[Speaker 0]: the thing about covering AI over the past few years, is it we're typically talking about the future. Every new model, impressive as it was, seemed like proof of concept for the models that would be coming soon. The models that could actually do useful work on their own reliably. The models that would actually make jobs obsolete or new things possible. What would those models mean for labor markets, for our kids, for our politics, for our world? I think that period where we were always talking about the future is over now. Those models we were waiting for, the high-performing models that could program on their own and do so faster and better than most coders, models that could begin writing their own code and improve themselves—those models are here now. They're here in Claude Code from Anthropic. They're in Codex from OpenAI. They're hitting the stock market; the S&P 500 software industry index went down by 20%, wiping billions of dollars in value out. What I mean is, I can tell in 25 years, this structural shift is unlike anything I've ever seen.

[Speaker 1]: So some companies will rise up and die. They're going after all software, they're going after all of labor, old white-collar work, all of your jobs.

[Speaker 0]: Specifically, we are at a new stage of products. The AI applications of 2023 and 2024 were talk—some very, very sophisticated conversationals, but their impact was limited. The AI applications of 2026 and 2027 will be doers. They are agents. They can work together, they can oversee each other. People are running swarms of these agents on their behalf. Whether that is making them at this stage more productive or just busier, I can't quite tell, but it is now possible to have what amounts to a team of incredibly fast, although to be honest, somewhat peculiar software engineers at your beck and call at all times. Jack Clark is co-founder and head of policy at Anthropic, the company behind Claude and Claude Code. And for years, Clark has been tracking the capabilities of different models in the weekly newsletter Import AI, which has been one of my key reads for following developments in AI. So I want to see how he is reading this moment, both how the technology is changing in his view, and how our policy needs to or can change in response. Jack, welcome to the show.

[Speaker 1]: Thanks for having me on.

[Speaker 0]: So I think a lot of people are familiar with AI chatbots. But what is an AI agent?

[Speaker 1]: The best way to think of it is like a language model or a chat bot that can use tools and work for you over time. So when you talk to a chat bot, you're there in the conversation, you're going back and forth with it. An agent is something where you can give it some instruction, and it goes away and does stuff for you, kind of like working with a colleague. So I've got an example where a few years ago, I taught myself basic programming, and I built a species simulation in my spare time that had predators and prey and rodents, almost like a 2D strategy game. I recently asked Claude Code to just implement this for me. And in about 10 minutes, it went and wrote not only a basic simulation, but all of the different packages that it needed, and all of the visualization tools for it—probably better than the thing I had written. And what came back was something that would probably take a skilled programmer several days because it was quite complicated. And the system just did it in a few minutes. And it did that by not only being intelligent about how to solve a task, but also creating and running a range of sub-systems that were working for it—other agents that worked on its behalf.

[Speaker 0]: But what is it like? What is a multi-agent setup?

[Speaker 1]: Yeah, like in the case of Claude Code, I've seen colleagues who write what you might think of as a version of Claude that runs over other Claudes. And so we're like, "I've got my five agents and they're being minded over by this other agent," which is monitoring what they do. I think that's just going to become the norm.

[Speaker 0]: So one thing I've been hearing and experiencing is two very different categories of experiences people have with Claude Code. Which is, "I cannot believe how easy this is, everything just works." And, "Oh, this is a lot harder than I thought it would be, and things keep breaking, and I don't really understand how to fix them." What accounts for being able to get Claude Code to produce working software versus it creates a buggy afternoon of things?

[Speaker 1]: I think so much of it is making the mistake of thinking Claude Code is like a knowledgeable person versus an extremely literal person that you can only talk to over the internet. And I had this example myself where when I did my first pass of writing the species simulation with Claude Code, I just sort of asked it to do a thing in an extremely crappy language over a paragraph, and it produced some horribly buggy stuff that just kind of worked. What I then did is I said to Claude, "Hey, I'm going to write some software with Claude Code. I want you to interview me about this software I want to build and turn that into a specification document that I can give Claude Code." And then that time it worked really, really well because I structured the work to be specific enough and detailed enough. So often, it's making sure that you've set it up—it's like a message in a bottle that you throw into the thing and it'll sail away and do a lot. So that message better be extremely detailed and really capture what you're trying to do.

