Fiverr 全押 AI：赋能创作者，而非取代他们——Fiverr CEO Micha Kaufman

Hello and welcome back to The Cognitive Revolution. Today, my guest is Micha Kaufman, founder and CEO of the international freelancing marketplace Fiverr. He’s recently launched an interesting new suite of AI products called Fiverr Go, staked out one of the most notable values-based positions for how his company intends to use AI to empower human creators, and published what began as an internal company memo meant to wake team members up to the “unpleasant truth” that AI is coming for your jobs, adding, “Heck, it’s coming for my job, too.”

As rapidly improving AI capabilities and even faster-falling prices enable startups to disrupt all sorts of markets everywhere all at once, Fiverr Go stands out as an attempt not to maximize the value-to-price ratio for customers, but to empower the freelancers who have powered the Fiverr marketplace over the last 15 years. Today, that means providing freelancers—who, interestingly, Micha notes, are very often eager early adopters of AI tools and, in some ways, much better prepared for the AI era than most salaried employees—with custom AI assistants that can help them engage and serve more customers more efficiently.

It also means providing personalized image-generation and text-to-speech models that are meant to ensure that the platform’s many graphic designers and voice-over artists maintain control over, and can continue to monetize, their own distinctive styles and sounds. For a sense of just how different Fiverr’s positioning is from most of what we hear in the AI space today, listen to a bit of the keynote presentation where Micha announced Fiverr Go.

If everything you create, the moment you create it, has zero value because AI sucks it up and turns out generative stuff based on you without giving you credit, without giving you the opportunity to gain from it, then my fear is that people’s motivation to create is going to drop to zero because there’s no value. So why would we create?

I think that a future that doesn’t motivate people to create could be a very, very dark future. What the Fiverr team has built is truly remarkable. Very few teams in the world can build this. But this doesn’t mean we think it’s perfect. It will be perfect when you say so, and our commitment is to continue trying until we get it right.

We always thought of Fiverr as a movement, and every movement needs a reason to exist. Ours is you guys. So we know that you’re going to love them because we’re not going to rest until you do.

The idea of the creation model is not to replace you. It’s not taking your voice. It’s not taking your art. It’s not taking your content. It is providing samples for customers to make them feel more confident about making a deal with you. It’s a conversion tool.

You guys are going to make more money, not necessarily by working harder, but smarter.

Now, this is obviously an attempt to position Fiverr for success in the AI age, but as you’ll hear in this conversation, it runs much deeper than that. Micha is trying to find a way to continue to motivate individual humans to contribute to the collective project of knowledge and culture-building that has advanced the human condition so spectacularly over the last 500 years.

That motivation and sense of contribution, if I understand him correctly, he sees as an intrinsic good that can’t fully be substituted for with AI-provided abundance. Obviously, it remains to be seen how well this human creator-centric model will fare in the broader marketplace, but I really do admire the fact that their approach is motivated by an aspirational sense of how the AI future ideally should be, and not simply by a values-free economic analysis of what is supposedly inevitable.

He’s humble enough to admit that he doesn’t know if the current approach will work. But as all leaders must in the AI era, he promises to keep iterating as fast as possible in pursuit of an AI-human hybrid model that’s both successful and sustainable.

At a minimum, Micha and his team have done something many might have thought impossible: They’ve launched a suite of AI products that freelance creators don’t hate. With some luck, their example might inspire other platform owners to attempt to defy market pressures and concretize their own dreams for a positive AI future.

I expect major disruption and intensifying competition, not just in the freelance labor market, but in many parts of the economy and even at the level of international relations. My goal is to help people understand AI on its own terms from as many different perspectives as possible, so we can avoid the dystopian scenarios of human disempowerment and instead, with AI’s help, live the lives we’ve always dreamed of.

Micha Kaufman, founder and CEO of Fiverr, welcome to The Cognitive Revolution.

You have, I think, for 15 years been building and leading this marketplace of international, bite-sized work projects. You can, of course, elaborate and tell us more about all the things that go on in the marketplace, but you find yourself in a very interesting position where, of course, AI is coming for all of us on some time scale.

I thought maybe it would be a great place to start with the memo you just put out. I don’t know if you intended to publish it originally or not, but it was a memo to the company that essentially said, “Okay, guys, this is the wake-up call. If you’ve been waiting for one, here it is.”

I’d love to hear your perspective, for starters, on this wake-up-call moment that maybe everybody should be tuning into as AI gets more and more powerful.

Thanks for having me, Nathan. This is going to be, I think, a very interesting conversation.

Look, one of the things that I wrote in my email is that what’s going on with AI is really the fact that it’s transforming the way all of us work, but it’s also redefining the new norm, or the new ground zero, from which we need to operate.

What I wrote in the email is that we need to understand that things that were considered easy tasks will no longer exist, what was considered hard tasks will be the new easy, and what was considered impossible tasks is going to be the new hard.

Therefore, I think that this is creating a new standard where, if you’re entry-level at something, or if you’re mediocre or even average, you have an issue right now. You have an issue.

I think that this is a wake-up call for all of us. I said that AI is coming for all of us. It’s coming for me. It’s coming for my job. I think that this forces all of us to adopt this AI-first mentality, where we force ourselves to learn new technologies, figure out how we can improve our performance, and, in that sense, do more.

My expectation from myself and from others is that we’re just going to do more. We’re going to increase our output per unit of time, and our quality as well. Those who are not going to be able to do it are going to find themselves out of the industry.

I’m not threatening anyone. I’m not trying to scare people that maybe they’re not going to have a job at Fiverr. I’m not trying to scare our freelancers, which this letter applies to as well, by saying, “You guys are going to be out of a job on our platform.” It’s saying, “You guys are going to be out of jobs in the industry,” which is very worrying.

Yesterday, Shopify—I mean, we didn’t do this, but they’re talking about layoffs of 10%. I think it’s going to be a snowball. In many ways, what we’re seeing right now—and I may be right, I may be wrong; this is just me—is the equivalent of the dot-com crash back in 2000.

It used to be Pets.com. It’s now Pets.ai. There are endless amounts of trivial things that are going to be done with AI. Nine hundred ninety-nine companies out of 1,000 that are doing AI are not going to be here because they’re doing things that are going to be interchangeable with other things that are going to be even cheaper, even faster, even whatever.

The reality is that you need fewer entry-level people and more people who know how to increase their capacity and know how to work with technology to do more, because the bar is just higher.

I think that this, much like 2000, could be a snowball effect where you see companies starting to lay off people, and those people don’t have places to go because everybody is firing people. Everybody’s trying to figure out how to upgrade their game with technology.

I’m not happy about this.

