.
What the interview did not resolve is how a company measures the thing it says it values. Culture is not reported in a quarterly filing, and the metrics that proxy for it, attrition, internal survey scores, the share of decisions reversed after debate, are easy to manage toward. The test will come when a business decision and a cultural one point in different directions, and when the person who disagrees is junior enough that ignoring them costs nothing in the short term.