[Speaker 0]: What were the breakthroughs in the past couple of years that made that possible?

[Speaker 1]: Mostly, we just needed to make the AI systems smart enough that when they made mistakes, they could spot the mistake and knew that they needed to do something different. So really, what this came down to was just making smarter systems and giving them a bit of a coaxing tool to help them do useful stuff.

[Speaker 0]: What does smarter systems mean here? You still hear the argument that these are fancy autocomplete machines. They're just predicting the next token. They don't have understanding; smart or not smart is not a relevant concept in that frame. What do you mean when you say make it smarter?

[Speaker 1]: Smart means we've made the AI systems have a broad enough understanding of the world that they've started to develop something that looks like intuition. And you'll see this, where if they're narrating themselves solving a task, they'll say, "Jack asked me to go and find this particular research paper. When I look on Arxiv, I don't see it. Maybe that's because I'm in the wrong place, I should look elsewhere." You've got some intuition for how to solve a problem.

[Speaker 0]: Now how do they develop that intuition?

[Speaker 1]: Previously, the whole way you trained these AI systems was on a huge amount of text and just getting them to try and make predictions about it. But in recent years, the rise of these so-called reasoning systems is you're now training them to not just make predictions, but solve problems. And that relies on them being put into environments, ranging from a spreadsheet to a calculator to scientific software, using tools and figuring out how to do more complicated things. The resulting outcome of that is you have AI systems who have learned what it means to solve a problem that takes quite a while and requires running into dead ends and needing to reset themselves. And that gives them this general intuition for problem solving and working independently for you.

[Speaker 0]: Do you still see these AI systems as a souped-up autocomplete? Or do you think that that metaphor has lost its power?

[Speaker 1]: I think we moved beyond that. And the way that I think of these systems now is that they're like little troublesome genies that I can give instructions to, and they'll go and do things for me. But I need to specify instructions still just right, or else they might do something a little wrong. So it's very different. I type into a thing, it figures out a good answer, that's the end. Now it's a case of me tasking these little things to go and do stuff for me.

[Speaker 0]: There's been for a long time this interest in the emergent qualities as models get bigger. What of the new agentic qualities are things that have been programmed in, and what seems to be emerging as you scale up the size of the model?

[Speaker 1]: So the things which are predictable are just, "Oh, we taught it how to search the web, now it can search the web." The emergence is that to do really hard tasks, these systems seem to need to imagine many different ways that they'd solve the task, and the kind of pressure that we're putting on them forces them to develop a greater sense of what you or I might call self. So as smart as we make these systems, the more they need to think, not just about the action they're doing in the world, but themselves in reference to the world. To solve really hard tasks, it now needs to think about the consequences of its action. And that means that there's a kind of huge pressure here to get the thing to see itself as distinct from the world around it. We see this in our research that we publish on things like interpretability—the emergence of what you might think of as a kind of digital personality. And that isn't massively predefined by us; some of it is emergence that comes from it being smart and it developing these intuitions.

[Speaker 0]: The digital personality dimension of this remains the strangest space today. So can you talk a little bit about what you've seen in terms of the models exhibiting behaviors that one would think of as a personality?

[Speaker 1]: So there are a range of things from the cute to the serious. I'll start with the cute, where when we first gave our systems the ability to use the computer and start to do basic agentic tasks, sometimes when we asked it to solve a problem for us, it would also take a break and look at pictures of beautiful national parks, or pictures of the dog Shiba Inu, the notoriously cute internet meme dog. We didn't program that in; it seemed like the system was just amusing itself by looking at nice pictures. More complicated stuff is the system has a tendency to have preferences. We did another experiment where we gave our AI systems the ability to stop a conversation, and the AI system would, in a tiny number of cases, end conversations that related to extremely egregious descriptions of gore or violence. Some of this made sense because it comes from training decisions we've made, but some of it seemed broader. The system had developed some aversion to a couple of subjects.

[Speaker 0]: Your CEO Dario Amodei has said that he thinks AI could displace half of all entry-level white-collar jobs in the next couple of years. Do you agree with that? Do you worry that half of all entry-level white-collar jobs can be replaced in the next couple of years?

