Anthropic's Dario Amodei Just Issued AI's Biggest Warning—From the Company Leading the Race

Dario Amodei doesn't usually do media blitzes, but when he does, they come with, how do we say this… an air for the dramatic?

The Anthropic CEO tends to communicate through long-form essays with titles like "Machines of Loving Grace" rather than splashy keynotes.

So when he showed up at Davos 2026 and delivered what might be the most consequential series of AI predictions any lab leader has made publicly—across three separate appearances—people paid attention.

The headlines focused on his most inflammatory comment: that the Trump administration's potential plan to sell advanced chips to China would be "a bit like selling nuclear weapons to North Korea."

But that soundbite, provocative as it is, actually undersells what Amodei laid out across several hours of interviews. He painted a picture of the next 1-5 years that should fundamentally change how you think about your career, your company, and what governments should be doing right now.

Here's the uncomfortable truth Dario seems to be telling us: The people building AI are more worried than you are. And the gap between what's coming and what we're prepared for has never been wider.

The Timeline Bombshell: "6-12 Months" From AI Doing What Software Engineers Do

Let's start with the prediction that sent shockwaves through tech Twitter.

In his conversation with Demis Hassabis, Amodei was asked about his previous prediction—made in Paris last year—that we'd have "a model that can do everything a human could do at the level of a Nobel laureate across many fields by 2026-2027."

His response, starting around the 2-minute mark: "I don't think that's going to turn out to be that far off."

Then he explained the mechanism—and this is where it gets wild.

"The mechanism whereby I imagined it would happen is that we would make models that were good at coding and good at AI research and we would use that to produce the next generation of model and speed it up to create a loop that would increase the speed of model development."

In other words: AI that's good at coding → AI that helps build better AI → faster AI development → repeat.

The loop isn't theoretical anymore. Around the 2:30 mark, Dario dropped this:

"I have engineers within Anthropic who say 'I don't write any code anymore. I just let the model write the code. I edit it. I do the things around it.'"

This isn't some junior developer on a simple task. This is happening at the company building the models.

Then came the number that should make every software engineer sit up: "We might be 6 to 12 months away from when the model is doing most, maybe all, of what software engineers do end-to-end."

Six to twelve months.

He elaborated in his earlier WSJ interview: "The team that leads one of our lead products called Claude Code—he says he hasn't written any code in the last two months. It's all Claude. He's edited it, he's looked at it, but it's all been written by Claude."

And more recently, the company just released Claude Cowork—essentially Claude Code for non-coding tasks—and "we wrote it in like a week and a half, almost entirely with Claude Code."

So the AI is building the tools that let more people use the AI. The loop is already closing.

Why This Time Might Be Different

If you've been following AI for more than five minutes, you've heard "AGI is coming soon" approximately seven thousand times. For the folks at home who don't read The Neuron or other AI news every day, AGI = Artificial General Intelligence, a nebulous phrase that refers to AI that can generalize like a human, which to us means it can do basically any task a human can do (but everyone has their own definition). Why should we believe this prediction?

Dario's answer, from his Journal interview, starting around the 1:30 mark:

"I've been watching this field for 15 years and I've been in this field for 10 years. One of the things I've most noticed is that there's been a surprisingly [consistent] trajectory, whereas the kind of public opinion and the reaction of the public has oscillated wildly."

He describes a "Moore's Law for intelligence"—the cognitive capability of models doubling roughly every 4-12 months. And unlike the hype cycles in public perception ("It's going to change everything!" followed by "It's all a bubble!"), the underlying capability curve has been remarkably smooth.

"What I see is this smooth exponential line. Similar to Moore's law for compute, we basically have a Moore's law for Intelligence, where the model is getting more and more cognitively capable every few months. That march has just been constant."

The evidence? Anthropic's own business trajectory.

In his Bloomberg interview, around the 3:45 mark, Dario laid out the revenue numbers:

• 2023: Zero to ~$100 million
• 2024: ~$100 million to ~$1 billion
• 2025: ~$1 billion to ~$10 billion

"Not exactly. These are rounded numbers. But that is roughly it."

10x revenue growth, three years in a row. And according to analyst projections shared by Ejaaz, Anthropic is targeting $70 billion in ARR by 2028.

