Dario Amodei reportedly stands before thousands of Anthropic employees, unpolished notes in hand, warning that his company risks becoming “a place people join for the paycheck rather than the mission” — according to Axios reporting via Business Insider. No teleprompter. No corporate gloss. Just a CEO worried his massive compensation packages are attracting the wrong crowd. A viral Futurism headline later repackaged this as Amodei calling his staff “a bunch of untrustworthy rats” — but no transcript or primary source confirms that phrase. What he actually said is concerning enough without the sensationalized framing.
The Compensation Catch-22
Anthropic pays lavishly, then worries the money attracts the wrong people.
Anthropic uses rigid level-based pay and refuses to negotiate against outside offers. Amodei’s logic: if Mark Zuckerberg “throws a dart at a dartboard and hits your name,” matching that number would “destroy” internal fairness, per Business Insider. He boasts that some OpenAI employees “wouldn’t even talk to Zuckerberg.” The company also lets staff donate 25% of equity to charity with matching funds, per Time — a filter designed to attract public-benefit types over pure mercenaries.
Amodei reportedly spends 30 to 40 percent of his time on culture alone, according to Fortune. He holds bi-weekly all-hands sessions and encourages employees to “just argue with Dario” on Slack. “We refuse to compromise our principles because we have the confidence that people are at Anthropic because they truly believe in the mission,” he told Business Insider.
Noble framing. Then the leaked memos surfaced.
An internal memo reported by The Information and summarized by Yahoo Finance revealed Amodei calling OpenAI employees “gullible,” its supporters “Twitter morons,” and framing OpenAI’s Pentagon deal as “straight up lies” and “gaslighting.” He later apologized, per Fortune. Investors now describe his style as “volatile,” according to the New York Post, with billions hanging on whether that volatility stays contained.
Anthropic publicly positions Claude as AI built for human benefit, not peril. Yet unverified reporting describes its Mythos model — flagged internally as highly dangerous — potentially being used in NSA cyber operations. Critics at adversarial outlets allege safety pledges have been quietly softened under commercial pressure. Speculative talk of a trillion-dollar IPO valuation circulates in tech media — a figure that remains unconfirmed by any regulatory filing.
The Sharpest Irony
The CEO building job-displacing AI resents employees who treat his company like a lifeboat.
Amodei warned on CNBC that AI will cause “unusually painful” disruption to white-collar entry-level work. Commentators inflated this into “wipe out half of all entry-level jobs” — his actual language described large-scale disruption risk, not a specific figure. The tension is real regardless: recruiting mission-driven believers while building technology likely to eliminate the careers those same believers might otherwise return to. Joining the AI gold rush while it still pays is simply the rational response to the world Amodei himself is constructing.
He left OpenAI citing trust failures and “disturbing patterns.” He built Anthropic as the antidote. Now he’s policing employee motivations, sending contemptuous memos about rivals, and presiding over alleged intelligence-agency entanglements — all while insisting culture determines “how honestly you debate” and “how fast you ship,” per Fortune. The mission and the market rarely travel together cleanly. No all-hands meeting fixes that.
