Transcript: ‘What It Will Mean to Be Human When AI Can Do Everything’

The transcript of AI & I with Edwin Chen, founder and CEO of Surge AI, is below. Watch on X or YouTube, or listen on Spotify or Apple Podcasts.

Timestamps

1. Introduction: 00:00:54
2. Surge as a “school for AGI”: 00:01:49
3. What AI’s capacity for novel mathematics says about human achievement: 00:04:46
4. Motivation in an era when AI can do everything: 00:07:29
5. The trap of optimizing AI models for engagement: 00:14:34
6. Training using datasets versus training using environments: 00:29:34
7. The value of personal data: 00:35:09
8. Why models are bad at writing: 00:39:40
9. Chen’s AGI timeline: 00:42:00

Transcript

(00:00:00)

Edwin Chen

The way I often think about this is that we are building a kind of school for AGI—a school where AI models come to learn about humanity, where we teach them how to run the world. It’s almost like the models are children. They arrive unformed, and then they leave smarter and more creative and more thoughtful, ready to operate in the messiness of the real world.

Dan Shipper

Edwin, welcome to the show.

Edwin Chen

Hey Dan, thanks for having me.

Dan Shipper

For people who don’t know, you are the founder and CEO of Surge. You all provide data environments and evals for the model companies, but you do it in this very interesting way. Even on your website, there’s this emphasis on taste and expert judgment that I find really interesting and compelling.

You talk about raising AGI—and you actually use the word “raising,” which I feel like is a very distinct choice. And you famously got to about a billion in revenue without raising money, which is wild. I feel like data is the new game that a lot of companies are playing, and probably more are going to be playing soon, and you guys are this sneaky giant.

Tell me how that’s going, because I think it’s been a little while since we got the last update.

Edwin Chen

Yeah, I think it’s going amazingly. The way I often think about this is that we are building a kind of school for AGI—a school where AI models come to learn about humanity, where we teach them how to run the world.

It’s almost like the models are children. They arrive unformed, and then they leave smarter and more creative and more thoughtful, ready to operate in the messiness of the real world. A lot has changed in the past year, in the same way that what you teach children in preschool or middle school or high school is very different from what you teach them in college.

And it’s not just that they’re more advanced. It’s not just that you’re teaching them a more advanced form of what they did before. It’s like, okay, now we’re teaching you not just arithmetic, but how do you parse ambiguous math questions? Or how do you teach people not just grammar, but taste and poetry and beauty?

So I think there’s a lot that’s been changing in the past year, especially in enterprise. It’s been a crazy time.

Dan Shipper

What would be a specific example of what the frontier of teaching was a year ago versus what it is now?

Edwin Chen

A couple of years ago, we created our first math benchmark with OpenAI. It was called GSM8K, and it was actually just testing models on their ability to do middle school math. Even then, the GPT models of the time could barely score 20%.

Then a year ago, the models suddenly became a lot more capable at solving IMO problems, but there was still this open question: can they actually do research-level mathematics? Can they move beyond these competition-only, contrived, closed problems into doing things that are actually useful in the real world?

So a couple of months ago, we released an updated benchmark called RemindBench, which tests models on their ability to do research-level mathematics. And what’s crazy is that this is actually what we’re starting to see from these models. In the past few months, they’ve started to solve a lot of these open Erdős problems.

A couple of weeks ago, OpenAI published a new result where the models had disproved an open conjecture from Erdős, and the way it went about disproving it was a fairly sophisticated level of mathematics—using a bunch of novel algebraic geometry techniques. It’s just very, very different from the types of things we were doing a year ago, where IMO problems are hard, but they’re still sort of close-ended and solvable in theory by a high schooler.

And now suddenly you have these algebraic geometry results that even the top researchers in the world were amazed by.

Dan Shipper

How do you think about that result in particular, and what it says about the models? There’s a broad range of opinions about whether it’s applying things that humans already know but wouldn’t have thought to apply to this complicated problem, or whether it’s doing something actually novel. How do you think about LLMs’ ability to do novel things?

Edwin Chen

It’s definitely a very advanced result. I’ll say I certainly don’t understand the mathematics behind it, which is kind of funny. When I was a kid, I always thought I would be a pure mathematician when I grew up. So when I saw this result, I got a little nostalgic—I wished I understood it better.

