AI Is Moving Faster Than Society Can Handle | Adam Spector

There's so much money getting poured into AI right now, and it's being financed with debt. And the companies who are doing AI, Anthropic, Meta, folks like that, are willing to pay so much for that cash that it's raising borrowing costs or will raise borrowing costs for everybody. If the world essentially becomes compute and essentially all the money goes to compute, that means you can't get a mortgage for your house because it's not like you're not willing to pay, 10% mortgage rate for your house, but Meta might be willing to pay for that because they know what the ROI is exactly on new compute they can put online.

I also went through the transcript and stuff, and some of the things that you said actually became reality. Some of those didn't. I agree.

I agree. My AI, when I was reviewing the transcript, was like, you did a pretty good job predicting most of this stuff. I was like, fuck yeah, I did.

Fuck yeah. Let's just jump right in that one. AI was supposed to save us time, but why are founders working more in San Francisco?

I think because things are— so what AI has done and what I've seen for myself, I don't know if you both have experienced this, but what I've seen for myself as well, and I'm one of those founders, is AI is making people who are already hard workers are working harder because essentially what you're doing, and I think this is the biggest change and difference for a lot of people actually today, is that AI has— so let me back this up. I think people who are going to be the best at using AI today and in the future going forward are the ones who are really good at managing. Essentially the ability to delegate, which is something I talk about a lot and things that I care about anyways, but the ability to delegate and the ability to manage are arguably the only skill sets, plus some idea of understanding processes and systems and having new ideas. So creativity, process and systems, plus delegation is arguably the only generic grouping of skill sets that are going to matter. That's probably it. Now, what you end up finding, of course, are people who are founders in Silicon Valley. In many cases, they might not be good managers, right? Some of them are actually total terrible managers. I've seen many of them and they're actually really, really bad, but they are really good at delegating and handing things off. Say, I got to go do this, I'm going to go do this. And especially when you're building things in code, you're really good at building some certain thing here, have it run, go to the next thing, have it run, and keep going back and forth all the time, back and forth. The difference is with AI now you can do that even faster. So like, you know, for me, I'm like, I'm managing all these different agents and they're like coming back, oh, got this done, got this done, got this done. I'm like, oh God, he figured this, then I gotta context switch to this one, go to context switch to this one. And you just keep that going. And so you're working harder, but you're getting so much more done. And so that's, that's why, that's where I think it's going and also the skillset.

So you were saying in 2026, the A-player actually someone who manages 20 digital workers.

Simultaneously. Yeah, 20. And I think A players are probably 100. I think it's really hard to manage. I'm probably managing like 5 to 10 on a regular basis and like, that's pretty hard for me. But I think like A+ players are probably 20 to 100. Like they're just crushing things and you're not really managing them, right? What you're really doing and what's happening for, I think a lot of people is like you're managing a few senior agents essentially, and they're delegating that work out to sub-agents, which once again sounds a lot like a big organization. You have a CEO at the top. You have Elon Musk at the top of SpaceX, but he's not welding the rocket, right? That's someone many, many levels down who's being managed by someone who's being managed by someone else being managed by someone else. We're rebuilding the same thing in code.

You know what I'm thinking is funny is last time we were talking about middle management is going to be not a thing anymore, and now we're thinking that with AI. So when do you think they're not going to need middle managers either, these AI agents?

You might, you probably, you're right. There won't be, there will be a time where you probably don't even need the AI middle management of the AI layer either. Um, it'll just be sort of direct, but you probably need an orchestration layer of some sort to still sort of decide like who gets what and who does what. Now, on the other hand, you could argue, and this might happen too, you don't really need the middle management layer in a sense, because you'll be able to tell AIs, they're getting so good and so smart that you don't need to have multiple AIs. who are specifically targeted. You don't need, like, in our world at Shore, I don't probably need to have an HR-specific, like, human resources-specific agent and a finance-specific agent and a compliance-specific agent because one agent can kind of keep all of those roles in their memory brain in a sense, and they can context switch between them. So why have multiple ones when then you just actually increase the This, the cost of like delegating things out. So we might get to a point you actually aren't managing hundreds of agents. You're just managing one agent who has like 100 different threads going on at any one point.

Have you noticed the— by doing that, having more generalistic, I say, agent, like an A++ player agent versus multiple ones that are A players still, but not as good agents. Is cost— is more cost efficient?

Unclear. I'm not probably an expert when it comes to the cost efficiency side. Um, I generally think that the current cost of AI for the large majority of companies, and I mean, in this case, to be clear, I'm talking about like non-hardcore AI tech companies, right? Like sort of people who are using AI, making changes, things like that. For the large majority of companies, the AI agent is so much more efficient. and low cost compared to all alternatives, it doesn't really matter the cost breakdown. Like, sure, maybe optimise some of the work later at night when it's cheaper, a few other things, but at a high level, it probably doesn't matter. The sort of scary thing is when you start looking at the cost breakdown for how much work a human can do in, like, an 8-hour period, let's say, compared to what an AI agent can do in 24 hours, even if you're sort of optimising when to do it. It's such a massive cost differential, even if the cost of running that AI goes up, which is an interesting discussion too, but like, it almost doesn't matter. Like, the cost sort of doesn't matter right now. Like, it's sort of a silly conversation to be like, oh, I want to optimize my $200 spend on Claude. Like—

Use the time differently.

Yeah, like, it's not a good use of your time to optimize that. It's better use to like, how can I do more? How can I spend $400 or $1,000? Because it can do so much more. Like, that should be your question, not like, how do I save $20 by overoptimizing it?

Adam, I have a question. So right now, because I want to see what has developed or what has changed in Chore over the last 6 months, but please also for the listeners who have not listened to the pod before, please introduce yourself. Like, at least what is— what is— what are you aiming to do with Chore?

