Dario Amodei – “The Silent Tsunami Over White-Collar Occupations”

Series: Laboratorio del Futuro – Voices Building Tomorrow

Five voices. Five perspectives on the future of work. From Musk's utopia to Hinton's warning, by way of Amodei's alert, Altman's reorganization and Li's humanization.

The laboratory of the future is not a place with robots assembling boxes. It is a space for debate, for collective decision-making, for building institutions. Because in the end, the question about work is not a technical question. It is a question about what kind of society we want to be.

And as Hinton said, the answer is not in the laboratories. It is in all of us.

Dario Amodei –

“The Silent Tsunami Over White-Collar Occupations”

The Scene at Davos, January 2026

The Swiss Alps wore their white winter mantle as the corridors of the Davos Congress Centre filled with dark suits and conversations in multiple languages. It was the 2026 edition of the World Economic Forum, and the air, already thin from the altitude, held an added density: the mix of power, money and anxiety that accompanies moments of historic transformation.

In one of the side rooms, before an audience that included the leaders of the planet's most powerful nations and the managers of the world's largest investment funds, a man in his thirties, with dark hair and thin-framed glasses, spoke with the calm of someone presenting scientific evidence. His name was Dario Amodei, he was the CEO of Anthropic, the company that created the language model Claude, and what he was saying made more than one attendee shift uncomfortably in their seat.

“Artificial intelligence could eliminate up to half of entry-level white-collar jobs within one to five years,” Amodei said, with a calm that contrasted with the force of his words. “The unemployment rate could spike to 10% or 20%.”

It was not an apocalyptic prophecy thrown into the void. It was a surgical diagnosis, based on data his own company was generating. A few meters from where Amodei spoke, in Anthropic's offices in San Francisco, software engineers had gone months without writing code from scratch. What they did was supervise, edit and guide what Claude generated for them.

The future was not in the think tanks' reports. The future had already arrived, and it was sitting in the same room.

A warning from inside the industry

The Warning from the Inside

What makes Amodei's warning particularly unsettling is that it does not come from an outside observer, or a speculative philosopher, or a politician looking for headlines. It comes from someone who is, literally, building the technology he describes. And not only that: his company is reaping the fruits of that construction at a speed that frightens even Silicon Valley's own executives.

The numbers are eloquent. Anthropic went from $100 million in revenue in 2023 to $1 billion in 2024, and from there to $10 billion in 2025. A hundredfold growth in three years. For 2026, projections point to $15 billion or even $20 billion. That is not the trajectory of just another tech company. It is the trajectory of an entire economic sector reconfiguring itself around a new raw material: artificial intelligence.

Amodei does not hide the contradiction. In his essays and interviews, he alternates between optimism about AI's potential to cure diseases, accelerate scientific discoveries and generate prosperity, and concern about the labor tsunami approaching. But when he talks about employment, his voice becomes graver, more precise. He knows that what is coming is not a repeat of earlier industrial revolutions.

Speed as a Critical Variable

If there is one word that sums up the difference between this transformation and all earlier ones, that word is speed. Amodei insists on it again and again. The agricultural revolution took centuries to move peasants from the fields to the factories. The industrial revolution took decades to reconfigure cities and professions. The artificial intelligence revolution is happening in years, perhaps in months.

“People talk about previous disruptions,” Amodei said in an interview with The New York Times. “They say: ‘Oh, yes, people used to be farmers. Then we all worked in industry. Then we all did knowledge work. People adapted.’ That happened over centuries or decades. This is happening in a very small number of years, a single digit. How do we get people to adapt fast enough?”

That is the question that obsesses Amodei. Not whether there will be adaptation, but whether adaptation can happen at the pace the change demands. Because when technology moves at exponential speed and institutions—governments, education systems, social safety nets—move at linear speed, the tear is inevitable.