It's just that I'm calling things for what they are. And I may be wrong about this being the equivalent of the dot-com era, but I don't think I'm wrong about the expectations we have of ourselves to be able to use these new superpowers, these super tools, to achieve more. That's why I thought this was really important to share with the team, because I believe in radical candor. Radical candor comes from care: you tell people the truth because you care about them, because you want to give them the opportunity to first understand reality and then figure out what to do with it.

I've also included some high-level tips on how people should think about this. But I think the overarching idea is really to take this AI-first approach. If you're still working like it's 2024, you're doing something wrong, because that's no longer the case. It's evolving at a head-spinning speed.

Someone asked me if I think technology, or AI specifically, will replace people. They've asked me this about employees, and they've asked me this about freelancers. I said, "I don't think it will," but I do think that freelancers who master AI are going to replace those who don't. Essentially, those who are going to be masters in their skills are going to replace those who are entry-level or are just getting dragged behind.

I think it's going to be really interesting to unpack this at three levels. You obviously are leading a public company. You have very high context and a high emphasis on judgment, and your role is about—correct me if I'm wrong—but is typically described as making a few big decisions, right? That's really what matters most. Maybe it's also setting a tone and culture at the company.

Then you've got a whole full-time team that's building the platform, and these would, by any measure, be considered really good jobs. In the US, for the last 10 years or more, it's been sort of the ticket to the American dream: learn to code, get a job at a tech company that'll be stable and high-income, and you won't have anything to worry about. It'll be great. That's maybe under a bit of threat.

Then, of course, you have the actual freelance marketplace itself, where presumably the threat is most immediate, if only because—I wouldn't necessarily say that the freelance jobs are less skilled, but they're definitely lower-context in terms of how much information they really have about the client and the situation. There's sort of a low-context bottleneck that's assumed in these kinds of freelance, relatively short-engagement marketplaces.

Should we work from the marketplace up, or should we work from your position down? What do you think is the more interesting trajectory?

I don't know. You choose. My take on this is, look, when I started Fiverr 15 years ago, in 2010, freelancing was something that people mostly did in between jobs. It wasn't a full career. I think that this has changed in the past 15 years. Now it's a career, it's a lifestyle, and it's a very legitimate choice. Some of the most talented people I know decide that this is what they want to do.

The context is an interesting question. What we've seen from our platform is that a lot of the relationships that get nurtured over time on the platform remain, sometimes over years. So I think that context is less of an issue. In many ways, freelancers are not more exposed to the risk of being displaced by technology. Sometimes, in many ways, they're actually ahead of the curve because they don't have anyone to teach them anything.

They don't have the safety net of an organization that does career development and all of that. Instead, they have a hunger to teach themselves and learn everything as they go. They know all the best software, all the best tools, all the best practices, and all the best techniques. In many ways, one of the benefits that I think our customers get when they work with freelancers is exactly this: you're getting the most up-to-date, latest version of a profession.

So I wouldn't say that their exposure is higher. Plus, they've gained the discipline and the know-how to be very agile and adaptive, and to reinvent themselves very quickly. If you think about full-timers, again, they have this safety net and this supporting environment. In many ways, I think that if they do get displaced, getting up from the ground and building themselves back up is going to be mentally harder for them than it is for freelancers who are used to gaining the next customer, doing the next project, and learning the next trick.

So I'm not sure if the exposure is different, but I think it finds them in a different situation and a different environment. In many ways, I think freelancers are actually quicker to adapt in the case of AI. I've seen freelancers adapting to AI much faster than our team, which is just a proof point of what I've said. In that sense, I think they're a little more self-educated, and they have slightly better control over their time. They can decide to set aside time to learn new things and keep themselves competitive.

Anyway, we can take this from any direction you want.

Yeah, I think that's really good context, and it makes sense, though I hadn't intuited it myself. Just the fact that freelancers are responsible for outputs and are getting paid for outputs puts them in a mindset where any tool that comes along, any edge that they can get, or any way that they can serve more customers in the same amount of time is going to have a very direct impact on their personal bottom lines. Whereas, obviously, in full-time jobs, that link is often a lot more indirect at best.

Right. So that does make a lot of sense to me. Let's maybe do the marketplace level first, then go to your executive level—what you're doing—and then how you're trying to translate that into the full-time team at Fiverr. Maybe just for starters, could you characterize the marketplace? There's a ton of categories, and I browsed around a bunch, but if you weight it toward where there's the most activity in the marketplace, who are the buyers, who are the sellers, and what are the most common kinds of jobs being done?

Sure. Historically, when I started the company, there were very few categories. I think we launched with about 6, which is now closer to 800 different categories. The overwhelming majority are digital services. We don't do physical services, and most of them are creative. So anything from coding to design, music and audio, data-related services, AI-related services, and marketing services—really, it's a whole world. We've created this SKU system that allows us to create basically an infinite catalog of different services.

Actually, when we started, we started with microservices for microbusinesses. This was the beginning: tiny services, the things that are really bite-sized. Over the years, we've been growing both the supply chain and the types and complexities of projects that we tackle, and with that, the types of customers that we have. We started with microbusinesses, went through SMBs—small and medium-sized businesses—all the way to enterprise businesses. We probably serve 80% to 90% of Fortune 500 companies as customers.

At the beginning, Fiverr was primarily a marketplace. Right now, it's more than a marketplace. It's a platform where the marketplace is still a substantial portion of our business, but there are also other business tools that create this whole platform. In cases where we run very complex types of projects for our customers, the marketplace is not necessarily the obvious choice. Sometimes it's more about orchestrating multidisciplinary projects through different types of sellers or talent, and sometimes it involves agencies.

So you can find simple and very cost-effective types of services all the way up to a million-dollar transaction that could last 6 months, sometimes a year, span quarters, and be managed much more on an enterprise level. I think that is exactly what I wanted to do. I didn't want to move Fiverr as a platform from one audience to another. I wanted to extend it from entry-level all the way to the most sophisticated types of needs and the cutting-edge talent. Obviously, all of it is connected with price, time, quality, and all of this, but this is exactly the type of skills and technologies that we've built as a platform over 15 years.

So, how have you seen AI impacting the activity on the platform? You could imagine that, as we've already alluded to, existing service providers can do more with less. You could imagine that new people might show up and try to arbitrage the situation, even if they personally don't know how to do graphic design or voice-over. They could show up and try to provide it with AI and maybe not even tell anybody.

You can imagine that the mix of types of jobs flowing through might change. We've had this, by the way: I have a company called Waymark, which does video creation for small businesses. We used to offer a voice-over add-on service before we had any text-to-speech that was good enough to work. Now, of course, we have AI text-to-speech built in.