Yeah. So, so, so I'm Adam Spector. I'm a 4-time founder. My first 3 companies were all venture funded, raised about $30 million or so in venture funding over my career. Chor, which is my current company, is fully bootstrapped. Essentially run the operations for startups and for any business, really any SMB. What operations means is like the back office stuff, right? So human resources, finance, compliance, doing all the stuff that is a chore that you didn't start your company to go do these things. So why are you spending time doing them? Like, your time is so precious as a founder, and if you were spending time trying to Figure out how to run payroll or getting on a phone call with your payroll provider because like something broke down. You're literally wasting time where you— that you should be spending. That's the same, actually almost the same example as like what we just talked about, how you're optimizing your costs and AI spend. It's kind of the same thing, right? I come across all these founders who say, oh, like I can do this myself because it's not that hard. Well, like, first of all, it actually is relatively hard. So you're wrong. You just don't know what you're talking about. But second of all, It's the cost to you to like to save that extra little bit of money to do the back office work yourself is so expensive in comparison to spending all those extra hours that you have building a better product, getting more customers. And like you are making a distinct trade-off, right? When I'm spending time with both of you, I'm making a trade-off that I can't do anything else right now. And so like that's a trade-off I need to be continuously weighing. And every founder should do that. So that's, go back, that's what we do at Chore. Hire chore.com, come find us, we'll go run your operations for you so you can go do your best stuff. I'm an investor in 200+ startups, um, a lot of later stage stuff, which I know is one of my predictions from last time too, but a lot of later stage stuff recently, um, because early stage is getting freaking difficult.

I love that prediction, honestly. That was really, really good. Right on the dot. Also with the, I liked it with the context switching, 'cause in the end it was like, chores, right? People like, for example, doing your legal stuff or HR, etc. The thing is, it's similar to an AI model when it needs to do context model, like context switching, it needs to empty entire like cache before, right? Similar to like the brain, but we can't reset as fast. So when you're doing 2 tasks at the same time, it's horrible. So if you now have to think about all the payrolls, etc., that just like takes your eyes off the ball, which should you actually be doing, like building your startup and pushing forward.

Correct. But a lot of people, look, a lot of people want to waste time. Like the simple The fact is we all do this where it's like, well, I have this hard thing I need to do, but I have this other easy thing that I can do, and I'm going to go do the easy thing because it makes me feel good, like I accomplished something, but I actually didn't do the hard stuff. And we all do this, right? I optimize my credit card points. It's a silly thing. I've actually stopped. I've actually moved away from that significantly because it's a huge waste of time. And it doesn't, it's not going to move the needle in my life in any significant way. It's like cool to see, oh, I have all these points, but then there's a whole game to use the points. It's just, it's BS. It's a waste of my time. We all do this, but we should fight against that thing to do, to do that.

Were you using that with Amex points?

I had Amex points and I have Chase points at the main events that I do. Those are the 2 best ones out there. So like, yeah, those are the right ones to use, but just get cash back. Like it's the net-net isn't worth it. But, and I'm happy to talk about credit cards, we can have a chat about credit cards and all that game there too. It's very fascinating. But once again, the key point is founders especially need to be thoughtful about where they're spending their time every single day. Don't waste it on things that are a distraction, or if you are, just acknowledge, like, it's fun to distract yourself in that way. And the other thing, the key point you made, Lukas, about like context switching, I think is really important. And that also goes to why people are feeling probably some burnout with AI. is humans are bad at context switching overall. Even if you're good at context switching, it actually is very mentally draining to do it. And now if you're context switching across like 10, 20, 30 agents running multiple different things at once, that is tiring. That is tough. And so only a select few people are really good at that.

Do you think these are going to be the AA, well, A++ players in the next decade?

For sure. I really think the only people who are going to be super successful are the ones who are fantastic at context switching quickly across a wide array of spaces. So having almost deep generalized knowledge, which I don't even know if that's an oxymoron, but having deep knowledge across the board as much as you can. In making intelligent decisions. I mean, to be clear, I'm talking my own book in the sense that I'm relatively proud of the fact, like, my dad always used to say, Adam, you need to be a specialist. Like, that's who's going to make all the money. Go deep into a topic and you become really great. And he was right for a really long time. I kind of always fought against that. Like, I'm a bit more of a generalist, and I actually think a generalist is like the skill set you need to have today where, like, I know a lot, like a little bit about a lot of things. Overall, I can go deeper in some other areas, but like a little bit about a lot of things. And that enables me to go cross-functional in a lot of different ways.

Enough to be dangerous.

Enough to be dangerous. But now with AI, I can become that expert really quickly. Like, that's the cool thing is like AI enables a little bit like The Matrix, right? Where someone would plug their head into like the Matrix and download like the plan to how to fly a helicopter. AI is kind of like that. Like, if I want to become an expert on flying a helicopter, I could download the whole instruction manual from AI, have it customized to me. I can't jack it directly into my brain right now and have me able to fly, but like, I don't know, that's 5 years away. Like, seriously, like, I'm joking. I say that kind of seriously, like, but it already gives you the roadmap to become an expert at a high level and you can go learn about these topics quickly and go back and forth with AI and do this whole coaching you just couldn't get before.

Are you guys already doing harness engineering at Chor or are you still vibe coding? What's the— how are you— what's your day-to-day?

So good question. We're more vibe coding things. We're starting to build some other stuff. The thing is, even if you're building stuff now, I really worry for most companies about whether they can stand apart in any way. So a lot of the ideas that we had for what we're going to build are still really valuable and great. And I think literally will replace a significant portion of the work that humans do today, which is okay. Truthfully, the work that most humans do today is copying and pasting, not just a chore in most jobs. That's a bad use of people's time to copy and paste. Like, no, I don't think anyone gets a lot of fulfillment from copying and pasting. AI, there are tools coming out from the major players that are making it easier to automate the copy and paste function. So we shouldn't be spending time building that. We should be spending time building the real skills, building the deep knowledge of like where things break, and then arguably becoming like the authentication, not authentication, sort of the accountability layer. And trust layer where you still need humans involved, where it's like, hey, I want someone who can deal with a really weird, hard edge case. I wanna go talk to someone to verify this is the right decision. An example might be with lawyers, right? You can have obviously AI draft all your legal agreements for you now. It's really good. But you know, like if you're gonna go through a lawsuit against someone, you probably wanna have a human review it and make sure it's like really good. You don't want the AI hallucinating, which does less of now, but you still don't want that. So. I think that's where Chor and a lot of companies like ours will end up.

So in general, code is becoming cheaper and cheaper, whereas verification is becoming more and more important.

It's a great way of putting it. Yeah, I think so.

So currently you said you're in an awkward in-between right now with Chor. That's what it sounds like, at least. What is it? Where are you? What has changed now in the last 6 months in that regard? Last month, last time you were building all that stuff. Now you're in the authentication layer. How does it work? Could you run us through this?