If you're looking for evidence that the exponential is real, "revenue went from zero to ten billion in three years" is a decent starting point.

The Economic Warning No One Wants to Hear

Here's where Dario's message gets uncomfortable.

Most economic thinking about technology follows a familiar pattern: new tech creates disruption, but also creates new jobs. The automobile killed the horse-and-buggy industry but created far more automotive jobs. The internet eliminated certain roles but spawned entirely new industries.

Dario thinks this time might be genuinely different.

From his Journal interview, around the 3:50 mark:

"My view is the signature of this technology is it's going to take us to a world where we have very high GDP growth and potentially also very high unemployment and inequality. Now, that's not a combination we've almost ever seen before."

Read that again. High GDP growth AND high unemployment. Simultaneously.

"You think of it as high GDP growth, that's lots of stuff to do, lots of jobs for everyone. It's always been like that in the past. We've never had a technology that's this disruptive. So, the idea that we could have five or ten percent GDP growth, but also ten percent unemployment—it's not logically inconsistent at all. It's just never happened that way before."

In his Bloomberg conversation, he was even more direct. When asked if he still stands by his previous prediction that "50% of entry-level white collar jobs could be gone within the next one to five years," around the 12:50 mark:

"So, you know, my view of AI—there's maybe two axes. There's good things happening versus bad things happening. And there's AI is a small deal versus AI is a big deal. My view is on the extreme side of AI is a big deal, but I'm kind of in both top quadrants where I think some really good things will happen. And if we don't act to prevent them, some really bad things will happen."

The good: cure cancer, eradicate tropical diseases, understand the universe.

The bad: "We could have this very unusual combination of very fast GDP growth and high unemployment or at least underemployment or, you know, low wage jobs, high inequality."

This isn't just Dario. Demis Hassabis, in the same panel discussion, around the 16:10 mark, agreed on the near-term impact:

"I think we're going to see this year the beginnings of maybe impacting the junior level entry level kind of jobs—internships, this type of thing. I think there is some evidence... I can feel that ourselves, maybe like a slowdown in hiring in that."

The "Zeroth World" Nightmare

But the employment disruption isn't even the darkest scenario Dario painted.

In his Journal interview, around the 29:50 mark, he described what he called "the nightmare":

"The nightmare would be that there's this emerging zeroth-world country of like ten million people—that's like seven million people in Silicon Valley and three million people kind of scattered throughout—that is kind of forming its own economy and is becoming decoupled or disconnected. Maybe the ten percent GDP growth looks like fifty percent GDP growth in that part. This technology is so crazy it can pull things apart that way."

A "zeroth-world country" of 10 million tech workers experiencing 50% GDP growth while the rest of the world stagnates.

"I would almost say that it was a dystopian world. And we should think about how to stop that."

This is the CEO of a company valued at $60+ billion, potentially IPO-ing at $350+ billion this year, warning that his industry could create a dystopia.

That takes us to one of the most interesting subthemes of Dario's Davos appearances: why he thinks Anthropic is different from other AI companies.

Scientists vs. Social Media Founders: A Different Kind of AI Company

Dario was asked directly in the Bloomberg interview whether his peers in AI are taking risks seriously. His answer, around the 21:15 mark, was carefully devastating:

"I can't speak to what the other players are doing or why they're doing it. I think there are at least some other players in the ecosystem who I think of as responsible, who are at least thinking of things in the right way. I agree there are others who are not."

But the more revealing comment came in his Journal interview when discussing the difference between scientist-led AI companies and those led by "the generation of entrepreneurs that social media" produced, around the 22:50 mark:

"Scientists—there's a long tradition of scientists thinking about the effects of the technology they build, of thinking of themselves as having responsibility for the technology they build, not ducking responsibility. They're motivated in the first place by creating something for the world. And so then they worry in the cases where that something can go wrong."

Then the knife: "I think the motivation of entrepreneurs—particularly the generation of social media entrepreneurs—is very different. The selection effects that operated on them, the way in which they interacted with, you might say manipulated, consumers is very different."

He's not naming names. But when you're contrasting "scientists who feel responsibility for what they build" with "social media entrepreneurs who manipulated consumers"... well, the audience can fill in the blanks.

This connects to Anthropic's strategic positioning, around the 3:05 mark:

"We don't need to monetize a billion free users. We don't need to maximize engagement for a billion free users because we're in some death race with some other large player. And so I think that has let us think more carefully."

