Every two weeks, Dario Amodei stands in front of Anthropic's entire company and talks for an hour. There is no slide deck. There is a memo he has written himself, usually three or four dense pages, covering product strategy, competitive dynamics, geopolitical risk, and whatever internal problems he thinks the company needs to hear about plainly. He calls the meeting the DVQ. The name stands for Dario Vision Quest, which he tried to resist because of its psychedelic associations but eventually accepted. The company uses it without apparent irony.
The 40% Rule
Amodei recently told Fortune that he probably spends "a third, maybe 40%, of my time making sure the culture of Anthropic is good." For the CEO of a company approaching an IPO at a near-$1 trillion valuation, building AI models used by millions, that is an unusual allocation. Most tech executives at this scale are consumed by product roadmaps, customer relationships, and investor management. Amodei has made an explicit bet that culture is not overhead — it is strategy. His premise is that building frontier AI requires a particular kind of collective reasoning, and that reasoning does not survive organizational scaling unless it is actively maintained by someone who can speak for the whole company.
Anthropic has grown from roughly 300 employees in 2023 to several thousand by mid-2026. That pace of growth typically destroys institutional memory and the shared values that early employees take as given. New hires from Google, OpenAI, and academia arrive with their own assumptions about how AI labs work and what trade-offs are acceptable. The DVQ is Amodei's answer to that drift. It forces a cadence at which the CEO articulates, in his own words, what the company currently believes and why.
Key Facts
- DVQ frequencyBiweekly all-hands
- Time Amodei spends on culture~33–40% of his schedule
- Anthropic employees (mid-2026)~4,000+
- Claude certifications issued10,000+
- Partner network applicants40,000+ firms
- Annualized revenue (May 2026)$47 billion
What Happens at a DVQ
The format, as described by Amodei and corroborated by employees, is denser than a standard all-hands. He writes the memo himself, not delegating it to communications staff, and walks through it at length with no filter on the difficult parts. Topics can range from how the company is positioned relative to OpenAI or Google DeepMind, to how Anthropic is thinking about specific geopolitical risks, to product decisions that did not go as planned. Questions are encouraged, and Amodei is known for giving substantive answers rather than deflecting to prepared talking points.
Between DVQs, he maintains a Slack channel where he responds to employee questions or shares his own reasoning on strategy shifts — sometimes in posts long enough to constitute memos in their own right. Employees say this creates a running archive of the company's internal debates, which builds shared context across teams that may never interact directly. At a company where interpretability researchers, enterprise sales teams, and infrastructure engineers all need to understand why Anthropic makes the decisions it does, that shared context has practical value.
"I want to get a reputation of telling the company the truth about what's happening, to call things what they are, to acknowledge problems, to avoid the sort of 'corpo speak,' the kind of defensive communication that often is necessary in public." Dario Amodei, Fortune, 2026
Culture as a Precondition for Safety
The theory behind the investment is not just about retention or morale. Anthropic's core argument — that it is possible to build increasingly powerful AI systems while remaining genuinely committed to safety — depends on employees believing it. Internal reporting on how Anthropic accelerated Claude Code showed a company capable of moving fast when it chose to, but also one that paused when safety questions arose. That willingness to slow down, or to decline to ship a model publicly as it did with Claude Mythos, requires staff who understand the reasoning and accept the commercial cost.
Amodei is aware the IPO will test this. Public markets put relentless pressure on quarterly results. Activist investors push for margin expansion. The culture that can hold a safety position through a bad quarter is a different animal from one that has only ever answered to private investors. His public remarks on US AI leadership suggest he has been thinking about these structural pressures for some time. The DVQ is, in part, a hedge: if the culture is genuinely internalized rather than just performed, it has a better chance of surviving the transition to public ownership.
The Spread Beyond the Company
One signal that the approach is working is the growth of Anthropic's partner network. Over 40,000 firms have applied to join the Claude Partner Network since its March launch, and more than 10,000 consultants have earned Claude certifications. The bar for working with Anthropic closely — whether as an employee or a certified partner — involves absorbing a specific set of values about what AI development should look like. The DVQ is one mechanism for keeping that bar consistent as the company's footprint expands.
Amodei's critics would argue that radical transparency is easier when a company has never faced an earnings call, and that quarterly guidance changes the calculus substantially. His supporters would point to the retention data and the fact that Anthropic has attracted researchers who left higher-paying roles elsewhere. Neither argument is settled. What is clear is that the DVQ represents a deliberate choice: at $47 billion in annualized revenue, with an IPO approaching and thousands of employees working on systems that may affect billions of people, Dario Amodei has decided that the most important thing he can do twice a month is stand up in front of his company and tell the truth.