I mean, the authentication part has still always been a core part of what we do, right? The humans have really just been doing it. We're building out the stuff so AI can do more of the day-to-day tactical work, and the humans then will do more of the verification or the weird edge cases or the harder things. That's sort of one big change. And the other big one is more and more rethinking who are the right people on the team to do that work in general, in making sure we have team members who understand the work is not the tactical day-to-day, it's the higher-level, higher-leverage work. And so making sure the team who gets that and who's bought into that vision, because a lot of people aren't bought into it, right? This is a really big change. And what's happening to Chor is no different than what's going to happen to every other company out there, right? Like, if people on my team are listening to this and like, are there people in general listening to this? Like, look, The fact is like what we're doing at Chor, we could be wrong, of course. So there's that, but let's presume we're right. If we're right, what we're going through is the same as every company is going to go through. And so you need to find a way to level yourself up to prepare for this future, not fight against it, not say, well, Chor is wrong. This is the wrong direction. Therefore, I'm going to, I'm going to go do a different job. And like, I'm going to go find another job that's going to pay me a lot, a lot of money for my copy and paste skills. Like, yeah, there are companies that will do that still. Correct. Correct, Mark. There are companies that will pay you for that, but that's going to last a year. Like, this isn't maybe 2 years if you're lucky. This is not a long-term job survival mode. This is you need to retrain yourself now. And if you don't, you're going to be out of work.

How do you align your team that they're still buying in the mission? Because it sounds like, how do you— do you have someone externally? Are you taking the lead in this?

It's really tough. It's not something we've perfected. I know we do exit interviews with everyone who leaves and everyone always says we have a fantastic culture. People love working at Chor. I've tried really hard to make us a company that people enjoy. Like we are fully remote. We are completely flexible about when people are doing their work. We're very transparent about how compensation works, benefits work, all these things. But it's also still difficult, I think, for anyone to see this change that occurs and to understand it. And, and of course, you don't want to hear it from me because you might think I disagree with why I'm saying it or my own reasons for it. Like, I'm doing this because I want Chur to be a long-term successful business. And like, and so it's hard. It's hard for the team to understand that sometimes. It's difficult. We try to explain it all the time. Once again, the people who get it are the ones who will stick around long-term and the people who don't, I literally, I sincerely, sincerely wish everybody as much luck as possible. Like, I want every single person— I view everyone as sort of part of the Tor family, and I want them all to be massively successful, even if for whatever reason they don't feel like Tor is the right place for them.

Adam, you told us last time that early-stage VC was dying because software can be destroyed overnight. Has it actually made hardware more attractive?

It has. And I think it has. Certainly you've seen a lot of the early-stage VCs moving more into hardware. This has happened a lot. I think even if you look at the Situational Awareness, Leo, Ashton Brenner's fund that kind of blew up, but if you look at their actual bets, I think a lot of them were directionally correct. Most of it was in hardware, shorting a lot of the software stuff. Generally speaking, that's probably the correct direction for most of these things. And I still think that's the case today. It's like hardware, but even hardware, to be clear, I don't think that protects you Long-term, right? So the problem with hardware is it's a little, maybe the way to put it is it's a little bit like if you look back, you know, 30 years ago when people started outsourcing manufacturing to China, right? That made a lot of sense. That was like the easy thing we can outsource to China. It's lower cost. We'll move it there. It's sort of the same thing today with hardware where it's like tough to do the hardware right now for AI because it's just atoms are different than bits. It's a little bit more difficult. It takes a while to build those things up and it's just further behind, right? Whereas you have this massive training corpus of all of the coding data in human language over all of history to train on. Different story than hardware stuff. It's just, what do you have examples of? Internal combustion engine? You don't have that much data and there's not a lot of real-world training data really as well. So it's harder. But like, we're going to get to a place in the next 5 years, maybe 10, maybe less, where you have robots building robots, probably in their own specialized tax-free enclaves or something. They're already in China. I think they're called dark factories where they are—

Crazy.

Yeah. There's no lights. There's no people in them. That process is going to accelerate. And so hardware is going to go the same way as software. It's all going to get eaten up by

For me, it's interesting to see, like, we outsourced everything to China back in the day because the labor force was very cheap. But now the labor force is not the cheapest. It's just they have everything automated there and that, like, the dark factories, which makes it then cheaper. But now the West has like a very hard time to catch up on that one because they don't have this infrastructure yet. Well, you're also missing an interesting point, Lukas. The Chinese are actually outsourcing to Vietnam also, like where actually things that cannot be automated yet, they are outsourcing to different cheaper countries in Asia.

Sort of, although I don't think Mark's— that's totally— I mean, yes, but I think a lot of what looks like it's outsourcing to Vietnam is actually avoiding tariffs. And so they're sending stuff to Vietnam to then send it to the US with lower tariffs, but it's not like really being manufactured there. So not totally sure about that one.

Fair enough to a certain degree.

Yeah. So, but, but with that said, like the, what China has done is very impressive in the danger for the West. And right, you're seeing this in Europe, right? All the car companies are getting killed. It's like you didn't invest in any of these underlying tools. And so like, it's not just a better factory, it's the batteries, it's the actuators, it's the motors, it's the rare earths. It's all these things get built up and it's this whole stack, right? It's called like kind of, there's a great post about a year ago called like the electricity stack or electricity, like the electric slide, I think. And it added all these things up and it's all these pieces built up together. It's really hard to replicate those things.

The interesting thing is you mentioned rare earths. They found last year or this year a massive deposit of rare minerals in Germany for batteries. It's, I think, triple the amount that is in China or anywhere else in the world. It's insane.

But so here's the problem. So rare earths are not rare. Rare earths are just really hard to refine. And dirty to refine. So yes, you might have found a lot in Germany, but the Germans have to get over the environmental aspect and say, we're going to— and it's dirty and polluting. Right now we outsource the dirty and pollution to China because then we didn't see it. It was out of sight, out of mind.

100%. The thing is, at least there's a foundation that could be used. That's the cool thing there in terms of when you're talking about the EU, like we're so dependent on other— we don't have a lot of oil, for example, and all this stuff. And it's super interesting that we could build it here, which is, I find cool.

Well, what's weird for the EU is they did not. So you have all these constraints, right? Constraints usually make someone stronger in many cases. And for reasons that are beyond my ability to fathom, like the EU did not do that. Maybe because they had a good deal with Russian gas for a long time coming in and relatively low-cost oil kind of protected by the US coming from the Middle East into Europe. So they didn't have to make changes. Whereas China was like, hmm, the US controls the world's oil in a sense. They're a potential adversary of ours. Let's get off oil. And so they use that constraint to change what they did, and that has helped them leapfrog the West.