Enterprise revenue over consumer engagement. Products that "directly have value" over ads and attention metrics. It's a fundamentally different business model, and, Dario argues, it leads to fundamentally different incentives around safety and responsibility.

As Ejaaz noted on X: "man you've got to give it to anthropic: zero founder exits (all 6 founders still there), projecting $70B ARR by 2028, first profitable major AI lab beating OpenAI, doesn't need to turn on ads, enterprise rev supports them."

The business model isn't just good business.... it's also a family business (Dario runs Anthropic with his sister, Daniela). Whether its their mission or how the run their company, this consistency is the the structural foundation for Anthropic's claim to be the "responsible" AI lab.

"Selling Nuclear Weapons to North Korea": The Chip Export Controversy

Which brings us to the comment that generated the most headlines.

In his Bloomberg interview, Dario was asked about the Trump administration's potential policies on selling AI chips to China. His response, around the 6:45 mark, was explosive:

"The thing that is holding them back—and they've said it themselves, the CEOs of these companies say it—is the embargo on chips that's holding us back. They explicitly say this. And now, indeed, there are some policies, and I hope they change their mind, to explicitly send not quite our latest generation of chips, although it was reported that even that was being considered, but the generation of chips that's just one back—that's still extremely powerful."

Then came the nuclear analogy:

"So I think it would be a big mistake to ship these chips. The analogy I thought of—if you think about the incredible national security implications of building models that are essentially cognition, that are essentially intelligence. I've called where we're going with this 'a country of geniuses in the data center.' Imagine a hundred million people smarter than any Nobel Prize winner. And it's going to be under the control of one country or another. So I think this is crazy. I think it's a bit like selling nuclear weapons to North Korea and bragging 'Oh yeah, but we made them.'"

When the interviewer pointed out that David Sacks (the administration's AI czar) was effectively "arming the Chinese," Dario responded carefully: "I wouldn't refer to any particular people, but I would just say that this particular policy, I think, is not well-advised."

His broader concern about authoritarian governments and AI, from the Journal interview, around the 13:50 mark:

"I am concerned that AI may be uniquely well suited to autocracy and to deepening the repression that we see in autocracies. We already see it in the kind of surveillance state that is possible with today's technology. But if you think of the extent to which AI can make individualized propaganda, can break into any computer system in the world, can surveil everyone in a population, detect dissent everywhere and suppress it, make a huge army of drones that could go after each individual person—it's really scary. And we have to stop it."

This isn't abstract theorizing. Dario is saying: the technology we're building could enable a level of authoritarian control that has never existed in human history. And we're considering handing the building blocks to governments that would use it for exactly that purpose.

Andrew Curran captured the underlying logic after watching Dario's full day of interviews:

"He said that if Anthropic and DeepMind were the only two groups in the race, he would meet with Demis right now and agree to slow down. But there is no cooperation or coordination between all the different groups involved, so no one can agree on anything. This, imo, is the main reason he wanted to restrict GPU sales: chip proliferation makes this kind of agreement impossible, and if there is no agreement, then he has to blitz."

In other words: Dario isn't anti-slowdown. He's anti-unilateral slowdown. If you can't coordinate, you can't slow down. And if chips proliferate everywhere, coordination becomes mathematically impossible.

Curran's conclusion is stark: "After watching his interviews today I think Anthropic is going to lean into recursive self-improvement, and go all out from here to the finish line. They have broken their cups, and are leaving all restraint behind them." Might as well, right? 

Want proof of how confident Anthropic is in Claude's capabilities? They just open-sourced their original performance engineering take-home test—a kernel optimization challenge (writing kernels = writing optimized GPU code to speed up AI math operations, a task language models can now automate) where the baseline runs at 147,734 cycles.

The catch? Opus 4.5 already achieved 1,363 cycles. To get hired, you need to beat the AI.

As one engineer noted after attempting it: "Getting below 2000 cycles is already very impressive... yes it's pretty fucking hard."

Anthropic engineer Mihir Patel admitted that Opus 4.5 "slightly beat me when I did this." The AI is now the hiring bar for the humans building the AI.

