Anthropic: как «миссионеры» ИИ обрушили рынок на 300 миллиардов

For a long time, Anthropic was considered something like an elite monastery of Silicon Valley. Founded by brothers Amodei, former OpenAI employees, the company positioned itself as a stronghold of ethics and safety. While Sam Altman was building an empire and negotiating with Apple, Anthropic was writing "constitutions" for its models and debating how to prevent AI from destroying the world. But this week, the masks came off. It turned out that the most ideologically-driven company in the industry knows how to destroy not the world, but the market capitalization of competitors. And it does it with frightening efficiency.

The trigger for panic was the release of the Computer Use feature and new capabilities of Claude 3.5 Sonnet. The market's reaction was instant and painful: the software and financial services sector lost about 300 billion dollars. Investors suddenly realized that if a neural network can fully use a computer — move the cursor, fill out forms, switch between windows — then a huge swath of corporate software turns into a pumpkin. Why pay for dozens of specialized subscriptions if one agent can replace an entire department of data entry operators?

Context matters more than the numbers themselves here. Anthropic has always emphasized its "missionary" culture. They don't just make a product, they bring the light of safe AI. But the irony is that this very "safe" approach made their Claude 3.5 Sonnet an incredibly accurate and reliable business tool. Unlike the hallucinating models of competitors, Claude demonstrates frightening execution. This transformed the company from an academic project into the number one commercial threat to the entire SaaS sector.

Traditional giants like Salesforce or Workday spent years building their interfaces so people could work in them. Now Anthropic says: "We don't need your convenient buttons, our AI will figure it all out on its own." This is a radical paradigm shift. If AI used to be a supplement to software (Copilot), now it's becoming its replacement. Investors saw in this the beginning of the end of the era of classic corporate software. If Claude can control any interface, then the value of the interface itself approaches zero.

Why is this happening right now? After years of hype around generative AI, business began demanding real automation. Chatbots that just write poems or code don't sell for high prices anymore. Anthropic was the first to offer a solution that directly impacts the bottom line — operating costs of companies. And it did this while keeping its "good guys" image. But for software company shareholders, these "good guys" became riders of the apocalypse.

Anthropic's culture has always been closed and even somewhat arrogant. They believed their approach to safety was the only correct one. Now this confidence has transformed into a product that is not just "smarter," but "more useful" in the most mundane sense of the word. This creates a unique situation: a company that fears AI risks more than anyone has created the biggest risk to the current structure of the knowledge economy. And apparently, this is just the beginning.

The bottom line: Anthropic has proven that "ethical" AI is not only about protection from machine uprising, but also about ruthless optimization of jobs and markets. Will the old giants be able to adapt, or will Claude just hit the "delete" button on their business models?