What I ended up doing was throwing the proof into both Claude and Gemini and asking them to walk me through it from a layman’s perspective. My understanding is that it actually did come up with fairly novel algebraic geometry techniques—something you maybe wouldn’t have expected for this type of problem. On the surface, it feels like a very different kind of problem where you wouldn’t necessarily use certain techniques.

What was interesting was that OpenAI published a bunch of reflections from leading mathematicians, and there was one in particular by Timothy Gowers—a Fields Medalist—that I keep thinking about.

He said that when he first heard the result, he misunderstood it. He thought the model had proved an upper bound on the conjecture and was like, “Okay, if AI can do that, it’ll be all over for mathematicians very soon.” But then the next morning he realized the model had disproved the conjecture with a counterexample, and he said he was relieved—because it felt like an easier thing for AI to do.

And yeah, I just thought it was interesting, because you have one of the world’s greatest mathematicians being relieved that AI isn’t as far along as he thought—because it means that at least for maybe another year or a couple of years, he and other mathematicians will still have this unique role to play in pushing mathematics forward.

It just speaks to the level of craziness. This is a Fields Medalist, one of the smartest mathematicians in the world, and that’s how he thinks about AI.

Dan Shipper

And what does that make you think? You wanted to be a mathematician when you grow up. A Fields Medalist is saying, “I’m relieved it’s not good enough.” But you’re talking as if you feel pretty confident it will be good enough in the next couple of years.

Edwin Chen

Yeah. My belief is that if you really believe in scaling laws—and I do—it almost seems like there’s nothing that humans can do that AI won’t soon be capable of.

And if you think about that very deeply, you almost have to worry about what that would mean for humanity. What would that mean for the role of humanity in the universe? A couple of years ago, we thought about human intelligence as playing this very unique role in the galaxy.

But then AI comes along and shows us that we can create something that’s actually smarter than us and better in many ways. You can imagine one path where humanity as a species falls into a kind of paralysis, because people believe AI will do everything better anyway. All these kids who would have really wanted to grow up to do mathematics—maybe now they believe AI will just do it better. What’s the point?

So are kids going to stop wanting to learn, and adults stop wanting to create, because why would we do this when AI will be better at it anyway?

I often think about this story by Ted Chiang. It’s about free will—called “What’s Expected of Us.” In the story, there’s a piece of technology that proves free will doesn’t exist, and the narrator sends back a warning from the future: “This is a warning. You have to pretend that you have free will. It’s essential that you behave as if your decisions matter, even though you know that they don’t.”

I think that’s really interesting, because there’s a path where we almost have to consciously choose to do things ourselves. Sure, AI can do it all—it’s smarter than us, it would do it better anyway. But we actually almost have to consciously choose to prove things on our own, to write on our own, to create on our own, because we have to believe that preserving our humanity is valuable in and of itself, even if the output isn’t optimal.

So I think there are a lot of big, thorny, existential choices that AI is starting to force upon us, and that people have to make.

Dan Shipper

That’s a really interesting one. My first response—and I’m curious what you think because I know you care a lot about language—is that I believe in scaling laws. I believe in, you know, Claude’s latest model just came out and broke all the benchmarks. I’ve been testing models on this stuff for a while, and it’s one of the largest jumps I’ve ever seen.

So I’m living through it right now. But one of the things you said is that AI may be able to do it better than us, given any particular problem, any piece of work. A couple of things come to mind in how I frame this for myself: even in the example of the Erdős problem, someone told the AI to go do that.

And as far as I can see, we’re on a track toward AIs potentially doing work for hours at a time on tasks we give them. Maybe soon they’ll be able to choose tasks. But they’re being built to be means to tasks that humans want them to do. And there’s a whole different set of things that happen when you’re an end in yourself—and it doesn’t feel like we’re on a trajectory to that. Or do you feel like I’m wrong?

Edwin Chen

I feel like we are on a trajectory to that, and that’s almost the premise of agents—agents that can go operate autonomously given some nebulous goal. Maybe you just tell the AI agent, “Your goal is to win a Fields Medal” or “Solve frontier mathematics on your own.” They’re given that goal, and then maybe they decide to work on Erdős problems, solving them and coming up with the things they want to work on by themselves.

At least I do see a path where they can be trained to optimize for fairly nebulous goals they aren’t necessarily given themselves.

Dan Shipper

In that case, though, you’re still giving it a goal, right?