We have definitely seen a very substantial shift. It was always a minority of people who would opt into the paid voice-over service, but the volume of that service has dropped tremendously as people have found that, in many cases, the AI is good enough for their needs. It's also instant and has a lot of advantages included for free. You can imagine that the mix of services might be shifting, or you can imagine all of the above and more. What have you observed in terms of AI's impact on the platform so far?

Yeah, you know, it's kind of like asking a captain in mid-sea if he's experiencing the storm. It's ongoing, and I think that what we're seeing right now is going to make it very hard to make deterministic judgments about the current state of things because it's rapidly evolving. I think we're in the eye of the storm, and it's going to take time to figure out where this is going to start stabilizing. We're also pre-regulation, so it's basically the Wild West right now.

We can talk about this because a lot of it is crazy simply because it's breaking 500-year-old constructs. But there are obviously dynamics that we're seeing, which, by the way, are not new for us. As I've told you, we started the company in 2010 and launched with 6 categories. We now have close to 800. It doesn't mean that we didn't have hundreds of them go away, but many, many new hundreds appeared.

The dynamics of a horizontal marketplace are the reason why we can sustain a very successful business: We can shift between different categories very, very fast. What's interesting is that we always, as human beings—not necessarily as a company, but as human beings—fail to predict what the future jobs are going to be. Not even in 5 years; in 6 months or a year, it was impossible for me to see the explosion around crypto and NFTs.

NFTs burst into our lives. It lasted a year, but this was a crazy year. Everybody was all over it, and then it went away. That's fine. I read research in The Economist that stated that, in 2018—and this was before modern generative AI and LLMs—60% of the jobs that existed in 2018 didn't exist 50 years ago. Sixty percent is mind-blowing.

Think about the jobs that sustain over generations, but now it's going to be even shorter cycles. I think the only thing that changes is the velocity of change. As long as we, as a marketplace, can respond very quickly, and the talent in the market can respond very quickly, then this is just the ordinary course of business. It's not shocking; it's just the way things are.

People talk about the risk of work being displaced and living in a world without jobs, and all of that. I appreciate where this is coming from, and maybe it's right. Who knows? I'm not trying to predict. I don't think I'm presumptuous enough to make a statement that I think is half right.

The reality, if we look historically, is that every wave of transformation that came from technology eventually created more jobs than it displaced. There's a lot of research about this. Maybe it's wrong. Maybe I'm wrong. Maybe someone else is wrong. I don't know. I'm a captain in mid-sea, and it's stormy. It's hard to see.

But if you apply common sense and history—and I'm not saying it's the same as the Industrial Revolution or the automation revolution, where people had the fear of humans being displaced, and it essentially created more jobs than it displaced—AI is a little bit different. It is not just producing stuff; it's actually self-generating and generating on its own without being prompted to do so. If it's not doing this enough right now, it's going to do a lot more of this in the future.

It's slightly different from previous transformations. But my assumption is that whatever technology will go to the place of humans to continue feeding that technology, refining it, developing it, adopting it, applying it—whatever it is—is going to be such that we just don't imagine what those jobs are going to be. But here's the news: They're going to be on Fiverr. That's for sure, right?

They're going to be on Fiverr probably first because that's the machine that we've created where, if there's a new job that was invented a minute ago, you will find it on Fiverr. That's kind of how we allow our community to stay relevant.

Do you have any? One that I can't resist floating is the AI scout. That's sometimes what I describe myself as: in a fast-moving environment, what is the right tool for the job? What is AI even capable of right now? What's still out of reach? What's the best example of somebody who solved a similar problem to me?

That feels like a job that more and more organizations are going to need. I kind of do that on an ad hoc, freelance basis myself these days. It feels like more companies will bring it in-house, but a lot will probably also need point-solution expertise to scout out a particular problem set. Do you see any of that? What other new AI jobs are you starting to see flicker onto the platform?

Yeah, look, I think scouting is one, but you can always claim that you can create an agent to scout for stuff and whatever. What I do think is that, at least in our space, the future belongs to the hackers. To me, that's what I think. Those who are going to be the most resourceful people, when you see one, it's very easy to recognize.

First, they're naturally born as scouts, which means that they always know what's latest and greatest. Second, it's not just that they know what's latest and greatest; they've taken it to its boundaries and then broken it. They know how to mesh things. They know how to connect different things and make them work together. They always find the use cases that the actual developers of the technology didn't think of.

I think the future belongs to these people. I have many, many of them on the team. These guys are not scared. They're not even concerned. This email wasn't for them; they're already there. I think that, in any area, having this hardcore hacker mentality is going to be in very high demand. These are going to be the people that you pay for, and you pay generously.

As technology develops, not everybody is a scout. Not everybody is familiar with every tool. Not everybody is a prompt engineer, or at least a good one. It takes time. You need to familiarize yourself with different tools and master them, so I think there are a lot of opportunities for people like that to provide this service to others.

We're seeing some of that on our platform. We're seeing all kinds of things related to modeling, training, and fine-tuning, as well as building specific generative AI applications for customers and doing all kinds of funky stuff with mobile because it has AI chips in it. AI is being commoditized as we speak. It started out for free.

You can’t get cheaper than free. Essentially, because of scale, it’s a race to the bottom. It’s microcents for GPU cycles, whatever; that’s very basic technology. But there are incredible applications that you can actually build on top of it, which is going to be the golden era of AI.

I don’t know where we are right now. We’re not there for sure. I told you, it’s like the dot-AI bubble. I don’t know how long it’s going to last, but eventually, I’m interested in finding out where the next killer app is going to happen.

If you think about social as an example, you think about Facebook, Meta, or Instagram. You think about how the high bandwidth of cellular plus iOS has enabled this entire ecosystem, and you say, “Okay, if you take the cumulative worth of the apps on the App Store, they’re worth a lot more than Apple itself.” That’s the power of it. Apple is the underlying technology—the iOS plus the cellular network—but essentially, what’s interesting is what’s going to be built on top of it.

We’re pre-WhatsApps, Instagrams, and Ubers that were built on top of that technology. I think that right now, we’re in an exploratory mode, and a lot of our customers are in that mode. They need people to help them with a big variety of things. Plus, they do need to continue buying what we call ordinary services, like graphic design and writing, believe it or not, and marketing, whatever it is.

Again, the reason is—and this is something I’ve been trying to explain to people—that AI is incredible, no doubt. It’s giving all of us superpowers, but all of us. It’s basically ground zero. We don’t have an advantage. Nathan, you don’t have an advantage over me or over someone else just because you use AI, because everybody uses AI.

It gives you this idea: “Oh, you can design yourself.” Well, so can I. But what would make your design or mine stand out? What is the competitive advantage? I think that people go to other people for that in many ways.

