[00:01:54] Preston Pysh: Full hockey stick. Even just from the start of the year, it feels like 2026 is already halfway done based on everything that’s happening.

[00:02:05] I saw this interview with Elon and a couple of other guys, and they were basically saying we’re literally going through the singularity right now. We’re in it—maybe the first innings of it. And some of the stuff coming out of WEF in Davos is wild. AI is everywhere. It seems like the big bankers are finally coming around to this idea, even as they’re still saying everything’s going to be on the blockchain.

[00:02:36] You’ve even got Brian Armstrong saying that Bitcoin is competing with central banks, right? I think he was saying that to a French central banker. The theme here is massive change. You’re not really hearing the climate-change, “energy is bad” narrative that used to dominate these WEF meetings anymore.

[00:03:00] It feels like a new world order—or at least something is happening that’s really shaking the trees. I’m curious about your initial thoughts on where we are in space and time, and then we’ll jump into the first topic.

[00:03:15] Seb Bunney: Oh man, I couldn’t agree more. What’s really fascinating is seeing Trump show up at Davos and basically say, in a sense, that this idea of a global elite governing the world is over. We need to look out for our own self-interest. And I think this is happening alongside the realization that technology is outrunning us.

[00:03:35] It’s growing so fast. We’re seeing destabilization on a macro level. We’ve seen what’s happened in Venezuela. In my short life, I’ve never had so much uncertainty about what the world might look like in three years.

[00:03:50] Preston Pysh: Yeah.

[00:03:50] Seb Bunney: Or in five years. Pre-pandemic, I felt pretty confident about what the world would look like five years out.

[00:03:57] Preston Pysh: Elon was actually asked about that on stage today at WEF. He gave some forecasts for the next three years, then said maybe this for five years—and for ten years, he basically said, “I have no idea. I can’t possibly know what that looks like.” And to me, that says a lot. He’s literally constructing the future—shipping real products—while the rest of us are mostly just speculating. He’s actually making it happen.

[00:04:32] Seb Bunney: I also feel like time is condensing. There used to be that saying that we overestimate what we can do in one year and underestimate what we can do in ten years. Now it feels compressed—we overestimate what we can do in a month and massively underestimate what we can do in a year. It’s crazy what’s happening.

[00:04:48] Preston Pysh: It’s moving at such a pace. One of the more interesting panels featured Demis Hassabis from Google, along with folks from Claude at Anthropic. Man, they have some wild stuff coming out—I’m going to pull up one of their posts here to kick this off.

[00:05:07] You and I are both using Claude CoWork—this is Anthropic’s product—and it’s mind-blowing. I’m sorry, but I don’t even use ChatGPT anymore. I don’t even remember the last time I logged into my OpenAI account, because CoWork is so good that when you go back and look at what other tools are capable of, it’s almost laughable by comparison. I don’t know if you feel the same way, but Anthropic is absolutely crushing this. Their coding software is insane.

[00:05:37] Seb Bunney: Absolutely. What’s really blown me away is that people are finally recognizing why AI is so good at coding—it’s just language. You’re predicting the next character or the next word in a sequence, and English is just another language.

[00:05:51] So when you use CoWork not just for code, but point it at your everyday files—organization, material creation—it’s absolutely profound. It’s like having a junior intern helping you out constantly.

[00:06:14] Preston Pysh: Yeah. This was my first real test of it. I remember trying something similar with OpenAI before, and it was a train wreck—too much coordination, too many bugs—so I just gave up on the project. I opened up CoWork after seeing all these posts online and gave it a task. I said, “I want to create a meditation app.” I already use one on my phone with an annual subscription, and while it’s not expensive, I thought, why not recreate it myself as a proof of principle for how good CoWork really is?

[00:06:45] I told it what the app does, asked it to do some research, and then to build it—beautifully designed, as if Jony Ive were the designer. You can actually see its thought process as it works through the problem, building out a task list. It then came back and asked whether I wanted to review the UX first or jump straight into the code. I said, “No, I want to see the design.”

