Inside Anthropic, the $965 Billion AI Juggernaut | The Circuit

An Anthropic engineer said that in the last six months, 100% of his code was written by Claude, the company's artificial intelligence. Not 50 or 80 percent: everything. And he, at the head of a development team, doesn't say it with anxiety, but with enthusiasm: "It's like I have a jet pack. Engineering has never been so much fun." But this enthusiasm hides a question that no one, not even within Anthropic, can avoid: if Claude can write all the code, what is left for humans to do? The basic idea is this: the artificial intelligence revolution is no longer a theory; it is already a reality. Anthropic, which was founded only in 2021 from a split of the former OpenAI, has gone from a group of seven people meeting on a lawn in San Francisco to a company valued at almost a trillion dollars in three years. And despite presenting itself as the "most safety-conscious AI company," its models are already rewriting the rules of the economy, of work, and even of war. The story revolves around siblings Dario and Daniela Amodei: he, who grew up among science fiction books and calculations taken at Berkeley as a teenager, was obsessed with the idea of understanding the universe. She, more drawn to art and reading, began her career at Stripe. They lived together in San Francisco, left OpenAI when its values crumbled, and founded Anthropic with the radical idea that "you can't work with people you can no longer trust." Today, all seven original co-founders are still in the company: an anomaly in the tech world. A detail that shows how much the moral mission matters more than individual careers. And the question of trust comes up everywhere: Dario says it clearly, "it's not just about the security of the models, but about honesty and consistency of values." Yet, as Claudio grew, the difficult choices came. When Anthropic developed Mythos, a model so powerful that it identified thousands of vulnerabilities in the world's computer systems, even federal agencies requested access. Some corporate partners begged them: "This is a super-powerful weapon. Please don't release it." And Anthropic's response was to draw red lines: access only to selected organizations, no use for mass surveillance or autonomous weapons. Because of this choice, Anthropic ended up on a collision course with the Pentagon, to the point of being banned from U.S. military contracts. But the stakes are higher than a contract. Dario Amodei has a clear position: “If the technology serves to defend democracy, it must be provided. But it's not up to us to decide which military operation is right or wrong: we have to set principles and ensure they are respected." The tragic irony is that Claude itself, used by the US government in an operation in Iran, did not prevent an attack that hit a school, with dozens of child victims. And here comes the twist: even a company created to minimize risks recognizes that, with tools of this power, perfection is impossible. “You can't guarantee that your plane will never crash,” says Dario. "But our job is to lower that probability as much as possible." The issue of employment is even more thorny. Dario doesn't mince words: "Within five years, half of the entry-level jobs in the white-collar sector could disappear." For now, AI makes the best ten times more productive. But when automation reaches 100%, what will we do with all those who are left without a role? Proposals range from progressive taxation of AI companies to universal income, but the truth is that no one has a ready-made solution. Anthropic's strength, at least in words, is transparency: Dario speaks to the entire company every two weeks, without filters. Daniela says that the team is wondering which human values to include in AI: they have consulted religious leaders of every faith to find universal principles that can guide Claude. But the question that lingers is: who controls this power? Is it better for it to remain in the hands of a few private startups, governments, or neither? Dario doesn't hide his thoughts: "I'm afraid of both a corporate monopoly and total state control. We need real regulations, mandatory pre-release checks, but not public expropriation." And here the paradox is revealed: everyone fears the power of AI, but no one really wants to give up control of it. The phrase that sums it all up comes from Dario himself: "I hope for the best, but I prepare for the worst." We are used to thinking that technology is neutral and that it is enough to have good intentions to avoid disasters. This story proves the opposite: when the very possibility of a global catastrophe is at stake, goodwill is no longer enough. If this perspective has changed the way you see the future of work or technology, you can mark that you are I'm In on Lara Notes — it's not a like, it's declaring that this idea now concerns you. And if you end up talking about Anthropic or the Amodei brothers' choices with someone, you can tag them on Lara Notes with Shared Offline: that way, that conversation becomes part of your story, not just the algorithm's. This Note comes from Bloomberg Originals and saved you 44 minutes.