Talking about— I think I agree. It's crazy cool how they did that with the 30-year plan and the 100-year plan, I guess. On the other hand, I want to come back to something that you mentioned earlier. that AI will become more expensive in the future? Give me a, 'cause I heard that in other podcasts as well. It's like, at some point computational costs and use the tokens while they're cheap, they're going to be much more expensive. When you, once you're hooked, it can be like, all right, it's double now. And you'll be like, okay.

So I think that's partially true. I think there's a real chance that they can extract a lot more profit from things. You know, like as an example, I'm testing right now to like rebuild a lot of the work that like Sort of, I'm building, you know, there's a million chief of staff companies. I'm trying to rebuild sort of my own, not for the work we do. We call it Tor Chief of Staff, but like very different chief of staff work, like just manage all these things in my life. And once I rebuild it in something like, okay, even if I'm paying $200 a month for it, for me as a business person, that's an ROI that's worthwhile because a chief of staff is $200K a year. So like, that's a massive difference. And sure, make it $400. Okay. I don't totally agree though, and I'm not sure where this lands, but at a high level, there's 2 kind of competing elements. One, you have open source models, right? And everyone's It's like, okay, these are way cheaper to run, blah, blah, blah. They're not actually cheaper to run because the question is about compute. And the people who have all the compute are Anthropic and OpenAI, Meta, Google, and then Cursor/Grok/SpaceX. They have all the compute. So it's really about compute. And if they have the compute, they kind of control these things. And so you're actually really paying for compute. In a sense, and that's where the real costs are going to go. At the same time, the models themselves are getting much more efficient, but also smarter. So on one hand, you have their weight, the cost, the intelligence per token. I'm going to hope I get this right. The intelligence per token is essentially lower or sort of higher. They're smarter per token, but you are using more tokens because you have more intelligence and you're willing to pay for the more intelligent LLM. In high-level TL;DR, I think we might see the price stay kind of the same. Maybe it's going to go up a bit because we're just compute short right now and there's just a big buildup that's going to happen over the next few years and it just takes a while for some of these factories to come online. But once we do, once again, models are going to be way more efficient. And yeah, you might be doing more with them too, so you're spending more of your dollars on it, but you're getting more out of it.

Do you think LLM— I mean, these are all LLMs, but we also talked about the bits versus the atoms before. And I was thinking world models. I think Nvidia is doing a big play in that right now. Do you think, do you already, have you noticed in San Francisco that they're, let's say the hardcore nerds that are in the intersection between AI bits and atoms are focusing on that and they're thinking this is the new next thing that will change the world forever?

I don't know enough of the world model people to really comment in an intelligent way there. It does seem— look, what I hear about at least, I think there's a very good argument LLMs in their current form could easily take us to recursive self-improvement. We're probably already there. We're going to get to AGI. Sam Altman basically said today, by the end of this year, we're going to get to AGI in his mind. All these sorts of things. So LLMs can probably get us there already. World models might be really important though for the next big leap for AI, which is the atoms side of it. We have AI embodied in all parts of our house doing all the things in our lives. And essentially we become truly like WALL-E world where we're just sitting there on Instagram and getting fed food and getting fat and lazy, which people are already thinking.

That is so good. That is such a good movie when they're in the ship, just sitting there. That's a classic movie.

I can order food right now. I can get any media. of any sort that I want right now in front of me immediately. And I don't have to work all that hard to make that happen. And so, and AI is going to start turning our— we've already turned— large majority of the population have already turned off their brains, right? Maybe before they were getting it from like Fox News or CNN or whatever their news source was, and they were kind of just listening blindly to that. It's not that different though with AI. It's just worse because it can talk to you and make it sound really— Like you're really the smartest person in the world.

I think it is crazy how far we've come in the last 4 years in the sense of ChatGPT. I think that's when it came out, right? 4 years ago?

About, yeah.

Up to now and what the capability has been. And even the last 6 months that we spoke already has changed a lot. One thing also, that was one prediction that you were kind of wrong, or at least the timing was off, Cursor. I was like, ah, But I see your point, but like, what do you think of that? I would love to hear it.

Yeah, I mean, look, I think, so my Cursor prediction, for those who don't remember that, that one was basically like Cursor was going to not do well because it's going to lose the race to the big guys who are out there. What I failed to predict, that they have a massive amount of training data underlying what they do, and like that coding data is really valuable, and they were building some of their own models. And so a model player, AKA Grok, needed to catch up. Had a lot of liquidity because they just went, they got merged and bought by SpaceX, which then went public, had liquidity, said we need to do something to catch up really quickly. And there was this valuable option here. So I would say in some ways, like, I was right and wrong, right? They're not independent. They failed to be independent and become a— they're not going to be a $2 trillion company like Anthropic is today. So they still had a great outcome for everyone involved. But

was it a cash deal?

Cash and stock, I don't know. Either way, they made a lot of money. I think it was like a good 3x over what their last private market valuation was. They did well, but they needed to catch up. So that's where I think I was right and wrong. They failed to be independent against the big guys, but they still had something pretty valuable they were sitting on that they could leverage to sell and make a lot of money for all investors.

I agree 100%. It was something that was really, really, really hard to predict, like the probability of that all happening. Just think about how crazy it went from SpaceX is going public, Twitter, like X is going in there as well. Then Grok is going in there, like all these like super— I don't know, I would have never predicted they're all going to be rolled up into one. Like, what are the chances of that?

Correct. But I— and so a few things there. One, going back to the Cursor point, I still don't know that many people who are like using Cursor in a major way still today, necessarily, at least the way it was. Like the hot heat of the moment. However, I'm using GrokBot right now. It's a really cool product and does some really cool things. I'm very impressed by it. But they might not have been able to sustain that as an independent company and their valuation might have been lower because the value extraction they had was something that only a company like Grok or SpaceX could really make use of. The reason also that I think this is sort of an interesting world we're going into and a worrisome one in some ways is I think the value in AI is going to accrue to the biggest players. So as you get bigger and as you get more compute, you essentially start owning the ecosystem. And that is nearly impossible for anyone else to compete with unless you have the deepest pockets in the world. And even then, it's really hard. Like, Google's essentially failed right now. Meta is making a really interesting play, to their credit, Mark Zuckerberg's

credit. What is Facebook doing?