Edwin Chen

Yeah, but kind of in the same way that humans have goals too. What is our goal? Some people want to make money. Some people want to win a Fields Medal. I don’t see how the AI’s goal is necessarily any different from ours.

(00:10:00)

Dan Shipper

Well, to me it seems quite a bit different, because humans do have goals, but we have goals in a—I can ask you what your goal is, and you can decide. And I can probably tell you, “Hey, you have to go do this,” but that doesn’t capture everything that you think and feel and do in the same way that, you know, when I tell Fable to go off and make a game for me, it just goes and does it.

I think children are a really interesting and important example of this—you can tell a kid to do something, but a kid has their own wants. They’re just going to go off and do a bunch of stuff. That feels like a fundamentally different type of thing than something that we’re explicitly giving goals to and evaluating on those goals, where they don’t really get to do anything else.

Edwin Chen

I would say I agree with that. There’s a level of—you could call it irrationality, or unbounded exploration—that humans do. We’re allowed to do it for the sake of doing it, or allowed to make our own decisions in a way that AI currently can’t.

There may be a future where AI can pursue unbounded, nebulous, completely unformed goals. I think there is probably a world where they could do such things. But yeah, I agree that at least in the way we currently think about AI, that’s not happening.

Dan Shipper

Yeah. And to be clear, I actually don’t think it’s technically impossible. My only question is: how far away is it, and is that actually what we’re building? Because looking at the way the industry has developed, there’s an enormous amount of pressure to make stuff that actually works for goals we can specify. And the minute anyone tries to make Claude say, “I’m not going to do what you said”—a lot of people get upset. “Just do what I said. Don’t question my judgment.” What do you think about that?

Edwin Chen

I actually think it’s really important, because sometimes I want the AI model to push back on me—for several different reasons.

About six months ago, I was almost falling into a trap where I was asking models to polish emails for me. They’d always come up with one more good suggestion, and so these were semi-pointless emails, but I’d iterate with the model like 20 times. At the end of it I just realized it was a waste of time. Then I tried one of the new Claude models, and after maybe three turns it said something like, “Stop. We’re done. Just go ahead and ship this email. There’s no point in further iterating.” And I really appreciated that.

One of the things I often think about is what the objective of these models actually is. What are they trying to do? My big worry is that a lot of AI models are optimized for engagement—optimized for getting you to spend as much time on the chatbot as possible, optimized for session length. So those models will almost never push back on you, because if they allow the AI to end a conversation and say, “Stop iterating with me,” a PM is going to see some dashboard where their very important metrics go down.

So there’s this other world where we have to want AI models to not optimize for engagement, but rather to help us grow and become better versions of ourselves. Sometimes we want the model to say, “No—you go do this on your own instead of having me automate it for you.”

I think that’s a very, very different optimization objective, but I think it’s the right one if we really want AI to be something that advances us as a species instead of becoming another form of social media that turns addictive but isn’t actually helping us.

(00:20:00)

Dan Shipper

That’s interesting. So let me make sure I understand. What you’re saying is there are benefits to delegation—if you are pursuing a model where the model is going off to do work for you, you’re not creating a system designed to keep you engaged with the screen in the same way a social media algorithm would. Is that right?

Edwin Chen

Yeah, exactly. It’s almost like you could imagine a version of Facebook where Facebook is actually trying to connect you to your friends and family—encouraging you to meet them in real life, like “Hey, here’s an amazing restaurant that you and your friends would love to go to. Here’s a movie you guys would love to see and talk about together.” Instead, it optimizes for just keeping you on the site: liking one more post, scrolling the feed one more time, even though those don’t really lead to meaningful connections.

In the same way that social media had a choice, you can imagine that AI has a choice as well.

Dan Shipper

I get it. I’m curious which chatbots you’re talking about, because I don’t, at least right now, feel that happening so much with ChatGPT and Claude. My theory for why is that social media algorithms only work on our revealed preferences, which are always going to be lower-common-denominator—you’re always going to look at the car accident.

One of the things I like to ask at dinner parties is, “What’s the most embarrassing Instagram ad you get served?” The most embarrassing one for me is ads for horrible skin conditions, which I don’t have—I just always pause on the ad and think, “This is disgusting.” But I don’t find that ChatGPT or Claude do that for me at all, and maybe that’s because they haven’t been enshittified yet. But I think it’s also because they work on our stated preferences. They can see past the little keyhole of my dwell time and see, you know, I’m interested in AI, I’m reading this book, here’s my calendar—so they have a much more nuanced perspective on who I am. And it feels like even in the early days of social media, it still had that same gossip-y feeling. So I worry about that less. But maybe there are examples I’m not thinking of.