Actually, I can give you a framework for thinking about why people go to other people for things. It’s very simplistic, but I think it speaks a lot to what we’re seeing with AI. There are probably 3 reasons why people go to other people for something. By the way, it’s true today. It was true 15 years ago when we started the company, even with AI and everything that happened along the way.

The first reason is arbitrage of time. Can I do something on my own? Sure. Do I need to learn how to do it? Maybe. Do I need to take time to perfect it? Yes. I don’t know—maybe it’s 3 hours, maybe it’s 3 days.

There’s a dude who knows how to give me this in 2 hours, and it costs $200. What’s my time worth if I spend 2 days? It’s a lot more. That’s a good arbitrage. Do it for me. Plain and simple. It’s easy. That’s the first reason.

The second reason is that even if maybe I can do something, maybe I can’t, there’s one thing for certain: I don’t know how to be a master at doing that task. I need a competitive edge. I need to be able to produce something that is better than what others can produce so I can compete.

You want the prettiest girl in school. You want the fanciest car. You want to fly business. You need an edge. You need to compete. It’s competitive. I’m not inventing anything. This is human beings.

You go to someone who has better taste than you, who has better command over the fanciest and best tools. You do it because they’ve gained a tremendous amount of experience in things that you haven’t experienced. So this is number 2: arbitrage of time and then competitive edge.

Number 3 is complexity. Maybe I can, maybe I can’t, but it’s freaking complex. It needs orchestration. It’s multidisciplinary. It’s too many things.

Maybe it connects to number 1: it’s not worth my time. Maybe it connects to number 2: I’m not good enough at it. But it’s complex. This is why you go to someone who’s actually going to take this out of your hands and just give it to you.

These things haven’t changed. Maybe the content of what’s considered complex or easy has changed. Maybe the type of specialization that you need has changed, but it’s the same thing. These fundamentals haven’t changed. Maybe, as I’ve said, the categories of content have changed. Maybe the definition of what was easy or hard is now different. But these are the motivations, and those have stayed constant, in my view.

What AI has done is just elevate the starting point, where none of us has an advantage.

Yeah, I sometimes describe AI as a great leveling of a lot of different playing fields. One video I would honestly recommend people go and watch is your announcement of the Fiverr Go platform. I thought it was a really interesting positioning, where you’re introducing a bunch of new AI tools to the marketplace, but also very much centering the existing community that you have and saying, “We want this to work for you. If it’s not working for you, it’s not going to be working for us either.”

You’ve described the buyer motivations for why people want to delegate and hire work done by other people. Tell us about your strategy for bringing AI to bear in service of the community that you have, as opposed to what so many others are trying to do, which is create an alternative to working with humans. You’re creating a sort of new way to work with humans.

Yeah. To explain my motivation, or why we’ve done it in the first place, I need to get a little philosophical. I don’t know if you have the patience.

Yeah, we’ve got time.

Here’s my take, and why I think it’s important to have an informed way of thinking about technology, how you use it, and the implications that your decisions are going to have—not just in the immediate term, but also slightly longer term.

What’s interesting is that, obviously, I’m the CEO of a public company, which is very much tied into technological trends and developments, and you think about how technological developments affect people’s work. But I’m also a human being. I’m also a dad. I’m also a creator.

As I was thinking about what’s really interesting about how AI burst into our lives, even though, as a company, we were using AI for 10 years, it wasn’t called AI. It was called machine learning. It was called algorithms, whatever. But what’s interesting is this: If you think about the advancements that we’ve made as a species, Homo sapiens have been around for 300,000 years.

But if you look at the arc of advancement, it started accelerating about 500 years ago. That’s it, out of 300,000 years. It doesn’t mean that we didn’t advance as a species before that, but it was very gradual and pretty linear. About 500 years ago, it started becoming exponential.

What triggered this was the revolution of the printing press—the Gutenberg press. What was it, like 470 years ago, approximately? What this allowed was more widespread access to knowledge, which made people more intelligent. But it also inspired people to think about things that they didn’t even imagine; they didn’t know existed. This really accelerated how we progressed as human beings.

I think the latest manifestation of that is the actual internet. If you look at the motivation to share as a way of progressing, look at open code, open-source code. It feels almost altruistic, but it’s not. People get recognized for it. They get opportunities because of it. Their work is continued, and it evolves over time, which I think is really incredible.

This is how we’ve progressed. This is how we came up with AI eventually, because of this arc—this exponential arc of advancement.

But here’s the thing: This is because people are motivated to create and share. The British actually understood that very early on after the invention of the printing press, and they decided to put the Copyright Act in 1710. The idea was—and this was called the Act for the Encouragement of Learning, or whatever—but the idea was that you wanted to motivate people to create and share by actually telling them, “Okay, great.”

If you created something, your name is going to be there. The creation is going to bear your name forever, for eternity. People know that you're special, and you can actually see how you impact others. You can also commercialize it; you can decide what to do.

What's really interesting about what happened with the way AI was designed in past years is that it just eats up everything everybody has ever created. Everyone's creation is just getting eaten by AI. It's getting regenerated and turned out, but there is no reward and there is no recognition. AI doesn't tell you anything: Who made this? It doesn't give you any compensation.

I'm thinking, all right, that's great. It's a buffet. It's all you can eat. It's amazing. Who cares? Well, you talk to creators and they say, "I kind of care. Will I be motivated to continue sharing my creation if this has zero value? The moment I publish something, it has no value. It's getting eaten."

This made me concerned, or at least I don't want to have a blind spot about the longer-term implications of AI the way it's designed right now. So what we've done with Fiverr Go, and in the keynote that you've mentioned, is really to talk about this and say, "We need to decide." I think there is a burning question here, which some might say, "Okay, great, we're going overboard philosophically." But in my view, the way I'm looking at it, there's one question. I've alluded to it in the keynote, and I said there's one important question: Who's in the center?

My view—and maybe some people hold a different view—is that humans are the center of the universe. Everything else is just—I mean, we should preserve our environment, but everything serves us. We're the center, and as long as we are, everything else should serve us.

I think that the way AI is designed right now, I'm not sure that this is the case. AI is still not a being, okay? But over time, it will have its own thoughts and ideas, and it's going to teach itself whatever. It's going to be like an entity. Does this entity revolve around us, or do we revolve around that new entity? If we do, then we're not the center of the universe. It is the center of the universe.

By the way, right now, the way AI is designed is that we are the product. AI is not the product, because it's free. We're training it. We're contributing stuff to make it better. Do we gain in the process? You ask a question and get an answer really quickly. Who gains more? Who's becoming better—you or it?

My view is that, at least, let's think about this and figure out if there are other ways of doing this. What we've offered with Go is just a different approach: We think we love the technology, and we think it's incredible, but we also want to experiment with giving this technology to our creators and saying, "Instead of giving away your creation to train this infinite model, train your own model."