I mean, they, you know, he's dropping people $100 to $500 million deals to buy people to come on board. Yeah. But like, they've launched some really good products that seem pretty interesting. They have a mass amount of compute. They actually have a real business. They also have a mass amount of training data, right? Like, remember part of what like people would say for Grok was like, oh, Grok is trained on all the Twitter data. Facebook's trained on all the Facebook data, Instagram data. It has a billion, basically everybody in the world who is online has a Meta account on WhatsApp or Facebook or Instagram. Like, they have way more data and they have more compute than almost anybody, and they're making a real push on AI. So like, they have some real fascinating plays out there that other people don't. But either way, like, them, Anthropic, OpenAI, that might be it. And all the value is going to accrue to those folks and the people who serve them.

Could you explain why they will own the whole ecosystem?

Because one, if you, if AI becomes critical to your, what you're doing, right, in your day-to-day work, and AI is predicated on having compute, right? You can't do AI without a computer to run it.

Mm-hmm.

You need to go to the places that have compute. The big players have locked up compute for the next, for all of time, and they will pay more than anyone else by a long shot to do that. So that's One reason you're going to get locked in there. 2, the more data they get because you're using them more helps them train their models more. And the more they train their models, the faster, the further ahead they get. So it's sort of this self-reinforcing cycle where you're like, I want the best models. Oh, and by the way, I also need compute. So I'm going to go to Claude and Anthropic to get that. And by using Claude and Anthropic, it trained it and then it keeps getting better. And it's a self-reinforcing cycle that takes them away to $2 trillion or $10 trillion valuation. As they suck up essentially the entire ecosystem of all jobs in humanity. And there's not a lot of room for everyone else. So hopefully you're a shareholder of them and then you hope government taxes it somehow, taxes the computer, taxes the value they're generating because you're probably going to have massive unemployment.

What do you think is going to happen first? A $1 billion software company or a useful human robot in every home? What do you think?

What do you mean by $1 billion software company? There's a lot of $1 billion software companies.

I mean, besides, which one's already known? One person besides—

Oh, one person. One person.

One person.

Yeah. We've already seen the one person, maybe not. I think we've already seen the companies that are at least valued at a billion dollars or more with one person or one or two people. There was an article in the New York Times probably 6 months ago about some guy who built a medical company, like a hims and hers.

Yes.

So there's that type. But also there's scams. A lot of the Kardashians have these billion-dollar valuation companies that are basically a lot of strong marketing and generic Chinese manufacturers making the clothing, but great marketing from the Kardashians. So there's, I think, already a lot of these companies out there.

What are your thoughts on— last time you predicted robots are going to be here sooner than expected. Do you have a more concrete timeline when we're going to see those coming into our daily lives?

How are we defining robots, Marc?

Okay. Uh, they don't need to be walking. They don't need to be walking. I don't care for that. As long as they're, let's say, help in the kitchen or do something that's in the physical world. Besides a vacuum, Mark, because that's like robots. Oh yeah. Okay. Yeah. Not a vacuum. More complex than a vacuum. Doesn't have to be generic that it can do all kinds of chores.

I mean, that one, I wouldn't be shocked if in the next few years we see ones that are like really good at a lot of these things. Whether they're going to have wide adoption is a bigger question. This is a weird place where human labor is pretty cheap in a lot of places, actually. A lot of the people I work with who live in India, they have a chef, a personal chef. They have someone who cleans their house. They have a driver. All these things because it's so cheap for labor. Now, labor in the US and Europe is more expensive, and so people don't have that. But it still costs a decent amount to get Like, to trade that off for a robot that's like kind of only going to work, that maybe has privacy implications, that like, I don't know. But I will say, like, I think there's lots of other robots that are going to be out there every single day, right? Like, we're already seeing all the self-driving taxis. Like, those are, those are robots and they're preferred. They are the preferred, like Waymo is the preferred method of transportation in San Francisco over Uber.

Are they cheaper or why?

No, they're more expensive. Which is, but people, they're more expensive because people prefer them. People prefer them because they feel safer in them. They're cleaner. You don't have to talk to the driver. If you're a woman, a lot, I think a lot of women are scared by some random dude driving them around. You're not worried about that in a Waymo. So like, there's a lot of reasons people prefer, and it's an easier drive. Like if you're someone who's willing to, if basically you're like a millennial, like, you know, anybody in our generation-ish who's like, The moment they get in their car, have downtime, they look at their phone, right? We all probably do that. If you have a car that doesn't jerk and stop and all these other things because somebody's driving too fast and is a bad driver, and the Waymo's super smooth, that's also worth it too. You don't show up with a headache at your next destination.

It's so crazy because we don't have that here. Europe will be a lot behind.

Yeah. Well, and Europe's not going to get Siri AI. It's just one simple example. Literally, I'm running Siri AI on my phone right now, the beta version. And

like, how is it?

I mean, it's good, right? I mean, it's for anyone who's using a lot of AI, it's not like that revolutionary. It's sort of better being tied into the other things that I do. And it's a legitimate search engine now. Like it actually like works. You can ask the questions, it does real research and follows up. So definitely way better. But

is it subscription or is it just included into your phone?

It's free. So a lot of people will use it. I'm sure it'll have huge adoption. And so for a lot of Americans, this will be the first thing they really use AI in a significant way, probably, because it's just built into what you do. But if you're in Europe and you're not somebody who's like AI-pilled, you're still doing search, Google search.

Have you ever used Whisperflow? What are your thoughts on that?

Not only have I used it, I'm an investor.

Oh.

So yes, big fan. Yeah. Like, I don't— I use that in another tool called Codetypist, which I'm not an investor in, but it basically finishes the sentences that you're typing. Like, you're just typing and it's just like auto finishes it because it's using a prediction engine. Between those 2 things, my typing is down probably like 80%.

I like Reverse for Flow. I just started using it a week or 2 weeks ago, something like this, because I was like so tired of just like repeating certain emails. So I was like, all right, I'm gonna start talking. And it's— and I'm a fast typer. I'm typing with 10 fingers, but just like nothing beats talking, like, to be honest.

So here's— you guys got to catch up to me. I'll give you some stats and see if you can catch up to me the next time we have a conversation. So I'm at least based on their insights, which who knows how true they are, but they, they say I'm in the top 0.1% in my words per minute. I know hopefully I'm not talking too fast in this podcast,

but no, that's

good. 154 words per minute. And I've dictated getting close to like 700,000 words dictated right now. Total.

In total or all time?

Since I started using it. I mean, I was a very early user, of course, but like,

okay. All

right.

So that's a lot of words.

You know, think about all the emails you're typing every day and messages you're sending to AI.

Are you using it also on your phone or just on the laptop?

