We saw how Claude released 200 lines of code and the B2B market collapsed. We lost $300 billion in market capitalization.
This is just the beginning, but the direction is already obvious. This is Reid Hoffman, co-founder of LinkedIn, a legendary investor who was one of the first to foresee every major technological revolution. And what he is saying now is important for everyone building a career in the age of artificial intelligence to hear.
There will no longer be individual specialists working alone. Each of us will act together with a set of agents.
What should I do? What should people like me do?
Even those who say, 'Yes, yes, I use AI,' are actually not using it seriously enough.
You often say that each of us will have our own team of agents working for us. Is this already a reality, or are we not quite there yet?
This is already completely possible.
And if a person works a regular job, where should they start to double their income this year?
Interesting question. Probably the simplest thing is...
I'm very glad we met again after a year. We can return to what you were talking about a year ago. The year was absolutely crazy. I see ChatGPT literally everywhere. People are launching projects, building products, and automating businesses with AI. Tell me, is this a real AI boom, or are we seeing just 5-10% of what awaits us?
Of course, the last couple of years have been about AI. But, by the way, I want to show one thing. Recently, I made a Christmas album with music I'd like to listen to at Christmas. And this is just the beginning of the opportunities that AI actually opens up. For example, I don't know how to record albums at all. I have neither the skills nor the experience for it. And this is what I think people greatly underestimate. Yes, everyone hears talk about Claude, coding, and other tools, but people don't realize how this spreads to all areas of human work and creativity. It's not just that each of us will have an agent as an assistant doing something for us. Yes, that will happen too, but the possibilities of programming are essentially generalized reasoning abilities that open up a host of other opportunities. For example, you can say, 'I've always wanted to work with a travel agent who really shares my love for certain types of archaeology. You can become that agent, figure out where to go, what to see, and even book everything.' This is just the beginning, although it's already clear where this is all heading. So I think it won't take long for everyone to understand what I've been saying for the last couple of years. There will no longer be individual specialists working alone. Everyone will act together with their own set of AI agents. And almost certainly in a year, when we record the next podcast, you'll have an AI agent on your tablet that will be prompting: 'Ask him about this.' Or when he said this, clarify like this, and such things will become the norm. That's why I say: not 10%, maybe 5%, or maybe even 2%.
In our last interview, you said we have about two years to adapt. If I'm not technical, if I don't have technical skills, but really want to start adapting right now, what should I do?
Well, the basic level is if you are already actively and purposefully interacting with simple agents, with chatbots. For example, you're thinking about how to take your podcast and media to the next level and asking, 'What should I think about? What new formats to try, what ideas to test?' Or, say, I'm going to a cool place and asking, 'What can I do there, where can I go?' And when I say 'basic,' I also mean the approach to prompting. It's not about typing seven words and seeing what happens. Now, in my opinion, it's very important to use your voice, just talk to AI. We speak much faster than we type. And the result is often much better if you just give it as many thoughts as possible. For example, when I interact with different AI agents, I can say, 'I'm interested in the prospects of thermonuclear energy. What companies are doing something really interesting right now?' And then I say, 'Now write me a prompt to conduct research on this topic.' In response, it returns a two-page prompt. I run it and get a really interesting result. That's what I call the basic level.
Okay, that's the basic level. What about the advanced level, but without programming?
Well, for example, one of the things you start to understand with experience is that these chatbots work very well with roles. So you might have some idea. Let's continue with the thermonuclear fusion example. I'm interested in thermonuclear fusion in the context of climate change, energy, and how it will affect AI. Let's look at this issue from different angles. What would a technologist say? What would a venture capitalist say? What would a government politician say? And what would a nuclear safety specialist say? And then you can ask, 'And are there any other roles I haven't thought of?' And then you ask it to answer, taking on each of these roles in turn. And in this way, you start looking at the topic through different lenses. This approach to AI as a perfect performer of a certain role is very useful because it works even for simple things. For example, when I'm writing text, I can say, 'Argue with me.' That's a simple role. Let it become an opponent, a skeptic, someone who disagrees with my position. And you can go the other way and ask what arguments supporting my position I'm missing.
But then it gets even more interesting. For example, you think, 'How will the next generation of AI affect social media, internet media?' That's a great question to ask AI. But here's what's important. You need to ask it in a way that you ask it to conduct a web research. This is a slightly more advanced level. But here's what people don't understand. They think they're talking to a super-informed AI, but they forget that its training ended 18 months ago, which means in the fast-moving field of AI, it's already a year and a half behind. Right? So when you ask a question like 'What are the most relevant AI tools right now?' you need to frame it as a research task. You need to say, 'Gather data, pull in fresh information, and prepare a report.' In ChatGPT, this is the thinking mode. Because if you rely only on the information in the model, you have to consider that the knowledge is 18 months behind. And that's a problem. It's a real problem.