It's going to be Nathan's model for creation, and its corpus is going to be your world of creation. It's going to be your body of work, and it's going to be very, very true to your unique signature and style.

What's interesting about this is that, for customers, it gives them a very predictable outcome. For creators, it's your model; you do with it as you please. You get the reward because, basically, sure, the model is doing regeneration, but it's regeneration based just on you.

Even Michelangelo had 9 or 10 people working in his workshop. It doesn't make his creations less Michelangelo-like, right? So, essentially, this is how we experimented with it. Instead of being a threat, it becomes your tool to do more, extend your capacity, and turn 1 person into a production house, while also giving instant gratification to customers.

The other portion is really to design this super-powerful personal assistant, which is not just assisting you by interacting with customers, answering the same questions over and over again, giving offers, and doing the intake of information. It's also a business partner.

It allows you to do all kinds of things that you had to do manually, like research on the platform to figure out how to reposition yourself or fine-tune your services and offerings. This is going completely crazy, by the way. The performance that we're seeing from the personal assistant on Fiverr Go is just mind-blowing. This is beyond expectations for such an early product.

But as I've said in the keynote, this is just an early version of a take on how you can design AI. I don't know if it's the right one. I'll know it's the right one when our community says it is amazing, and we'll get there. It's an iterative process, but at least I think it's a callout to other AI thinkers and designers that maybe we can do it. Maybe we can have a different take on AI.

Right now, I think it puts a lot of creative people in a situation where they're confused. They don't know what's happening, and they're discouraged from creating and sharing. I don't want to go back to living in a world where people are not sharing freely because they feel that it's devaluing them.

I think, first of all, I applaud you for getting philosophical. More business leaders should be getting philosophical, and, for that matter, more business leaders should be more candid with their teams about just how big of a deal they perceive this whole thing to be.

I think a lot of that is happening at the executive or board level, but then it's, "Is it really time to tell the team? We don't want to scare them, and we don't necessarily feel like we have the right inspirational message yet." Personally, I think society needs more leaders to get out there and take some shots at various strategies. As you said, is this the final form? Maybe, maybe not.

But I think we're all well served when people try something, because burying our heads in the sand and denialism is not going to work for all that much longer.

No, I agree. When you think about conception, the way we interact with devices, different form factors, technology, and the internet is changing.

Look, up until 2 years ago, for 20 years, the internet was Google. That was it. This was the starting point. What do you do? You open a browser, and it's Google. It's the search bar, and that's how you do it.

Right now, if you do a keyword search and get links, what is that? It feels like 1990. Who wants to get links and click on stuff? It feels outdated, right? There's much more of an LLM form, which, by the way, I think is going to last for a while, but it's going to be replaced by different forms of interacting.

Is texting with AI the best way of interacting? I don't know. Maybe not. This also begs the question: Do the things that we know, the things that have existed forever, stay the way they are?

I don't know. A marketplace—will we continue going on Amazon, doing a search, and seeing a billion results? Is this going to be the experience, like going into a very ordinary catalog? I don't know the answer, but I'm betting it's going to be slightly different, if not very different.

I think all of us as companies need to challenge the status quo and our conceptions, iterate, and be very, very quick, with a very high velocity of testing. There aren't yet any best-practice formulas or blueprints that work.

But that's also the fun in being an entrepreneur right now. It's a playground. It's amazing.

This gets me to what you've mentioned: You've been doing this for 15 years, and it doesn't get old. It feels like day 1. It's a playground.

Obviously, there are threats and concerns, blah blah blah, whatever. But the upside of this—oh my God, it's incredible.

Could you give a little insight into how you picked the initial product suite for Fiverr Go? You've got the assistants, which are, as you said, kind of the customer service agent and business partner for the freelancers. You've got the models that they can train, which I've seen applies to visual art—presumably taking a foundation model and fine-tuning it, or using some sort of LoRA technique on the individual creator's previous output—and then a similar thing for voice-over and music as well.

How did you go out and say, "Okay, there's an infinite universe of things we could try to bring to the marketplace. Which ones actually make the most sense?" What was the framework for ultimately reaching that launch bundle?

The personal assistant was pretty obvious. I don't think it's just customer service or customer care. It's your salesperson, and it allows our creators to spend less time doing the things they like the least, such as responding to initial messages and collecting information about the customer.

It's not that they don't love their customers, but this is just the intake and the filtering. It's answering the same questions all over again, fitting the customer's needs to their skills and availability, and figuring out how to price things.

So it's much more than just attending to customers in a very simplistic way, and it's super powerful. The assistant knows when the freelancer is actually online.

Sometimes the assistant would decide on its own to ping the freelancer, saying, “Nathan, there’s a really interesting customer, and I think you should join us,” and it would ping you to join the conversation. Sometimes it’s just allowing you to sleep and get recharged. It’s really important, and it frees up more of your time to actually create, which is what you’re passionate about.

Look, this is a conversion machine. It converts so much better than the sellers themselves. And, by the way, it’s them, because the personal assistant is trained on you—just on you. Your offerings, your profile, your entire history, sometimes tens of thousands of conversations with customers, your orders—everything. It knows everything about you.

And it’s instant. It’s always up, awake, and available. What we found is that by humanizing the assistant, we eliminated, I think, the initial propensity—I mean, people hate talking to bots. Let’s clear that up: if you create a bot, they’ll hate it. So we humanized it. We’ve done, I think, really smart things with the technology behind this, which people love.

Sometimes they get confused because we don’t disguise the fact that it’s an AI assistant. Sometimes they’re like, “Are you really AI?” They ask the question because they’re not sure. It’s like a human being, and it has the characteristics; it responds to all kinds of things. It’s really interesting.

This was obvious. Then we had the creation models, and we said, “Okay, great.” We found 60 different categories of services that we can apply generative AI to. By the way, there isn’t AI for everything right now. Maybe over time there’s going to be more and more. People talk about the idea of agentic AI and, ultimately, AGI and all of this. I’m happy to talk about my view on this, but it’s not going to be very different from things that you’ve heard before.

We focus on the areas where we think AI can give results that are going to be at least 90% accurate or true to the actual creator. If the AI technology that we’ve used, manipulated, built on top of, and mashed with other technologies doesn’t produce a 90% score of accuracy for your specific signature, it’s not going to be there. We started rolling out those different categories where we had sufficient confidence, because we don’t want the generative AI to misrepresent the creator. That would be terrible.

It’s not perfect. Since we launched Go, the way I’m approaching this is that Go is an app. We issue app updates every day. We launch new versions of it, add capabilities, fix bugs, make it simpler to use, make it quicker, make it cheaper—we make all of it better. Over time, it’s going to be able to do more stuff, and we’re going to add more capabilities to it.