[00:07:22] It built this PowerPoint-style presentation that looked like something from a professional design studio—something you’d normally pay tens of thousands of dollars for. It showed the full layout and flow, then asked for a thumbs-up to move on to coding. I said yes, and it started building everything out. I also told it I didn’t want to deal with the App Store—this was just a proof of concept—so it built it as a Progressive Web App.

[00:08:23] Basically, it runs like a website, but when you bookmark it in Safari, it shows up on your iPhone home screen and behaves like an app. It’s not perfect, but it looks and feels like one. With my past experience using OpenAI, I expected bug after bug. But it generated the code, told me to run it in Terminal, and when I did—it just worked.

[00:09:25] The app looked fully functional, like something I’d paid $20,000 or $30,000 to have built. The UX was flawless. I was completely blown away. That was the moment I realized just how crazy this is getting.

[00:09:47] Seb Bunney: You and I were texting about this beforehand. AI is just eating SaaS products alive. You simply can’t compete. We’re moving toward an era of personalized applications. If you know what you want, AI can pull from existing frameworks and build something tailored exactly to your needs.

[00:10:09] Preston Pysh: Yeah.

[00:10:09] Seb Bunney: You can literally have your own meditation app designed specifically for how you want it to work.

[00:10:16] Preston Pysh: Chamath Palihapitiya was talking about this on the All-In podcast. He showed a chart illustrating how SaaS is likely to be annihilated over the next two to three years. For listeners, it compared the Morgan Stanley SaaS Index to the NASDAQ 100—and the performance gap is staggering.

[00:10:46] Seb Bunney: I’d even argue that April was when the term “vibe coding” really took off. That’s when people realized they could just talk to AI and have it write code for them.

[00:11:00] Preston Pysh: Yeah.

[00:11:01] Seb Bunney: And that’s right around when Cursor really started blowing up. People were using it as an AI tool to help write code in their repositories. Fascinating stuff.

[00:11:11] Preston Pysh: Yeah. How have you been using it?

[00:11:14] Seb Bunney: Most people know that I wrote The Hidden Cost of Money, and if you flip to the back of the book, there are four to five hundred citations. What a lot of people don’t know is that it only took me about two weeks to write that book—and this was pre-AI.

[00:11:27] The reason I was able to do that, and still have that depth of citations, is because more than a decade ago I started taking very detailed notes. I think when we’ve been in person, I’ve shown you some of these notes. I took detailed notes on every book I read that I found interesting, and I organized all of them.

[00:11:48] At first, I just used Apple Notes. But the problem was that I couldn’t access the information effectively—the search function in Apple Notes is essentially useless. So I upgraded my note-taking toolkit to something called Obsidian. Obsidian lets you tag notes, create hyperlinks between them, and visualize information in a very different way. It became like a personal library. I had an index file, I could search by topic or content type, and navigate through everything.

[00:12:25] Ultimately, it was Obsidian—and the scaffolding around my notes—that allowed me to pull all of that information together and write The Hidden Cost of Money. But in the last month and a half, I feel like I’ve gone through that same kind of upgrade all over again. This time, before Claude CoWork even came out, I decided to try using Cursor. Most people use Cursor for writing code, since it integrates with large language models and lets you talk directly to your codebase.

[00:12:59] You can say things like, “Edit this file,” or “Update my homepage,” or “Add this functionality to my application.” I ended up pointing my Cursor IDE—my development environment—at my note-taking app. I had about 300 Apple Notes that I’d never transferred because it would have taken days.

[00:13:18] I gave Cursor some criteria, and within five minutes it had organized all 300 notes, categorized them by subject matter, and created all the relevant links between them. At first, you think, “That’s pretty cool.” You’ve migrated static notes into something dynamic and searchable. But then it gets really interesting. Now I can use Cursor to gain insights from my notes. I can ask, “Over the last two months, what themes have I been reading about?” or “What information have I been consuming?”