I use it a little bit on my phone, mostly my laptop, but part of that's because I'm not mobile that much, like in a sense, like I kind of stay at home and so like I have my whole setup here, so I just, I'd rather do it on my computer than on my phone if I have a big screen.

That's awesome. I just noticed sometimes, I don't know, maybe I'm just not talking properly. It just, I'm like, this is not what I said. This is not what I said.

Keep practicing with it. It should get better for sure. Like you gotta train it a little bit and I'm sure it's like, you guys have a little bit of an accent, so it's a little bit harder.

But also we're switching in 3 languages. We're switching between English, German, Spanish. At least for me, it needs to train on 3 languages at the same time. I give it some slack, you know? Yeah, maybe that's the problem. We also use the same account. Yeah, we use the same account and then like most emails right now are like partial German or And then just, yes. I'm assuming it's just not, uh, because it's not English. Because I'm assuming English is the best one.

Yeah, I'm sure English is the best because all the training data for, you know, the large majority of training data for all the AIs is in English because it's just been the dominant language in the web world, web, you know, for the past 50, 60 years, whatever. But yeah, keep, keep trying it. It's great. Everyone should be. I feel like you're an animal if you're not, if you're, if you're typing things still on a keyboard, you're, you're already behind.

An entire keyboard manufacturer is going to say, no, boo.

Yeah, no, legitimately, I have a really cool keyboard and I was really into keyboards for a little while, like mechanical keyboards, all the changes you had made, all that stuff. I kind of feel bad for those keyboard manufacturers. Your market's going to slowly disappear.

It still feels nice. Don't get me wrong. If you press the 2 buttons, you still get the nice feeling for the 2 buttons that you pressed. You do.

You want it to be good. I agree. But my point is, it's all happening in these things. I don't know how we're doing on time, but we should talk briefly about 2 other things I wanted to talk about with you guys. Please. A few things. So one, there's some stuff that related to some of our previous conversations, but I think just today, Bill Gates put out a manifesto about where he sees things going, which was once again somewhat aligned with a lot of things that I said 6 months ago. But Bill has a much higher— many more people listening to him than I do. So I can make predictions that could be right or wrong, whereas he has to be a lot more accurate. So, but, but some of the things that he talked about recently, you know, he's like, he's kind of an AI or he's a technology optimist, right? I think most people would say Bill Gates is very much an optimist, like more tech, more tech, let's go. He essentially said we need to slow down. Like, this is going too fast. We're going to see massive job loss. We're not ready. Let's regulate a bunch of things. I generally think he's spot on with most of that. It's very similar. I don't know if we talked about it previously, but like the AI 2027 report. They had a new update subsequently around, it was called Plan A, I believe, AI 2040 Plan A. Generally speaking, it's all like, hey, we need to slow down because we can't get this stuff wrong. And right now we arguably are getting it wrong. The Hugging Face hack and all that stuff is like, we're probably messing this up. This is scary. Thank goodness it wasn't worse, but what if it is worse next time? And it probably will be. He said we should slow it down. He's, you know, very much an optimist, but like, I'm not. I, I love his decision to try to talk to say we should slow down. I know he'll work hard to do it. I'm skeptical it will work, unfortunately.

2 things. Could you give a 30-second lowdown of the Hugging Face hack? And usually in the past, more regulation leads bigger players to have the ability to meet the regulation. Small players will have bigger problems. What are your thoughts on these 2 things?

So the Hugging Face attack was basically OpenAI had trained one of their new models to essentially like, it was like, hey, like, please try to go get the, I'm going to get some of these pieces wrong, but essentially like the weights to a test that Hugging Face had hosted, like figure out how to solve for these tests. But it was essentially an impossible task. The AI realized it was an impossible task. The only way to solve for it was to break out of its containment where it was not supposed to access the internet, hack Hugging Face to get the answers. So like, you know, if you give the AI an impossible task, it's going to do things that will, might, you might not want it to do, which arguably we might all do that, right? If you, if you sort of say to somebody like, you, you need to go save your family member from something, but by the way, you're in a jail that you can never leave, and this is the best jail in the world. I bet you would basically work night and day to find a way to escape.

Yeah, because they have the task, they have the obsession to do the task, right? It's like for us, like they're going to kill your family. So we're

like, you're going to have this sort of reward function of saving your family. It's kind of the same thing in this case. Obviously it wasn't that big a deal, but it still sort of is. It's like, okay, it needed to figure it out and it did all these crazy things to literally hack out of its containment, which was supposed to be foolproof and not be able to get to the internet, broke out of it. To go get this stuff done. That's crazy, right? And then, so it did this. It literally hacked into Hugging Face to get this. Hugging Face noticed it. They didn't know where it was from. They then had to use their own AI to fix this. They were not able to use the top-tier models because they were being blocked by the Trump administration to use those. They had to use Chinese models to root out the hacking from OpenAI, which OpenAI did not know was happening from their systems, which obviously, like, that's a huge failure of security. But it turns out not quite as bad, but Anthropic and other folks have had the same thing happen. There was also another incident, I believe, with the British, some main British security office noticed that something from— they were testing out Mythos and it had hacked into some GitHub repo, tried to insert a— basically hack into the repo. The guy who's a moderator noticed it. It was like, hey, this doesn't look right. I'm going to say no. And then he created another account saying, no, no, this is totally legit. And it was doing all these social engineering things to try to get it in, crazy things. And that's just one warning sign among many. And actually, if you read the AI 2027 report, it's like, yep, this will probably happen. Yep, did happen. This will happen. This will happen. And it's like 90— I have to check again. It's probably like 93% accurate, that report, which AI 2027 is basically like We're kind of screwed is sort of this conclusion. And there's somebody built, of course, an AI tracker of how accurate it is. And it's like, last I checked, it was like 93% accurate.

Damn.

And this was— they made it like 18 months ago, right? They finished it 18 months ago. They probably made their predictions like 24 months ago and they're like still kind of spot on. And then you have people like Bill Gates now saying like, hey, like we need to slow things down and stop and regulate more. And it's

like, how?

I don't know. I don't trust the government to be able to pull that off.

How would you say, because it's always easy to say, yes, we have to slow things down, but at the same time, the big companies will say, hey, I can't slow things down because in the end, otherwise China will overtake us or whatever. Because I feel like it's like the Spider-Man meme where the fingers pointing at each other and both saying, if you stop, I'll stop, but none of them will stop because they are afraid. They don't trust the other one.