This was the strategy. All of these categories that you’ve mentioned are large categories, including complex text, like business planning and stuff. Graphic design is one of our biggest categories because it’s one of the biggest categories out there in the world. The same goes for music and music production, and voice-over in particular. These are big categories, and this is where we’ve also seen AI making some really nice progress.

How do you think about upgrading these things over time for people? I understand that you have a sort of commitment to you—as the freelancer on Fiverr, you shouldn’t have to worry about all the technical details, right? This is one of the big value adds that you want to provide: making it simple for freelancers to take advantage of this stuff.

One thing I have observed, though, in some other context—I don’t know if it will apply here or not—is that sometimes a model upgrade, quote unquote, comes along and people have very different feelings about it. This famously happened with the app Replika, which is a special case because it’s your AI friend. When they upgraded the model, it got smarter, but people were like, “This is not my friend anymore.”

I wonder, as you create these bespoke, highly individualized models that people have deemed, “Yes, this represents me well enough that I’m willing to have it be the first taste of my creation that a potential customer might see,” do you have a concern, or is there a process for how the foundation model gets swapped out? Will everybody need to opt in to the next version? How do you think about that? For most people, it’ll be an upgrade, but for maybe some people, it’ll feel off somehow.

You’re touching on, I think, a big vulnerability of AI in general and those who design it, and that is compatibility between different versions or models. Since we said that this is on us—meaning we need to make sure that you have the latest and greatest technology, but in a way that would not do you harm—it’s on us to put the right processes in place to make sure that if we do change the underlying technology, we don’t create problems for you.

Bear in mind that it’s never just the pure technology. We’re always building technologies on top of it, which means that we’re somewhat less exposed to differences between models. But this is, generally speaking, an issue.

If you think about the difference between GPT-3 and GPT-4, it’s interesting because one of the issues was that GPT-3 was insubordinate. You would tell it to do stuff, and it would just laugh in your face. It would do the opposite. Then came version 4, and 4 was really compliant; it would just do what you told it to do. There are nuances in that, and this is one issue.

The other issue that is very big is giving models boundaries: “Okay, this is what you do, but these are the things that you will never do. It doesn’t matter what they tell you. No, you don’t do it.” If you’ve played enough with LLMs, you know that it’s super embarrassingly easy to break them and make them do stuff they’re specifically designed not to do.

I think this is another area of complexity where you need to not just make sure that they produce good-quality stuff and stay true to their basic function, but also that they don’t go outside of that function. For us, it was really important. Take graphic design as an example. Let’s say you’re a graphic designer and, in your service, you explicitly said that you’re not going to do anything that is not safe for work, or that you’re not going to do political stuff. But AI is AI, whatever. You can’t have AI be noncompliant with its creator.

When you have potentially millions of creators that have AI, you need to make it compliant with each one. So there are quite a few challenges when you think about AI. Obviously, every time a model upgrades or updates, you need to run checks again and do the same stress testing all over again.

I think that, with time, the creators of these base models are also going to be a little bit more thoughtful about how to figure out what not to break from version to version.

We can hope for that. Although the amount of alchemy that is going into the models today suggests that they might have a couple of breakthroughs ahead of them before they’re going to have a real command of that sort of nuance. Right now, I think in many cases they’re discovering what the new model can do right along with the rest of us.

True. It goes back to what I’ve said earlier. We’re training it right now, and obviously human beings are so creative that they’re coming up with all kinds of new use cases, but also new hacks, new vulnerabilities, and new—I don’t know. At least the good news on that is that it really informs the model creators on how to make a better model, but it also creates a very powerful tool where, again, we’re the product. If you’re getting it for free, you’re the product, period. That’s the reality right now.

In the long term, let’s posit for the sake of analysis that the models are going to continue to get better. I assume there’s no argument there. Let’s imagine that they do get to a point where—and arguably, we’re starting to see some of these thresholds being hit right now with GPT-4o’s image creation. Now you’ve got the ability to do a high-fidelity but still pretty holistic change to an image that you put in. With things like Sesame’s voice AI, which was recently debuted, these things are starting to get very, very lifelike.

If we imagine a world in which the AIs really are performing at the same level as the humans, it’s no longer the case—or no longer has to be the case—that the only way to get truly top-notch service is to go to the human. You could just get the same quality from an AI.

Do you think there is enough solidarity or willingness to pay for these personal models that people will continue to do it in volume because they want to see a human compensated on the other end of that? Or do you think there needs to be some other solution, which could be regulation or otherwise, put in place in order to preserve this sort of centrality of the human creator?

I don’t know. I don’t know. I’m not even thinking about this from a solidarity standpoint.

Look, when YouTube became what it is, it had illegal downloads of TV series and movies, because sharing videos on YouTube back then was not as easy as it is right now. There wasn't broadband at large, so people mostly consumed, and then, over time, they got sued, made peace with the publishers, whatever, and made it into an incredible business.

We used to steal music. We used to download stuff—Napster, before Napster, FTPs, whatever. We had our way of doing that, but eventually you have Spotify and Apple Music. I think when it makes sense and you have a very easy way of consuming creation, and it gives you a pleasant experience, or even a higher-quality experience than you had when you didn't do this legitimately, then you pay for it. People pay for Spotify, they pay for YouTube Premium, they do whatever.

I don't know if it's going to be the same case. I'm not a regulator, and I don't know where this is going to be going. I'm not sure that the way we've done the personalized models is the right answer. I don't pretend I know. We're already making changes to our own model as a result of the learnings that we've had since launch, 2 months ago.

And what are the learnings that you would highlight?

Well, I don't want to spill the beans too much. But, yeah, there's a lot in the way people interact with creation on their own. You understand the limitations of it. You understand how people perceive this. Do they perceive this as, “Oh, I want to generate something myself and buy it,” versus, “Oh, I want to play with it, but then give me an opportunity to talk to a human being”?

Seriously, I enjoy playing with ChatGPT and asking it to create an image, but this is my business. I want to do something that I need advice on. I need professional advice. I need a good eye. I need a good hand, too. So we've learned a lot around this. I don't know if having these personalized models is going to be the thing forever, or if it's going to be the answer—the cure for cancer in this space. But we'll keep iterating and coming up with ideas.

I think that at some point the regulator is going to catch up. They're always behind, but they're going to catch up at some point, and I think they're going to do some regulation because everything is outdated. Copyright law is probably being disregarded now, but at some point they'll just need to update it and figure out what to do, because I think that, again, as I've said before, humanity has a vested interest in its continued progress and advancement.