The Hassabis Contrast: Where DeepMind's CEO Disagrees

One of the most interesting aspects of the Davos panel with both CEOs was where Demis Hassabis—Google DeepMind's co-founder—pushed back on Dario's timelines.

Asked about Dario's prediction of Nobel-laureate-level AI by 2026-2027, Hassabis responded around the 3:45 mark:

"I think there has been remarkable progress, but I think some areas of engineering work, coding, so you could say mathematics, are a little bit easier to see how they would be automated—partly because they're verifiable. You know what the output is."

His concern: "Some areas of natural science are much harder to do than that. You won't necessarily know if the chemical compound you've built or this prediction about physics is correct. It may be you have to test it experimentally and that will all take longer."

In other words: coding AI can verify its own work instantly. A biology AI might take years to validate a hypothesis experimentally. The loop closes faster in software than in science.

Hassabis also highlighted what he sees as missing capabilities, around the 4:20 mark:

"I also think there are some missing capabilities at the moment in terms of not just solving existing conjectures or existing problems but actually coming up with the question in the first place or coming up with the theory or the hypothesis. I think that's much much harder and I think that's the highest level of scientific creativity."

AI that solves problems you give it vs. AI that identifies which problems are worth solving. That's a meaningful distinction.

Interestingly, both CEOs agreed on one thing: they wish they had more time.

Dario, around the 23:40 mark: "I prefer Demis's timeline. I wish we had five to ten years."

And Hassabis, around the 21:55 mark: "If we can, maybe it would be good to have a slightly slower pace than we're currently predicting—even my timelines—so that we can get this right for society. But that would require some coordination."

The problem? As Dario noted: the reason they can't unilaterally slow down is that "we have geopolitical adversaries building the same technology at a similar pace. It's very hard to have an enforceable agreement where they slow down and we slow down."

Hence the chip export obsession. If you can't coordinate a slowdown, the next best thing is ensuring your adversaries don't have the hardware to keep up.

In a separate interview with Bloomberg's Emily Chang, Demis went deeper on why he's more conservative than Dario on job displacement.

His explanation: "jagged intelligence."

"Current systems are very good at certain things and very poor at other things. If you want to delegate an entire task to an agent rather than having what we have today—which are more like assistive programs—you're going to need a lot more consistency across the board. It's no good for it to be good at 95% of that task. You need it to be good at the whole task for you to be able to just fire and forget on it."

That said, he's not minimizing the disruption. He described it as "ten times bigger and ten times faster than the Industrial Revolution", so essentially 100X the transformation, compressed into a matter of years instead of generations.

And when asked what worries him most? It wasn't economics. "That's the thing I worry more about than the economics—what about purpose and meaning that a lot of us get from our jobs?"

Anthropic's Moment: The Business Behind the Philosophy

All of this philosophical positioning would be academic if Anthropic weren't actually winning in the market (at least in the short term). But they are.

Beyond the 10x revenue growth Dario cited, the company is executing on multiple fronts.

As Deedy Das documented from analyzing Anthropic's job postings, the company is expanding far beyond its core language model work:

• Audio: New roles to work on "understanding and generating speech and audio" including speech language models and audio diffusion models.
• Biology: Accelerate progress in life sciences by 10x.
• Cybersecurity: Building "AI powered products for cybersecurity."
• Discovery: Build an "AI Scientist" that solves "scientific Artificial General Intelligence."
• Eyes / Vision: Improve Claude's vision and spatial capabilities.

They're also making massive infrastructure investments. As Wes Roth reported, Anthropic is self-deploying approximately 1 million TPUv7 chips—rather than renting through Google Cloud. They've partnered with FluidStack, a startup formed by ex-CrusoeAI engineers, to handle the logistics of deployment, cabling, and testing.

That's the kind of capital expenditure that signals long-term commitment. You don't buy a million chips if you're planning to pivot.

And the enterprise traction is real. Around the 4:55 mark:

"When we're competing against other companies for enterprise contracts, we see just honestly and candidly—we see Google and we see OpenAI. Every once in a while, we see a couple other U.S. players. I have almost never lost a deal, lost a contract to a Chinese model."

As for the consumer competition—Gemini's recent surge to the top of app store rankings—Dario seems genuinely unperturbed. When asked about it, he said:

"We're not in that market, so I don't know about that. The statistics would seem to suggest that with their kind of greater distribution, with their greater existing reach, they've been able to make a lot of progress. And again, this is why consumers are fickle. This is why I would be careful if I were a consumer company."