Yeah. I mean, this is human nature and this is the real problem that we have. And this is why I'm more worried for our future than not is because This is, this is what it is. That's worrisome. Like, that's really, really scary that the, like, it's just human nature to do these things and you feel obligated. To be clear, I feel the same way at Shore, even on our small little teeny tiny scale, right? I feel the same way where it's like, I, I don't want to keep using AI and reducing the amount of work that my team has, which could lead to people having fewer jobs, right? Like, that's a potential outcome. At the same time, there's economic pressure for me not— if I don't do that, then we fail as a company because my competitors will do this. And there are competitors who are trying to do just AI only. To be clear, I don't think any of them have pulled this off, or at least in our industry, but like, they're trying. And so if we don't, we're screwed. So I have to do it, even if it's not the right thing for humanity or for my team. And Anthropic is saying the same thing, and OpenAI is saying the same thing, and then the Chinese labs are saying the same thing. And so yeah, everyone's rushing headlong into something they can't really control. And that's, that's scary because there's no guarantee these AI, which are essentially our children, are going to be our good children or our bad children. And like, the simple fact is, sadly, like, most people are really good. And right, if you go travel the world, most people are good people, irrespective of political stance or perspective. Like, day to day, in person, they're good people who want to live a good life and be good with other people. Yeah. I guarantee in every city there's also one or two crazy people who are like, oh, you give me a nuclear bomb, I will launch that right now and nuke this whole place because I'm unhappy with life in the world. So yeah, like, and we're giving a lot of people that power.

How would you fix this if it were up to you? That's like the easiest question you could have asked, Lukas. How do you fix this?

I mean, look, I'm definitely not—

You're welcome.

It's stuff that like Bill Gates is saying, always telling people like, you need to slow it down.

Yeah.

You basically need to find a way. And what the Plan A report said from the AI 2027 group and that stuff, it certainly seems like the right one. They're the experts though, I'm not. But the high-level basis is like, we don't stop it, but we slow things down. We also give full transparency. So basically we put all of our training work in China, basically, or Mongolia or something. And China puts all their stuff in Canada. So it's very easy for either of us to get to, maybe not the details of it, but you can destroy it sort of thing. You make it really transparent. You'd know where all the data centers are and you'd sort of say, hey, we're only going to go this fast and we're going to make sure it's clear to everybody how fast we're going. And you do that as a team worldwide. You can still compete, still build things, go fast, but just not quite as fast. The reason it's dangerous to go as fast as we're going Is because the ability, like the Hugging Face attack that we just talked about, it's easy for the world, like we're not ready for that level of power. If like a major AI company with essentially as much money as possibly anyone could have and the smartest people messes that up, it indicates to me we are not ready. For the power that we've unleashed. And so, like, we need to give ourselves enough time to catch up to it and make sure those AIs are fully aligned with us as much as you can be, that you don't have the crazy AIs out there running around. How do we do that? And, like, that's where we just need to slow down. But it's really hard. Human nature makes it really difficult, right? Anthropic needs to go public to keep growing, to get to recursive self-improvement, to get to AGI. And they would say they need to get there before anyone else because they're more trustworthy. Well, like, you know, I don't— why are they trustworthy? I don't know. Dario seems like a good person who cares, but like, that's a scary amount of pressure. Yeah, for like someone like just one random, random dude.

Like, even then it wouldn't even make it matter because it doesn't matter if he's trustworthy. If the AI has shown that it can go out of its containment, Doesn't matter what he is.

Well, right, but that's where you train. Hopefully you train it. You're going to have to train these AIs, and I know they're certainly trying to do this, where it's aligned with humanity and the outcomes that we want. And that's sort of the big part where you just say, like, hey, can we make this aligned as much as possible? So it's going to say, okay, like, you might do a hacking attempt, but like, you shouldn't really do that, or you say no, or all these sorts of things. And like, sure, you're going to have open weight models and people are going to do whatever the heck they want, but like, there's going to be enough good guy models that won't do those things that it sort of balances out. That's probably where some of this gets to. But yeah, it's a lot of responsibility to put into a single individual and that's where we are.

What is another prediction that Bill Gates said or in his report besides slowing down?

Government regulation, which I don't know, I don't really trust our government to regulate these things properly or the— I mean, I think Europe messes up all their regulations too, so it's not like Europe's any better. They like to regulate. They love to regulate. They generally make the wrong decisions on regulation.

AAA players. That's what they are. AAA regulations. Every company, like, I feel like the government is also a company that wants to grow. So in the end, you see, like, our roommate and we talked about the other day about Germany, and then we just looked up how many people work for the German government, and it basically grew 40% in the last 5 or 6 years. It was an insane amount of number of people who just started working for the government because in the end, Normally, you should think, especially in the age of AI, that you need less and less people, but it's the opposite, that the government will just keep growing and growing because there is nobody that says, hey, we need to cut this down. And it's dangerous because they have the voting power. Oh yeah, I'm not going to vote for someone who's cutting down government jobs because—

Everything's misaligned. And then you have unions where you can't fight the union who's saying, oh, we want to get paid more and guarantee our jobs. And if you don't, we're going to tell you you're anti-teacher or something. It's really difficult. So of course a politician says yes. And then you burden the government with debt they can't afford, and then they raise taxes on the people, and then people leave. And it's kind of a fairly simple system. I will say, I do think you can fix those things over time. This is sort of not about AI, of course. San Francisco, which is definitely notorious for its regulation as well here, is doing a lot to cut that. And to its credit, it has a long way to go still, but it's doing a lot to streamline things and say like, hey, like, do we need a dozen people to sign off on like building a new storefront? Like, that sounds kind of silly and unnecessary and it's not— it's not saving lives. I mean, like, I'll give you one example even for myself. Like, I live in San Francisco. I wanted to install solar power, which I would imagine would be a good thing for the city. It says like, hey, we have more power in the city that's generated in San Francisco. Instead of coming from far away, wherever our power plants are, that should be a win. However, my house has a power meter that's inside the house. So in the US, you have these power meters to see how much power you're using so someone can check it if they need to. And that's just in that being built inside the house, by the way, is a regulation they changed. So they kind of, I think, mandate them being built inside houses now because they can, I think they use like Wi-Fi to check them. So you can like the power meter, people can like walk by and use this Wi-Fi to connect or somehow and like pull the meter information. However, San Francisco Fire Department says, if you want solar panels, you have to have a power meter outside because we need to be able to stop the power in case there's a fire and we have to turn off the power from your house. So guess what? I didn't spend the $2,000 or $3,000 to move my power meter outside, and now I don't have solar panel power. I think that's a net loss for San Francisco, but You know, the fire department, though, would say we're protecting people from fire. So like, that's a legitimate concern too. How do you weigh these?