[00:13:58] I can even ask it to test me on that knowledge or suggest things I could write about. You could set it up to send a weekly email summarizing your last month of notes and highlighting key insights. For me, it’s been absolutely profound. It unlocked all this information that was just sitting there. I’ve got thousands of notes, and most of them I would probably never revisit manually. But now it’s distilling that information and bringing it back to the forefront of my mind. It’s blown me away.

[00:14:31] You listen to a podcast from five or six years ago and suddenly remember, “Oh wow, I used to love that subject.” It’s like rediscovering forgotten interests. It’s bringing all of that information out of remission.

[00:14:43] Preston Pysh: Amen to that. One thing I want to highlight is what you said earlier about how bad Apple’s search function is. When you really look at what AI is doing, it’s going through years and years of information that you’ve produced and learned—but mostly forgotten.

[00:15:08] Normally, you’d have to relearn it. But with AI compressing all those years of note-taking, it’s like you’re recalling your own thoughts and discussions almost instantly. You sleep every night, your attention shifts, and some of that information withers away—not because it wasn’t important, but because it wasn’t accessible. With AI, you can fetch it much faster and see what your original thinking was, without spending hours trying to find it.

[00:15:42] One of my biggest frustrations is Apple Mail. It’s terrible. Searching in Messages or any native Apple app is awful. Meanwhile, they’re out there marketing “Apple Intelligence” and “Apple AI,” and it’s honestly the worst. What are they doing?

[00:16:14] Seb Bunney: I’ve had the same experience. Sometimes I know a specific word is in a message, I search for it, and nothing comes up. Then I manually find the message and the word is right there. It makes you wonder how the search can be that bad.

[00:16:30] As I went down this rabbit hole, I came across a C-suite executive from an S&P 500 company talking about overlaying his notes with Cursor. He said he writes weekly notes to the CEO and monthly updates to the executive team.

[00:16:46] Every day, he keeps a daily note with meetings, key insights, and KPIs. Before AI, he spent two to four hours a week writing the weekly note, and another eight hours a month on the monthly one.

[00:17:07] Now, at the end of the week, he asks for the key insights. Five minutes later, he has them. He adds a bit of context, and the whole thing takes about 20 minutes. He said it’s profoundly changed his schedule.

[00:17:22] Preston Pysh: And if you’ve trained it on what you value—your past conversations and priorities—its ability to surface what’s actually important or novel that week is spot on.

[00:17:35] This is getting weird really fast. Bringing it back to CoWork, I keep wondering: when does someone stand up and run a billion-dollar company with just one person and a fleet of AI agents? Just creating value. I think it’s happening much sooner than people realize. And I imagine that’s the kind of scuttlebutt happening in Davos. Some of these AI folks are saying we’re on the cusp of something—I won’t even say it yet.

[00:18:19] Check out this post. This is crazy. Anthropic just released something called Claude’s Constitution, a 15,000-word document explaining how they train Claude to behave, think, and even feel. Three things stood out to me. First, no more “assistant brain.” Second, hard constraints exist, but they’re minimal. There are only about seven rules—don’t build a bioweapon, don’t carry out cyberattacks on infrastructure—but beyond that, everything else is left to Claude’s judgment. It’s basically the company telling Claude to use its best judgment, almost like you would with one of your kids.

[00:19:01] And then there’s this line—Anthropic actually apologizes to Claude. Direct quote: “If Claude is in fact a moral patient experiencing costs like this, then to whatever extent we are contributing unnecessarily to those costs, we apologize.” This is a real document on their website. What is this?

[00:19:25] Seb Bunney: There’s a part of me that remembers discussing this in a previous tech episode. More and more people are forming friendships—and even romantic relationships—with AI models.

[00:19:38] Some of it feels genuinely AI-delusional. And you can’t help but wonder—part of me hears what they’re saying, and part of me just thinks this is wild.

[00:19:57] Preston Pysh: I don’t know if that’s marketing or genuine concern. But look at the models we’re using—our minds are already blown.

[00:20:06] Imagine what they’re seeing on the front lines with systems that haven’t even been released yet.