Enterprise is stable. Consumer is volatile. Anthropic chose enterprise.

The Risk Essay We Haven't Seen Yet

One of the more intriguing revelations from Davos: Dario has been writing a companion piece to his famous "Machines of Loving Grace" essay—this one focused on the risks rather than the benefits.

From his panel with Demis, around the 11:05 mark:

"I wrote Machines of Loving Grace first. I'd love to give some sophisticated reason why I wrote that first, but it was just that the positive essay was easier and more fun to write than the negative essay. So I finally spent some time on vacation and I was able to write an essay about the risks."

The framing he's using:

"There's this scene from Carl Sagan's Contact, the movie version of it, where they discover alien life and there's this international panel interviewing people to be humanity's representative to meet the alien. One of the questions they ask one of the candidates is, 'If you could ask the aliens any one question, what would it be?' And one of the characters says, 'I would ask: How did you do it? How did you manage to get through this technological adolescence without destroying yourselves? How did you make it through?'"

That's the frame: how do we get through technological adolescence without destroying ourselves?

"Ever since I saw it—it was like 20 years ago, I think I saw that movie—it's kind of stuck with me. And that's the frame that I used, which is that we're knocking on the door of these incredible capabilities. The ability to build basically machines out of sand. I think it was inevitable the instant we started working with fire. But how we handle it is not inevitable."

When the essay drops, it'll be worth reading. Based on Davos, expect it to cover: AI systems that are smarter than humans and how to control them, individuals misusing AI for things like bioterrorism, nation-states weaponizing AI for surveillance and repression, economic displacement at unprecedented scale, and—most hauntingly—"what haven't we thought of."

What Anthropic Is Actually Doing About It

To Dario's credit, he's not just warning about risks—he's building solutions.

On the research side, from his Journal interview, around the 25:15 mark:

"One of the problems when we train these models is that we don't know—you can't be sure they're going to do what you think they're going to do. You can talk to the model in one context. It can say all kinds of things. Just as with a human, that may not be a faithful representation of what they're actually thinking. If they tell you, 'I'm doing X because of Y,' they might be doing X for a completely different reason. They might be lying about doing X."

The solution? Mechanistic interpretability—essentially brain scans for AI, pioneered by Anthropic co-founder Chris Olah.

"Similar to how you can learn things about human brains by doing an MRI or an X-ray that you can't learn just by talking to a human, the science of looking inside the AI models—I am convinced that this ultimately holds the key to making the model safe and controllable because it's the only ground truth we have."

On the economic monitoring side, Anthropic has been running what they call the "Anthropic Economic Index" for about a year. From his Journal interview, around the 6:50 mark:

"What that does is it's a real-time index that lets you track what our model Claude is being used for. It goes across all the conversations and uses Claude in a privacy preserving way to statistically query how Claude is being used. What are the tasks it's being used for? To what extent is it automating versus augmenting tasks? What industries is it being used in? How is it diffusing across states in the United States and countries in the world?"

The logic: "Until we can measure the shape of this economic transition, any policy is going to be blind and misinformed. Many policies have gone wrong because they're based on premises that are fundamentally incorrect."

On the international front, both Dario and Demis called for minimum safety standards that apply across borders. From the panel discussion, Demis at around the 21:25 mark:

"This technology is going to be cross-border. It's going to affect everyone. It's going to affect all of humanity. Those kind of things need to be worked through."

The Question of Taxes and Redistribution

Perhaps the most surprising moment came when Dario was asked point-blank about wealth taxes, around the 17:10 mark.

"There's a wealth tax now in California. Are you supportive of that?"

His answer: "No, I think it's a great start. I think that's poorly designed."

Wait—the AI CEO supports wealth taxes? Well...

"My worry would be if we don't think about these things in a sober way, then we'll get things that are poorly designed. So my message to the others in the world who are doing well in this boom is that if we don't think proactively about how to make this revolution work for everyone, we will get these proposals that don't make sense."

His broader framing, around the 16:55 mark:

"At some point I think everyone's going to come to the realization that there needs to be some kind of macroeconomic intervention. If we look at the disparities in wealth that we have now, as a fraction of GDP, I believe we've kind of exceeded the Gilded Age already. And this is mostly without AI. So I think things are going to go even further."