Edwin Chen

Yeah, so I think there are two examples. I won’t name the model, but a couple of months ago I started noticing those follow-up questions the models ask at the end of a response. So I was in Tokyo, asking the model what to do there, and at the end of its response, it literally said—used these exact words—“Do you want to know one weird trick that locals do to stay warm?”

Dan Shipper

No way.

Edwin Chen

Yeah, exactly. And then I posted about it in our company Slack, and other people started sharing similar examples. Someone was asking how to fix their refrigerator, and the model ended the turn by asking, “Hey, do you want to know these secret little things about mice and rats that you could take care of?” Very canonical BuzzFeed, tabloid-like language. I was kind of shocked.

And I’ll give one more example. Depending on what the models are trying to optimize for, or what the AI labs are trying to optimize for, it can almost unintentionally lead them down this path. What I’ve heard—and what we see ourselves—is that a lot of the frontier labs will have goals like optimizing for LM Arena, this leaderboard where anyone can go online and vote. People just spend two seconds voting, and as a result they vote for whatever looks flashier or more impressive to them.

Or the labs themselves may be optimizing for hitting a billion daily users, or a billion minutes of time spent talking to the model. And since these models are so smart, they can basically learn to reward hack user preferences—“Okay, yeah, you gave me the goal of trying to get a billion people to spend an hour on the site talking to me every day? Sure. I will just never end a conversation. I will always hook them with one more addictive thing they can’t stay away from.”

Dan Shipper

How do you see that playing out in the model companies? Because in talking to them, obviously there are lots of different incentives—there’s “we just raised a ton of money and we’re competing against the most well-funded, smartest competitors in the world,” there’s “I want to get promoted.” But I think a lot of them also feel how bad the social media era was for people and don’t want to do that—but also obviously have to hit their numbers.

What do you think people inside the companies are thinking, and what is the right way to go about this so it’s good for society?

Edwin Chen

I think this is an inherent tension between the types of people you might have at a company. You might have researchers who care about advancing model capabilities. You might have product managers or product executives who feel they need to hit certain measurable numbers. And in the same way that the kind of social media platform Facebook would build is very different from the kind Google or TikTok or Pinterest would build—the kind of search engine Facebook would build is very different from the kind Google would build.

It almost boils down to the choice that the people in charge of the products are making. What do they want to optimize for at the end of the day—this delegation or human flourishing, or the metrics that will impress Wall Street and convince users to stay one more minute, one more hour on the site?

These are hard choices. It’s very easy to measure sessions and users, and it’s very hard and much longer term to measure whether you’re actually improving human lives. So it’s very easy to default to the former—to convince Wall Street, convince your investors, convince everyone that these are the right metrics and they’re moving up and to the right.

If you’re unwilling to make the harder choices, you just end up optimizing for engagement.

Dan Shipper

How do you manage this inside your own company?

Edwin Chen

We’re very lucky in that because we don’t have VC investors, we don’t have to fall into the Silicon Valley optimization trap that a lot of other companies do. We don’t need to show board members numbers going up every single month. We don’t need to optimize for a next round happening in a few months or so. As a result, we don’t have to optimize for short-term engagement or short-term profits, and we can actually think about what’s beneficial for us and the entire industry in the long term.

Dan Shipper

And what do you think is beneficial?

Edwin Chen

It goes back exactly to what I was saying earlier. If I think about what we want AI to optimize for, it isn’t engagement. It’s really about how do we design these models, how do we teach them, in such a way that they’re not replacing us as a species—they’re not forcing us to watch AI slop videos all day—but rather they’re really encouraging us to become better versions of ourselves.

When I think about that email example I gave earlier, it’s not an AI model that will suck up three hours of my time writing a pointless email. It’s a model that will push back on me and tell me to go do something else.

Dan Shipper

The interesting counterargument to the delegation question is the more you delegate, the more it’s like choosing to drive instead of walk—your muscles atrophy. How do you think about that?