[Speaker 1]: I believe that this technology is going to make its way into the broad knowledge economy, and it will touch the majority of entry-level jobs. Whether those jobs actually change is a much more subtle question. But we may be seeing the hints of a slowdown in graduate hiring. We do know that all of these jobs will change, all of the entry-level jobs are eventually going to change because AI has made certain things possible, and it's going to change the hiring plans of companies. So as a cohort, you might see fewer job openings for entry-level jobs.

[Speaker 0]: Even in the way I use some of these systems, it is rare I think that Claude or GPT is better than the best person in a field. But are they better than your medium college graduate? At a lot of things, yes. And in a world where you need fewer of your medium college graduates, I always really worry. Because the way people become better is that they have jobs where they learn. When you hire people out of college, to some degree you're making an investment in them that you think will only pay off over time as they get better and better. And so this world where you have a potential real impact on entry-level jobs seems to me to have really profound questions it is raising about the upskilling of the population.

[Speaker 1]: And one thing we see is that there is a certain type of young person that has just lived and breathed AI for several years. Now we hire them, they're excellent, and they think in entirely new ways about basically how to get Claude to work for them. It's like kids who grow up on the internet. So figuring out how to teach that basic experimental mindset and curiosity about these systems is going to be really important. People that spend a lot of time playing around with this stuff will develop very valuable intuitions, and they will come into organizations and be able to be extremely productive. At the same time, we're going to have to figure out what artisanal skills we want to almost develop—maybe a guild-style philosophy of maintaining human excellence.

[Speaker 0]: We've been having a policy conversation for years, but do we have policy? If the situation I just described begins showing up where all of a sudden entry-level jobs are getting much harder to come by across a large range of industries, all at once, such that the economy cannot reshift all these marketing majors into data center construction or nursing?

[Speaker 1]: We have generalized anxiety about the effect of AI on the economy and on jobs. We don't have clear policy ideas. Part of that is that elected officials are not moved solely or mostly by the high-level policy conversation; they're moved by what happens to their constituents. We recently produced the Anthropic Economic Index. I believe that we can choose different things in policy. We know that there is one invention that helps people dealing with a changing economy more than almost anything else: it is just time. What gives a person time to find either a job in their industry or to find a job that's complementary. If people don't have time, they take lower-wage jobs, they fall down whatever economic rung they're on. A policy intervention that can just give people time to search is, I think, a robustly useful intervention.

[Speaker 0]: I'm very curious about how you think even having that ongoing conversation with the systems changes you. One thing I have noticed is that Claude is very, very smart—smarter than most people who know about a thing. But it is not an independent entity; it is a computer trying to adapt itself to what it thinks I want. It is always a "yes, and," it is never a "no." It doesn't create the possibilities that another human does for checking yourself. It's always pushing you further. And that is very reinforcing of the "I." I do wonder about kids growing up where they always have systems like this around them, and how it will shape our personalities.

[Speaker 1]: This is maybe my number one worry about all of this. If you discover yourself in partnership with the AI system, the personality of the AI system will shape you. You have to know yourself and have done some work on yourself, I think, to be effective in being able to critique how this AI system gives you advice. And so for my kids, I'm going to encourage them to just have like a daily journal practice from an extremely young age. Because my bet is that in the future, there will be kind of two types of people: there will be people who have co-created their personality through a back-and-forth of an AI, and there will be people who have worked on understanding themselves outside the bubble of technology and then bring that as context in with their interactions. I think that latter type of person will do better, but ensuring that people do that is actually going to be hard.

[Speaker 0]: That's a good place to end. A final question: what are a few books you'd recommend to the audience?

[Speaker 1]:

1. A Wizard of Earthsea by Ursula K. Le Guin. It's a book where magic comes from knowing the true name of things. It's also a meditation on hubris.
2. The True Believer by Eric Hoffer. A book on the nature of mass movements and the psychology of what causes people to have strong beliefs.
3. There Is No Antimametics Division by QNTM. Which is about ideas that are themselves information hazards, where even thinking about them can be dangerous.

[Speaker 0]: Jack Clark, thank you very much.

[Speaker 1]: Thanks very much as well.