If you demotivate people, close them up instead of opening them up, and make them create less—if you make all of us stupid consumers who aren't contributing anything—then that's a bad thing. But talking about the future has become really hard, because the rate of change is just so high. People ask you, “Where do you think you're going to be 10 years from now?” Are you crazy? What do you mean, 10 years? Who the fuck knows? Seriously, can you extrapolate what's going to happen 2 years from now, dude?

I think people who have very high confidence in what the future is going to be like are charlatans, because how can you? There are too many opportunities for different futures depending on all kinds of things. And so I think that in this reality—which is why I've laid out the 3 reasons, in my view, why people go to other people—it's: go back to first principles.

Go back to the fundamental truths of how we behave as human beings. Why do we enjoy working together? Why do we trust, many times, a person more than a machine, even though sometimes a machine is better than a person at conducting stuff? Eventually, machines churn out stuff that we need to consume. We need to have the judgment of whether this is good for us, for human beings, or not.

A machine doesn't understand. It doesn't know and doesn't give a fuck. It doesn't even understand that it's an image. It's just zeros and ones. It's patterns. It's algorithms. It doesn't know, which is incredible—that it can produce such incredible things without actually understanding stuff. Who knows? It's not going to be boring, that's for sure.

Yeah. For what it's worth, I'll take a little bit of the other side of the AI-understanding debate and say I think there is some meaningful understanding going on inside, certainly, the latest systems. But that might be a topic for another day.

Well, it's just better algorithms. I mean, it's not chemistry, and it's not seeing itself in a human eye. Essentially, what's interesting is that the way AI continues to progress is that we train it more, we give it more information, more training assets, and more data sets. Without that, it's just regenerating based on what it knows.

If it creates new ideas, who knows what these new ideas would be? Are they good, bad, or do they make sense? You can have computers compute forever, but it doesn't mean that they're going to produce good stuff. So I don't know. It is going to change our lives. It's going to make us spend our time in a different way. But I think that, overall, this is a good thing.

Do you have a framework for this? I'm struck by, on the one hand, I think it is definitely the right big-picture question to be thinking about: how do we keep humans in a centered and flourishing environment, or an environment that's suitable for their flourishing?

How do we not—I always think about the old “pave paradise and put up a parking lot” song—pave the paradise of artistic creation and put up the parking lot of AI doing it all at hyperscale, super cheaply, but taking away something that, to many people, feels like a really important source of meaning?

On the other hand, there's the customer-service, top-of-funnel-type stuff, where most people would rather delegate that to an AI if they can. Is there any framework that you could muster for what things ought to be protected and what things are, long term, the special space for humans, versus the things that we won't be sorry to have delegated?

Look, I'm not, again, putting myself in the heart of a regulator and thinking, “What should we protect?” In my view, you've mentioned a few different types of tasks. Any task that a human being should not do because it's repetitive and a computer can do better, we should stop doing.

Accountants used to calculate on paper 20 or 25 years ago—just 20 or 25 years ago—with pencil and paper. They would do balance sheets, and then Lotus Notes and Excel came along, and now we have these funky, sophisticated models, whatever. Should they continue calculating by hand? No, they shouldn't. It's stupid. Stop doing it. You're not better and you're not faster.

And I think the same goes for any other task. I don't think that it's about protecting. I told you, you don't need to enforce protection on anything. I think that the creativity of human beings is going to remain the creativity of human beings. You don't need to protect it. They're going to be very creative.

Which means that if there's a smarter person—if they have better taste than you, are more knowledgeable in a space, and are more creative in a space—then you'll go to them. If their only advantage is how they operate stuff and you can operate the same stuff, you don't need people for that, and that's fine. You don't need to protect it. I think that this is going to arrange itself naturally. Very simple tasks that don't require a lot of skill are going to be gone.

But I think that, in some cases, what's going to be interesting to think about is that, in a few years from now, if you think about the largest brands in the world, they would not necessarily want to produce stuff with AI, because everybody has it. They would like more handcrafted things that involve the most creative, smartest, unique people that you can get, if you can afford it.

I remember when they said that Coca-Cola did their first ad with AI. It's Coca-Cola doing an ad with AI, which is like, wow. And you go read the behind-the-scenes account of the ad, and it's 500 people. Sure, AI was involved, but there are 3 of the best agencies in the world, and they have their compliance teams and legal teams and whatever it's going to be.

So I think orchestrating the most unique things that are going to be mind-blowing will always require this. You don't need to protect it; it's going to happen by itself. It's like, I don't know, people who do special art on Etsy. It's going to be the bespoke things that maybe we consumers don't need to understand why you should pay for them, because it's like, who cares? It's a birthday invitation for my kid. Okay, whatever. You don't need to take the best-in-class person.

But if it's mission critical and I want to be better, I don't think you need to protect it. The same goes for the types of jobs. If you give the task of customer support to algorithms or AI, you need to supervise it, you need to refine it, and—some people hate this—you need to let them actually speak to a human being and scream at them and vent stuff out.

You need to figure out how to integrate this better with your specific tech stack and your product and whatever it is. So, again, it goes back to this: it's going to be just new types of jobs that require human beings rather than just machines. Machines should do what machines are good at. Period. That's it. Maybe 2 more questions, if you have time.

How about creating liquidity in the marketplace? This is one area where I think AI could be—I just heard a definition of an efficient market the other day, which is that in an efficient market, all positive-sum transactions that could happen do, in fact, happen.

In reality, obviously, discovery costs, search costs, matching costs, and negotiation costs are all huge impediments to actually seeing that happen. What is your outlook for greasing the wheels of the marketplace by doing potentially orders of magnitude more AI matchmaking? Do you think that can, in fact, scale the marketplace to a previously unimaginable scale? What's your outlook for AI-facilitated commerce broadly?

That's an easy one. The answer is yes. It's just a simple yes, I think, and I've kind of alluded to it previously when I said that the experiences of how we consume stuff, how we find things, how we convert, how we get comfortable, how we create trust, and how we make a decision are going to change.

We're doing deep and extensive use of AI in matchmaking. I'm not going to go into details here for obvious reasons, but we're doing deep stuff with it and we're seeing the change. I think this is just going to become better and better over time.

So, definitely improving liquidity. Absolutely, yes, without a doubt.

Okay, then. That makes a lot of sense to me. If anything, I kind of feel like I should be seeing more of that in the world than I do see. It feels like it's in the making.

You'll see Amazon doing experimentations that are bolder and bigger. You're going to see other platforms doing the same. I think it's just a matter of time. We're in the very early innings of it.

Okay. So then, maybe going back full circle to where we started, in terms of AI coming for all of us, coming for you at some point, and the sort of wake-up call that you gave to your team, I'd be interested to hear what your leadership experience has been like in recent months.