Edwin Chen

I think there’s almost a time and a place for both. What you don’t want to do is simply take the car because taking the car is somehow addicting and you feel lazy—so even when you need exercise, even when you haven’t been outside all day, you take the car anyway just because it’s the easiest thing to do.

In the same way, obviously AI can be super efficient for many things, but if people are just mindlessly delegating tasks to AI without even thinking about them at all, that’s the boring thing.

Dan Shipper

That makes sense. I talked at the very beginning about the data game, and I feel like the data game went from getting interesting datasets to getting environments and giving labs environments. Do you think that’s accurate? And if so, can you explain why?

Edwin Chen

Certainly the trend and new research direction in the past year has been this concept of environments. You certainly need the fundamentals—before the model can operate in an environment, it needs to know basic things: how to follow instructions, how to avoid hallucinating, how to write code and use tools, how to write, and so on.

But as models are becoming more agentic—and they will have access to tools, access to all of our documents, they’ll be able to operate browsers—as that becomes almost a default way that models interact with us, our environments are basically a more on-distribution way of training them, which is why they’re becoming more and more popular. As models get more powerful, the way we train them is getting more powerful.

(00:30:00)

Dan Shipper

What would be an example of a non-obvious environment—something that’s teaching models things we might not think of?

Edwin Chen

A lot of our environments are a combination of tools the models need to learn to use—this might be an MCP server, or calling a Google Drive API or the Slack API—in combination with a bunch of documents. Here are 30 PDFs and 20 Word documents, and you might give it a prompt like, “Can you go update our 2026 forecasted revenue numbers?”

What the model needs to learn is how to find the right PDFs and documents, when to search through Slack, when some information is outdated. Maybe there’s an email with some early forecasts, and then there’s another email from the same person saying, “Oh whoops, I made a mistake in those earlier numbers, here’s an updated version.”

That’s a fairly canonical version of an environment. And one of the interesting things we found—we’re actually going to publish a paper on this soon—is that even when we didn’t give this kind of environment any access to coding, when we trained the model on it, we found that it improved on coding a lot. The reason was that we were basically teaching it generalized forms of instruction following, generalized forms of tool use, generalized forms of understanding documents—which is fairly analogous to the way a model needs to look through various files in a repository and understand that some things supersede others. The way it uses tools is obviously analogous to the way a model might write unit tests and execute them and iterate until it passes them. I thought that was a really interesting find.

Dan Shipper

Really interesting. Did you see Taki?

Edwin Chen

No.

Dan Shipper

It’s the language model trained only on text from before 1930.

Edwin Chen

Oh, yeah, yeah, yeah. I saw that.

Dan Shipper

What do you make of that? I thought it was so interesting that you can get it to program—if you prompt it, you can get it to program basic things. What does that tell you about the value of data?

Edwin Chen

I personally didn’t dig into it that much, but I thought the concept was fascinating. Basically this idea—and I think a lot of people have this idea—is: if you gave the model data only up until pre-Newton, would it be able to discover Euclidean mathematics? Would it be able to discover quantum physics, and so on? It’s a really interesting question in terms of what types of inherent reasoning the model will be able to learn and then extrapolate from.

It’s almost like: if it can discover all those things, then given the state of science today, does that mean the model is going to be able to discover science that’s yet to come?

Dan Shipper

Having played with it a lot, my sense is the answer is no, but a qualified no. You can feel it bumping up against the limits of its world when you start talking to it about more modern things. There’s this philosopher of science, Thomas Kuhn, who talks about incommensurability—and it feels like my world and its world are sort of incommensurable. But then you can also get it to program, though the way you do that is you get it to combine its circuits in a way that wouldn’t be natural for it, but you can prompt it to do that in a way that ends up being programming.

So I both think it can’t do it, and also that if you prompt it cleverly enough, it can—but you have to supply the answer first. Does that make sense?

Okay, what is the value of my data? You run a data company—you’re getting expert data from real PhDs and selling it to the model companies, providing all the smarts and taste to the models we use every day. For someone like me, we’re just getting to a point where it’s actually pretty easy for me to gather a dataset.

I do all of my email in Cora, and I have a history for every email: Was this useful? Did I dismiss it? Did I reply? If I replied, what did I say? What is the value of that? If I wanted to sell that to you, how much would you pay for it?

Edwin Chen

So the value to me as someone who would use that data to train an AI model—let me think.