The Specific Target: White-Collar Occupations

For decades, manual workers looked with a certain envy at office workers, at professionals, at the “white collars.” That was the safe haven, the territory automation did not reach because it required discernment, judgment, analytical ability. Artificial intelligence, it was thought, would affect drivers, cashiers, factory workers. Lawyers, financial analysts and programmers were safe.

Amodei has come to dispel that illusion. And he does so naming names.

His most quoted—and most disturbing—prediction is that half of entry-level white-collar jobs will disappear within one to five years. But he does not stop there. In his remarks at Davos, he was even more specific: the sectors hit hardest will be precisely those we considered safest: law, programming and finance.

Let us put names to those categories. We are talking about the young lawyers who spend months reviewing documents in law offices. We are talking about the financial analysts who prepare valuation reports. We are talking, above all, about programmers, the profession that over the last two decades has been the symbol of a promising future.

Programmers and the short reign of the centaur

The Case of Programmers: The Centaur and Its Brief Reign

If there is one sector where Amodei's diagnosis can be verified in real time, it is software engineering. What is happening inside Anthropic is a window into the future of the entire profession.

In remarks made at the Davos Forum, Amodei dropped a bombshell: within six to twelve months, artificial intelligence could do most, if not all, of the work software engineers do today. Human programmers, in that scenario, would be relegated to mere editors of machine-generated code.

But this is not a distant prediction. It is already happening. At Dreamforce 2025, held in San Francisco, Amodei revealed that, inside his own company, 90% of the code in many teams is already written by Claude. Human engineers devote themselves to supervising, to reviewing, to handling the more complex 10%. In his own words: “Six months ago, I made the prediction that in six months 90% of code would be written by AI models. Inside Anthropic and in several companies we work with, that is absolutely true.”

Amodei uses a classic metaphor to describe this phase: that of the centaur, the mythological creature that is half man, half horse. In the modern version, the centaur is the combination of human and artificial intelligence working together. The human contributes strategic direction, judgment, the ability to decide what is worth building; the machine contributes speed, precision, the capacity to execute.

The problem, Amodei warns, is that the centaur phase may be very brief. “During that centaur phase, if anything, the demand for software engineers may increase. But the period may be very brief.” What comes after is the complete, or nearly complete, replacement of the human function.

The Productivity Paradox

There is a fascinating paradox in all of this. If artificial intelligence can do the work of programmers, lawyers and analysts, who will buy the products and services those companies generate? The short answer is that, probably, the same artificial intelligence systems will, in a loop that excludes humans.

Amodei is aware of this paradox. In his January 2026 essay, titled “The Adolescence of Technology: Confronting and Overcoming the Risks of Powerful AI,” he analyzes in detail the characteristics that make this revolution different. He points to three key factors:

First, speed, which we have already discussed. Second, cognitive breadth: unlike earlier tools that automated specific tasks, AI is beginning to match humans across a broad range of mental work, from programming to legal analysis, blurring the lines between different types of jobs. Third, general labor substitution: because AI is capable of performing many cognitive tasks, it can also take on the new jobs that would traditionally arise after the old ones were automated, squeezing the labor market more than ever.

In other words, the safety cushion of earlier transitions—when one sector disappeared, workers retrained for another emerging one—might not work this time. Because AI is not a substitute for specific jobs, but a general labor substitute for humans.

Claude CoWork and the cost math for companies

The Products That Embody the Change

While Amodei speaks in Davos, his product teams are rolling out the tools that will make his prophecies real. Claude Code, the programming tool, already lets developers delegate the writing of code. But there is more.

In January 2026, Anthropic launched Claude CoWork, a software tool (initially only for macOS) capable of processing documents, organizing data and generating reports autonomously. The key is not only what it does, but for how long it can do it: according to Amodei, Claude 4 can “work independently for almost seven hours” . That is 87.5% of a standard eight-hour workday. No coffee breaks, no distractions, no need for constant supervision.