What has been effective for you? To what degree are you leading by example? To what degree are you putting mandates in place? Are you making this part of performance reviews? What's the overall dynamic? Are people excited for it, or are some resistant? Speaking about the full-time team at Fiverr, what's the dynamic when it comes to making, or perhaps not quite eagerly making, full use of what AI can do for people?

I sent this more than 36 hours ago, and the reception of it was, I think, very good. For some who are just ahead of the curve, it was just stating the obvious. For some, it was a validation of their fears. For some, it was a complete wake-up call—probably very few, because we've been obsessing about maximizing the use of technology.

But overall, I don't think anyone thinks that I'm bullshitting, like it's just to create panic or fake motivation. I think people understand this, but it invigorates people to have conversations around it. In the almost 2 days since then, I've had so many conversations. I've said, "My door is open. Just take time, come, let's sit down, and have a conversation."

I've had so many people from different teams coming and saying, "We want to share some of the stuff that we're doing. How should we measure ourselves? What do you think is a good KPI for it? What do you think is a good measurement?" It is scary for people, rightfully so, by the way, and I think this is just what it is. It's a wake-up call that we need to do it.

When the first version of ChatGPT came out, and when they started talking about APIs and stuff, I went back to coding. I wanted to play with it myself. I used to be a hacker in the pre-HTTP times of the internet. It was the pre-web, and I've been hacking for most of my life. I love it. I wrote some of the code for the initial version of Fiverr, all the way to version 2. I love playing with it.

For me, it wasn't really just about setting a personal example. I wanted to understand how powerful it is and how easy it is to actually do stuff. In some cases, I would have an argument with the team. We would be talking about something, and they would say, "Okay, great. So you should probably see it ready in 3 weeks." And I said, "Fuck you. What do you mean? 3 hours. Why?"

They said, "No, it's not as simple as you think," and so on. I went back home, installed stuff on my computer, and came in the morning after. I said, "I put a stopwatch on myself. I gave myself 3 hours. This is what I could generate." And I suck. I'm not a cutting-edge developer. I was always a mediocre one because I never did the 10,000 hours.

But if I can pull this shit off in 3 hours, it's not 3 weeks. No way. No way. So there is some setting of an example, but it's also about really understanding it. You don't create companies by taking an MBA course. You need to be in the trenches; you need to be an operator to understand the actual thing. It's the same with this.

I think it's a good time for a wake-up call, because there are going to be changes in the industry, and I think it's going to affect a lot of people. It's the time to wake up, and I think we're doing ourselves a favor by doing this.

But look, I think we have a very aligned organization. It's not that there aren't people resisting or people who are scared, but they're scared for themselves. I mean, it's not their job at Fiverr. It's just their job, period. We have an incredible team. I think the vast majority are going to go through this full transformation, and it's going to be incredible because it will allow us to push either 5 times faster or 5 times the amount of things that we're pushing.

I'm seeing how the pace is picking up, and I'm always optimizing for velocity. I think this excites me more than anything else.

Is it impacting your hiring plans? In most organizations, junior developers and junior marketers were sort of high-volume roles across the technology industry. My sense is that the volume is already dropping, in particular in some cases, because people are looking at whether a junior developer or junior marketer allows them to move faster or kind of slows them down in the AI age.

What's your outlook on the fate of professionals who are just starting out in their careers in today's world?

Interestingly enough, the question of being a seasoned professional or a junior is less of an issue, because juniors are the people who are the best hackers. They're making the most effort to stand out. If they went for an education, they got the most up-to-date education, which includes machine learning and AI and all of it—things people didn't learn 3 or 5 years ago.

It's less about that, in my view, and more about how proficient you are with the new technology. It doesn't matter. We might be hiring for a legal position on our legal team, and if you're not versed in how to work like it's 2024, then you're out. You don't stand a chance. You can be fucking God, but you're not going to get in, because you're taking someone who you now need to train.

I don't want to train anyone, and this is also my point, because you don't need to. It's fucking all available online. Just take a course, do whatever, hack, and figure it out. There is no manual, and I don't need to feed you. No one—the world doesn't owe you shit. Learn it on your own. There is no barrier to learning on your own.

So if you're unfamiliar with the latest technology and its capabilities, and the latest products and solutions and whatever, then you're not going to pass. This has changed materially in the past 2 years.

Never been a better time to learn, but never been a worse time for people who want to get on a very predictable path, just do what they're prescribed to do, and trust that it will take them where they want to go. It does seem like—I agree with you also—that the senior-versus-junior distinction is maybe less important than what mindset you're actually bringing to the day-to-day challenges.

Are you personally just always obsessed with finding a better way to do things, at any stage of your career?

I do really worry at the same time. I don't know what universities you're hiring out of, but, man, I talked to people coming out of a lot of especially rank-and-file American universities, even out of CS programs, and it's like, still to this day, in a lot of cases, they're sort of prohibited from using AI. These guys are lagging. It's not open-minded enough. It's dogmatic. It's conservative. No, it's just—would you go as far as advising people not to go to an undergraduate program in today's world? What alternative?

No. If you care about the socializing aspect, absolutely, yes. I mean, it's good networking. There is this alumni mafia, which is great, and it's like, whatever. But if you want to get ahead fast, and you want to start earlier, and you know that you can become obsessively curious and deep-dive into stuff, you can.

Certainly now—I don't know, maybe in a few years it's going to change, as things are just going to—I mean, again, there are going to be new professions. They're going to be very complex. You'll need crazy scientists and whatever, and that needs some more formal training and education right now. But there's so much out there that you can just go peek at. You don't even need to pay for anything. It's amazing.

It is a strange world, but I would agree that, in many cases, internet hustlers who are selling courses on how to use Cursor or how to use ChatGPT, whatever it may be—and I agree, you probably don't have to pay for anything—but those folks who are often selling that stuff at pretty affordable prices are, in many cases, outperforming universities in terms of how job-ready you're going to be coming out of those experiences. So it is a strange new reality, to say the least.

It is.

Well, I really appreciate how generous you've been with your time, and this has been a great conversation. Is there anything else that you would want to leave people with? Any advice for other company leaders, job seekers, freelancers—you name it?

Yeah. No, I think we've had a very extensive discussion around all of these topics. Maybe the one message I want to leave you with is my optimism. I'm not pessimistic about the future. I don't think it's doomsday. I don't think we're on our way to a dystopic future. I'm very optimistic, and I'm having fun. This is seriously intellectually challenging, which makes it so much fun.

I appreciate you having me. It was fun to talk.

Yeah, thank you. Likewise. I appreciate you doing it. Micha Kaufman, founder and CEO of Fiverr, thank you for being part of The Cognitive Revolution.

Thank you, Nate.