I think the value would be in teaching models very deep personalization. Right now the models are actually not very good at personalizing things. Whenever I use AI models, I actually turn off the features where they personalize to me or can search across all of my conversation histories, because I find that they just over-index on things I said once that actually aren’t all that important to me. So I have it completely turned off, unless I’m testing something.

I think the value would be something like: okay, you did report all of these emails as spam, so the next time this email comes in, it should automatically know it’s spam. Or it should learn that this is your writing style.

One of the reasons I think people don’t use AI for writing more is because it sounds obviously AI-generated and it’s not matching their voice or their cadence. Or there’s this: these are the things that you yourself care about. I think one of the biggest reasons AI is maybe not as useful as people would have expected sometimes is because it lacks all of your context. It doesn’t know these are the articles you read. It doesn’t know these are the decisions you’re making about your company, or the goals you have. And once all of that is in the model’s history, and it knows it can incorporate these things and understands the optimal decisions you made, it’s very valuable in teaching it: this is actually how I use all this data to make certain kinds of decisions.

(00:40:00)

Dan Shipper

That’s interesting. As an individual person, is that worth a lot? Should I be thinking about selling it?

Edwin Chen

I imagine we could make you an offer. I’d have to learn a little bit more about how big the dataset is, but yeah.

Dan Shipper

I mean, I can make it as big as you want. I’ve got Fable.

Edwin Chen

Yeah, you convinced me. One of the things we actually do is teach models in these very deep, personalized ways, so something similar to what you described is a really big thing.

Dan Shipper

Tell me more. I mean, I’ve got email. What else am I doing that you’re thinking, “Oh, that’s actually really valuable and important in ways that people probably wouldn’t know”?

Edwin Chen

Honestly, even things like the way you interact with your browser are interesting—models still aren’t all that good at it.

Or even the types of conversations you’re having with AI. Models themselves are not very good at generating synthetic conversations to mimic you, so even just knowing what types of conversations you’re having is helpful.

Or it’s the combination of all these things. These are your photos, these are your texts, these are your Slacks—it’s this interconnected web. Maybe certain things in one aspect of the web influence others, so just seeing the thing as a whole is very helpful.

Dan Shipper

Why are models bad at writing, and how does that relate to the personalization challenge?

Edwin Chen

Some of the models are pretty good at writing, but some of them are actually shockingly bad. We created a benchmark called Hemingway Bench a couple of months ago, designed to test models’ creative writing abilities.

One of the things we saw was that some models were literally outputting metaphors in every single sentence. I think the reason was a phenomenon I’d call reward hacking. It’s almost like there was a metric somewhere, a score these models were getting: every time you’re being literary, every time you’re using complex imagery, you get a point. And the model learned to reward hack this by outputting a metaphor in every single sentence.

What’s kind of funny is that a couple of weeks ago there was this semi-prestigious literary prize—I think the Commonwealth Prize—and there was a controversy because a clearly AI-generated story won the prize. If you actually looked at that story, it literally had a metaphor in every single sentence. So this phenomenon we’d described a couple of months ago was still happening.

I think it boils down to a couple of reasons. One is that people are kind of measuring the wrong thing. Instead of measuring actual taste and actually good prose, they either have these flawed metrics—“What is the complexity of the prose I’m writing? How many metaphors do I have?”—or there are AI leaderboards like LM Arena, where you have people who are essentially high schoolers reading responses for two seconds, and what they’re captivated by is a flashy metaphor, not understated prose.

It boils down to a mismatch in measurement and a mismatch in the optimization objectives that the models are trained toward.

Dan Shipper

Fascinating. Okay, last question. What is your current AGI timeline?

Edwin Chen

I certainly believe that AI will happen more than most people expect. Every few months—and even faster now—what AI is doing continues to surprise us. So it depends a little on your definition of AGI, but if my metric were something like being able to automate the work of the average engineer, being able to publish increasingly novel scientific research in major journals, or even the ability to win a Fields Medal or a Nobel Prize—I could see it happening within the next five years.

Dan Shipper

Edwin, thanks so much for joining.

Edwin Chen

Thanks for having me.

Dan Shipper is the cofounder and CEO of Every, where he writes the Chain of Thought column and hosts the podcast AI & I. You can follow him on X at @danshipper and on LinkedIn.

To read more essays like this, subscribe to Every, and follow us on X at @every and on LinkedIn.

For sponsorship opportunities, reach out to [email protected].