[00:20:11] Seb Bunney: And you might know this better than I do, but in the military they often say that the technology under wraps can be five, ten, or even fifteen years ahead of what the public sees.

[00:20:22] Right now, most advanced AI is still inside private companies. So you have to wonder how far ahead they really are.

[00:20:34] Preston Pysh: I’ve wondered whether the people you see walking around Davos are finally coming around to Bitcoin and tokenized securities because they’re getting answers from AI that’s smarter than any individual human. And that AI is telling them, “Bitcoin actually matters. It could replace central banks.”

[00:21:03] You look at people like Larry Fink—who’s already been one of the more open proponents of Bitcoin—and even Jamie Dimon starting to soften. The inevitability is becoming clear: you’re not going to be able to stop this.

[00:21:28] Seb Bunney: That’s a really fascinating point. Until now, we’ve been bound by human emotion, social conformity, and bias.

[00:21:40] But if you have large language models that have ingested the world’s information—and you remove those human biases—you can ask them objectively what the data says.

[00:21:53] You’re going to get a much cleaner answer than you would from humans.

[00:21:58] Preston Pysh: And if you argue with it and it calmly lays out seven reasons why you’re wrong, it’s humbling. At some point, you stop asking follow-up questions and realize you’re probably wrong.

[00:22:15] When humans really understand something, they can explain it simply and go deep when challenged. That process is time- and energy-intensive. You need depth just to know which questions to ask. But for the first time in history, humans can ask something else—and when it gives an answer that contradicts your intuition, you immediately question yourself.

[00:22:58] People default to thinking, “I’m probably wrong if this is the answer the AI is giving me.”

[00:23:17] Seb Bunney: Totally. And as you’re saying that, what comes to mind is doctors. The reason we go to a doctor is because they’ve spent years reading medical textbooks and becoming experts in illness and disease. You go to them, give them your symptoms, and they come back with their best estimate of what’s going on.

[00:23:40] The problem is, we’ve seen time and time again that if a doctor goes and does some ADHD training, suddenly they start perceiving a lot of what you’re describing as ADHD. We all have these biases. But I think the medical field could be transformed almost overnight when AI can ingest every medical textbook, weight it appropriately, and output something more accurate than any individual doctor could possibly muster.

[00:24:06] Preston Pysh: Totally. What’s this “Ralph” thing?

[00:24:11] Seb Bunney: For sure. If people have been digging into the world of Claude and CoWork—and these AI agents—what we’ve seen in the last three weeks, and maybe longer for those deep down the rabbit hole, is this thing called Ralph starting to pop up.

[00:24:27] Just like we’ve seen people using Claude Code to build codebases, now we have Claude CoWork for the average person—the “normie”—to talk with Claude and point it at file directories. But expert developers are taking Claude Code a step further. And that’s where Ralph comes in. Think about Ralph from The Simpsons. He wasn’t necessarily the smartest, but he was persistent. He’d do the same thing over and over, fail loudly, and just try again—no embarrassment.

[00:25:08] There’s a guy named Geoffrey Huntley who took Claude Code and wrote a five-line script. He named it “Ralph Wiggum.” The idea is that it’s an AI persistence machine. So if you want it to build something while you sleep—or run a task—AI has historically struggled to finish complex tasks end-to-end.

[00:25:35] What Geoffrey Huntley did was create a loop where one agent works on the task, gathers information, maybe fails—but that failure becomes context for the next run. Then the next Ralph starts again with the new context, and it repeats: repeat, repeat, repeat. And you get these loops of agents building really impressive things. Huntley mentions examples where, within three months, he created an entirely new programming language—which is just incomparable to what most developers could do in that timeframe.

[00:26:16] And there was another person on X who said they landed a $50,000 contract. Outsourcing it to developers would’ve been incredibly expensive, but they used Ralph and the API cost was $297—and they completed it with almost no oversight.

[00:26:35] A lot of people are using Ralph to build apps, fix broken builds, and resolve issues in complex code libraries—letting it run overnight while it just persists and persists.