The cost analysis is devastating. A junior data analyst, with salary and social security, costs a company at least $4,000 a month. Claude CoWork costs $200 a month in its most advanced version. Even if it replaces only half of that analyst's work, the savings for the company are 48%. The return on investment is 950%.

Companies do not need to be convinced. They are already doing the math.

The Founders' Dilemma

There is something deeply uncomfortable about Amodei's position. He acknowledges it himself. On one hand, he warns about mass unemployment and social instability. On the other, his company is growing at a pace that demands continuing to automate, continuing to replace, continuing to sell tools that eliminate jobs.

The contradiction is not only moral; it is also practical. At the Davos Forum, alongside Demis Hassabis, CEO of Google DeepMind, Amodei admitted that inside Anthropic they are already seeing the effects. “I can see it within Anthropic, where I can anticipate a moment when, at the most junior end and then at the more intermediate end, we really need fewer and not more people. And we are thinking about how to handle that within Anthropic in a sensible way.”

That “handling it in a sensible way” probably includes role conversion, internal training, perhaps negotiated exits. But it does not solve the underlying problem: the company Amodei runs is building the technology that will make millions of people lose their jobs, and there is no way to stop it without also stopping the company's progress.

Some analysts point out that, unlike OpenAI, which with Operator is betting on personal assistants that make life easier, or Google DeepMind, which insists on keeping human oversight at every step, Anthropic has opted for a more direct strategy: building substitutes, not assistants. Claude does not need you to give it permission to act; it acts, and then you supervise. Or you don't.

The “boring but effective” measures and the timeline

The “Boring but Effective” Solutions

Amodei does not limit himself to diagnosing the problem. He also proposes solutions. In his essay and in his interviews, he advocates what he calls “boring but effective” measures.

Among them: transparency laws that require companies to disclose how they use AI and what impact it has on employment; chip export controls to slow development in certain regions; disclosure obligations about the behavior of models; and incremental regulatory interventions that manage to buy institutional time without freezing progress.

He also insists on the need for large-scale retraining programs. “If we don't invest in large-scale retraining programs right now, we are going to face serious social instability,” he warned in Davos. The problem, once again, is time. These programs take years to design and implement. AI, meanwhile, keeps advancing.

Amodei is critical both of those who deny the problem and of those who fall into “doomerism”—the belief that disaster is inevitable. He proposes a sober, fact-based approach that avoids both hysteria and complacency. But when asked about timelines, his diagnosis is unsettlingly concrete.

The Timeline Map

Let us gather the dates Amodei has been dropping in recent months, because they draw a map of the immediate future:

• 6 to 12 months: AI could do most of software engineers' work.
• 1 to 5 years: half of entry-level white-collar jobs could disappear.
• 1 to 2 years: we could have AI with general capabilities superior to human ones in many domains.
• Before 2030: a “very strong chance” that AI radically transforms the labor market.

The pattern is clear: what is coming is not a distant wave, it is a wall approaching at full speed.

The Question That Remains

When Amodei finishes his remarks in Davos and steps down from the stage, attendees crowd around him. They want more details, more dates, more forecasts. They want to know whether their own companies, their own jobs, are on the list.

Amodei cannot give them individual answers. But the general direction is clear. Artificial intelligence is not just another tool. It is a transformative force that is rewriting the rules of the economic game. And those who are seeing it fastest are precisely those who are building the new board.

The question hanging in the air of Davos, the same one hanging in Anthropic's offices in San Francisco, the same one beginning to hang over the conversations of millions of workers around the world, is the one Amodei asks himself without finding a definitive answer:

How do we get people to adapt fast enough?

The tsunami is approaching. And the sound you hear is not that of alarms, but of the uncomfortable silence of those who still do not know whether they will be safe when the waters arrive.

In the next article in the Laboratorio del Futuro series: we will delve into the vision of Sam Altman, CEO of OpenAI, who watches the reconfiguration of companies and the new economy that is emerging. We call it “Sam Altman and the Ten-Person Company That Bills a Billion”.