Can you evaluate my system? We have a system where my entire team of 35 people works. We also maintain a database for each podcast episode, adding transcripts. Recently, we made Claude projects for each social network we manage. We added all the analytics and episode scripts to Claude. So Claude works as our strategist. We gave it instructions, it knows my goals. What do you think, is this a basic, intermediate, or advanced level? And where to go?
I would say this is an intermediate level.
Part of it is that you have AI agents with assigned roles that are constantly embedded in the process and work as an active part of it. They are used from case to case, and this is what raises your level from basic to intermediate. At the advanced level, you start asking different questions: what additional skills can be added, what other data sources to connect. Maybe we need a separate agent that will talk to all other agents or projects and look for cross-cutting patterns, what we can learn from what works and what doesn't, are there common patterns that will suggest what to do in the next month or two. Understand? That is, to reach a meta-level where you analyze the process itself and look not only at internal data but also at external data.
Like other podcasters, right?
For example, you noticed that another podcast did a cool thing. I'm just giving a random example with Cryptopunks. And you think, maybe we can also use this somehow.
Yes, gather ideas from other fields for you.
And that's the point. It's important to think of agents as scalable intelligence. It processes information at the cost of electricity, that is, at the cost of computation. And computation is essentially electricity. So everything depends on the vector you give it. You don't want it to calculate endlessly. You can easily spend $200,000 on AI computations that won't bring any benefit. But if you direct it to a clearly defined useful vector, then even a conditional $10,000 on a serious amount of computation can suddenly give you 15 really valuable results.
And that's already obvious value. This is the second time Reid and I are talking about the future, AI, and the changing labor market. But the paradox is that the more digitalization there is around, the more valuable offline, live communication and space become. And Dima and I are very happy that we bought real estate in Hawaii a few years ago. For us, it's also a place where we can go with family and friends, and at the same time, it's an investment. The condo is growing in value, it can be rented out. It's a fairly liquid asset in dollars. If you're thinking about where to invest money offline, real estate remains one of the most understandable and stable options. The sponsor of this episode is the international real estate analytics agency Neginsky. And for those who are planning to start investing in real estate this year or are looking for a property to live in but don't know where to start, they have prepared two new catalogs. The first is a catalog for the UAE, which includes the top five properties with an entry threshold from 2.5 million rubles. It's suitable for those who are just looking at the market and want to understand in literally 20 minutes whether it's worth moving in this direction at all. The entry threshold for the UAE market looks very attractive. For foreigners, interest-free installments are available for up to 6 years. Rental income can reach 12% per year, and property value growth up to 40% upon sale. And all this is tax-free. In the Neginsky catalog, you can select properties with an expected yield of 9% per year and post-installment payments of up to 30% of the cost. You can pay within 3 years after receiving the keys, that is, you can pay from the rent. And by the way, in the UAE, service companies can rent out your apartment. You don't need to get involved. The second catalog is for those looking at apartments in Moscow for living or investment. If you're thinking about buying an apartment in Moscow, but a 20% mortgage rate scares you, don't rush to give up. High rates reduce demand, so developers offer their own conditions. The Neginsky catalog includes seven residential complexes with interest-free installments. Moreover, this is not economy, but business and premium class. In the current market, such offers are indeed a rarity. It's important to catch them in time. Both catalogs can be obtained for free by following the QR code on the screen and studying them at your own pace. And if you want to analyze your specific task, buying for family, investment, or just to save capital, the Neginsky team consults and selects a solution individually. All links are in the description. Go ahead, study. And we're returning to the interview with Reid Hoffman. What should those who listen to us but work a regular job with a salary of 80,000 do? How can they use artificial intelligence to double their income this year?
How to double your income? That's an interesting question. Possibly the simplest path. Everyone who runs businesses or invests in them already understands that they urgently need AI transformation and AI specialists. And so part of the strategy is to start demonstrating your involvement and knowledge in the field of AI so that you are easy to find, whether through LinkedIn, social media, or other platforms. People are literally saying, 'We're looking for such specialists. We need this transformation.' And so I think that highly paid positions are not just something like 'We need an AI researcher for huge money.' But now, I think that this year we will really start to see more practical applications. And businesses are asking questions: 'How can I manage my company more efficiently? How to better analyze the supply chain? How to analyze finances, risks, marketing, sales? How to do all this better?' And they will start actively seeking answers. Yes, first they will go to their internal teams. But then many will think, 'I've been doing this my whole life. Learning something completely new is difficult for me.' And this is where it's important to start transitioning to the new.