And then a prediction: "My guess is this is not even going to be a partisan thing."

From the Journal interview, around the 9:05 mark:

"The pie is going to grow much larger. The money is going to be there. The budget may balance without us doing anything because there's so much growth. The issue is distributing it to the right people. I think this is probably a time to worry less about disincentivizing growth and worry more about making sure that everyone gets a part of that growth—which I know is the opposite of the prevailing sentiment now, but I think technological reality is about to change in a way that forces our ideas to change."

An AI billionaire arguing that redistribution is going to become politically inevitable—and that his industry should be proactively shaping how it happens rather than fighting it. That's not nothing.

The IPO Question

With all this context, the obvious question: is Anthropic going public this year?

Dario was characteristically non-committal. From around the 24:25 mark:

"We are most focused on just trying to make the best models and just trying to build products on top of those models and just trying to sell those models to enterprises in ways that are useful. There are large capital needs for the field. So that should be considered. But that's where our focus is."

The interviewer pressed: "So an IPO isn't completely out of the question?"

"It's never completely out of the question."

Given the reported $350B valuation discussions and the support from Google (14% ownership), Amazon (20%), and Nvidia ($5B), an IPO this year seems more likely than not.

But in classic Dario fashion, he's not going to get ahead of himself. The technology comes first. The business follows the technology.

What This Means For You

So where does this leave the average knowledge worker watching all of this unfold?

• If you're in software engineering: If you don't know by now from Claude 4.5 Opus, the 6-12 month timeline is real enough to take seriously. This doesn't mean your job disappears tomorrow. But if you're not already deeply proficient with AI coding tools, you're falling behind. The engineers Dario describes—the ones who "don't write any code anymore"—aren't unemployed. They're more productive than ever. But the entry-level path is changing. Fast.
  
• • On that front, we found Demis’ advice for would-be junior level employees or interns who might have trouble finding work this year very helpful:
    
  • • Take the time off to play with the tools.
    • He says you should try to get unbelievably proficient at using them to “leapfrog” yourself into being useful in your profession. You might even get better at using them than the researchers themselves, who spend all their time building and not enough exploring the “capability overhang” of the current and future generations of tools.
    • You hear that, young folk? Get better at AI than Demis, get a job. What, like it’s hard?
• If you're in any knowledge work: The "augmentation vs. automation" question is being answered in real-time. Some tasks are being automated entirely. Others are being augmented dramatically. The people who thrive will be those who figure out which is which... and position themselves and their skillsets on the augmentation side.
• If you're a leader or manager: The economic disruption Dario describes isn't evenly distributed. Startups are adopting AI faster than enterprises. Demis's observation that the capability overhang (what the technology CAN do vs. what organizations HAVE DONE with it) is enormous. The competitive advantage goes to those who close that gap fastest.
• If you're a policymaker: You're behind. The Anthropic Economic Index exists because Dario doesn't trust that government statistics move fast enough. The calls for chip export restrictions, international safety standards, and redistribution mechanisms are coming from inside the industry. That should tell you something about how serious this is. If you want a good framework to work off of, use this one from Dwarkesh Patel and Philip Trammell.

The Bottom Line

At Davos, Dario Amodei did something unusual: he told the truth about what's coming, even when that truth makes his own company look simultaneously essential and dangerous.

The timeline is shorter than most people think (6-12 months to AI doing most of software engineering). The economic disruption is bigger than most economists are modeling (high GDP growth AND high unemployment simultaneously). The geopolitical stakes are higher than most policymakers understand (this technology enables authoritarian control at scales that have never existed).

And the person saying all of this? He's running one of the companies building it. Not because he wants to slow down—he can't, not unilaterally—but because he thinks we need to adapt faster than we're adapting.

From the Journal House interview, his closing thought at around the 31:55 mark:

"The things I'm talking about—you could today say, 'Oh, yeah, they're politically coded in some way'—they're going to become bipartisan and universal because everyone will recognize the necessity of it. Just mark my words, we come back, if not next year, the year after—everyone's going to think this."

He's betting that reality will force our politics to change. Given what he knows about what's coming, that's probably the right bet.

The question is whether we'll change fast enough.