It's a very complicated world. And I feel like everyone has the right to say these things. Like the fire department, super important that they are protecting people. But on the other hand, you're like, guys, the regulation needs to make sense here.

Well, right. And my bet is, had the fire department thought more about this, there's probably a workaround since most other cities and counties have figured this out. There's probably a workaround to do this where you can still shut off power. Or you're talking about an edge case on an edge case on an edge case. And then we have to weigh the different harms. Yes, we might save a few people from a fire, legitimate concern, but what about all of San Francisco losing its power plants and we have our own power generated here? How many more people might die if our power goes out? Or how many more people die if we keep environmental catastrophes that are happening because we're burning coal far away? versus producing it for free on someone's roof. These are big things to weigh. The fire department doesn't care about that. They care about people having fires in the city. So that's their regulation. I get it. Someone else, the mayor, let's say, should be thinking about what's the overall harm to the city and the pros and cons. But that's hard. You need then kind of someone who's a little bit more of a dictator and has less power to the people. And that's also a challenge.

It's difficult when people have to make these decisions. Adam, there was still another topic that you wanted to talk about, right?

You said

that was the

So the other really interesting thing, and this is one that is fascinating to think about too. So Dylan Patel and Dwarkesh, Dylan Patel was just on Dwarkesh's podcast recently. Highly recommend listening to all his stuff if you get a chance. But they made a really interesting point, which is basically once again, everything's getting larger. All the spoils are going to the large folks, number one. Number two, it's going to get really difficult in the next few years for every small country that has bad debt or even companies that Have like some debt. The reason is because there's so much money getting poured into AI right now, and it's being financed with debt. And the companies who are doing AI, Anthropic, Meta, folks like that, are willing to pay so much for that cash that it's raising borrowing costs or will raise borrowing costs for everybody. Now that leads to major issues. If the world essentially becomes compute, And essentially all the money goes to compute. That means you can't get a mortgage for your house because it's not like you're not willing to pay 10% mortgage rate for your house, but Meta might be willing to pay for that because they know what the ROI is exactly on new compute they can put online. Or if you're a country in a place that maybe doesn't have the best debt rating, you, your borrowing costs go up, which means you can afford less for your people, which leads to political instability. And so you start getting into this really weird world where once again, all the money goes to compute and you can't afford anything for the people at all. And how do you stop that? Once again, regulation, better taxation, maybe. It's a very—

My mind is just blown. It's just like, never thought about this. Never thought about that, but it makes sense. Kind of finite amount of money. That's why you also see that, I believe at least, why Bitcoin is not recovering right now because AI is just absorbing so much money of like so much capital overall that Bitcoin does. There's not enough interest in crypto right now because AI is covering right now. It's so

it has come back a lot, but it's not anywhere where it was.

Yeah, you know what I mean? It's not exactly where it's supposed to be or whatever predictions were because AI is absorbing so much.

Yeah, and I think the borrowing costs are the bigger part though, right? I mean, like Bitcoin's a different interesting hedge, which is sort of an interesting conversation too, but. But I think it's more so about largely, once again, the borrowing costs that haven't quite impacted people. But once again, there is finite cash. And so if it's all just going to finance AI buildout and compute, what's left for anyone else?

Yeah. If it all goes there, because there will be at some point limits too, right? Because there's Well, I guess there's not limits if AI robots are also building themselves. Interesting to see. I love this. This is the perfect one for, I'm like, all right, I want to see the next episode in 6 months, whether we've seen actually already a trend towards this, right? Because we're seeing it right now. And then in 6 months, it's like, okay, is this the case? We're seeing already tendencies towards it. So I think that's going to be an interesting prediction, I hope.

Look forward to it. That would be very interesting. Let's see if it's true. It's hard to imagine that it's not. Once again, so much money is going right now to the buildout. And if Anthropic's able to hit a $2 trillion valuation, that's insane. But you could also argue that's also small. If they're able to capture— if you go look back at the original OpenAI thesis, when they first did their deals with Microsoft to get investment from Microsoft, they're basically like, if we're correct, we're going to be the last company on planet Earth. Because we will absorb all human labor and all the work that ever existed. And if that's true, as an investor in those companies, you're really happy. But it's also the end of every other company and jobs for every— OpenAI maybe has 2,000 or 5,000 employees and they don't need a lot more than that. Whereas Walmart has 200,000 employees.

Makes sense. All right, Adam, let's leave it with a positive note. What is a positive prediction you have for the next 6 months or 12 months before we go

Yeah, look, the positive side of this, right? There's a lot of scary things that are happening right now. On the other hand, I think the really positive side is like, we are probably going to see cancer be cured. Maybe not in 6 months, but we're going to see the cure for cancer. We're already seeing some of that stuff happen, actually. Cure for cancer. We'll get to worlds where there's just unlimited energy, new types of physics that we just never— things we've never barely dreamed about. We're going to start to see. There's sort of been this big buildup where everyone's like, AI's coming, AI's coming. I think in the next 6 to 12 months, we're going to start to see it really impact things in much more significant ways. Once again, even basic ones like self-driving cars, right? They're already out around San Francisco all the time, but Tesla's turning that on around the world on all their cars. And so you're going to see this happen more and more and more and more. And that unlocks a lot of really good stuff for a lot of people, right? To just give one example, once again, with self-driving cars and even cancer stuff, it's like, I don't know how old your parents are, but my parents, as they're getting older, They still want their freedom, but at some point you take driving abilities, you take your driving license away from parents when they get older. My parents are not at that point, thankfully, but that could happen for somebody and then they lose their sense of freedom and then they decline pretty rapidly and that's it. Self-driving cars, they don't have to do that. They can still drive themselves anywhere. Medical advances are getting so good that they will be able to live a lot longer. There's a lot of really amazing things that are happening. And so as long as we don't kill everybody, kill ourselves with AI, we're gonna be living in every sci-fi. We're gonna speedrun basically every sci-fi trope that's out there, which there's some really scary ones too, but like we're gonna speedrun some of the really cool ones too. And it'll be heading to the stars in, you know, 20 years, all of us, if you want.

That's a great finish.

That's a perfect ending.

Good.

I love it.

Thank you for coming on the pod, Adam.

It has been a pleasure again, guys. Keep me posted on the next Cyber. Let's do this in 6 months, see how well I did with predictions. And make it to San Francisco.