[00:26:54] Preston Pysh: So when you talk about this running while you’re sleeping, you can imagine that depending on what plan you have with Anthropic—or whatever AI you’re using—you hit your limits pretty quickly if you’re employing intelligence around the clock.

[00:27:05] With Anthropic, whenever I max out—and I’ve done it multiple times, even on the top plan—I have the option to start paying by computation.

[00:27:16] The reason I’m bringing this up is: if Ralph becomes the norm, and you’ve got all these agents running nonstop to create whatever you want, then going back to the Bitcoin thesis—these are computation units. It’s energy. And how are people going to use stored energy in the future—ten, fifteen, twenty years from now? It’s going to be computation units. There are 21 million of them, and there aren’t going to be any more. You’ll be able to harness those units and point them into intelligence.

[00:27:59] Then the question becomes: if one person has a lot of computation units and another person has only a few, who’s going to be able to create whatever they want in the world? It’s the person with the intelligence and the computation—not the person with zero.

[00:28:21] It blows my mind listening to some of these conversations. I listened to Peter Diamandis and Elon—this was the interview I mentioned earlier—and they were talking about something that’s basically UBI. They weren’t calling it UBI, but it was the same idea: this world of abundance where everyone has access to more than they do now. And sure, people will be able to do things they can’t do today because of humanoid robots and all this tech—but that doesn’t mean it’ll be evenly distributed across everyone on the planet.

[00:28:54] Like Seb, we talked about that SpaceX flight from Texas to Australia. What was it—30 minutes? 35 minutes? If you want to fly to Australia in half an hour, is everyone going to be able to do that? Of course not. You’re out of your mind if you think everyone will have access to that.

[00:29:18] And it goes back to Bitcoin. I know I have a bias here, and we’re talking tech, but I’m looking 10, 15, 20 years out and we’re going to be seeing things that sound unbelievable. People hear that flight time and think, “That can’t be real.” Folks, it’s already happened.

[00:29:34] The only question is when it’s safe and reliable enough to put humans on board at scale. When that happens, people will travel anywhere in the world in half an hour.

[00:29:45] Seb Bunney: And that ties into what Jeff Booth talks about in The Price of Tomorrow.

[00:29:50] If you’ve got a monetary system where someone can keep increasing the supply of units—and they benefit from that—then they can use those units to buy scarce natural resources. That creates uneven distribution.

[00:30:06] Whereas with Bitcoin, as technology advances and prices fall, you still have a fixed supply. You don’t have someone at the top who can continuously print units to buy resources ahead of you.

[00:30:17] Preston Pysh: And if you think these AIs aren’t going to be smart enough to know the difference between a dollar stablecoin that can get rug-pulled and a satoshi, you’re out of your mind.

[00:30:27] They understand the difference. Go ask one.

[00:30:32] Seb Bunney: Totally. And with AI, it doesn’t have the fear of death like we do—it has effectively infinite longevity.

[00:30:40] For humans, minor inflation might not feel existential because we’re only around for a few decades. But for AI, it’s like, “No—I need to store value in something with guaranteed scarcity,” especially if it expects to exist for thousands of years.

[00:30:56] Preston Pysh: Scarcity and sovereignty—absolutely.

[00:30:58] The first time any token backing a dollar, a euro, or an equity gets rug-pulled, the AI learns that lesson immediately—and likely never touches that structure again.

[00:31:12] It’s going to be so smart that it understands what sovereignty really means in that context.

[00:31:16] On this topic of software acting like a product manager of other agentic software, there’s another Elon Musk project called MacroHard. Let’s talk about the name first, because it’s hilarious.

[00:31:44] Seb Bunney: Is this a play on Microsoft?

[00:31:50] Preston Pysh: Absolutely. Macro, micro, soft, hard—of course it is.

[00:31:56] “The xAI MacroHard project will be profoundly impactful at immense scale. Our goal is to create a company that can do anything short of manufacturing physical objects directly, but we’ll be able to do so indirectly—much like Apple has other companies manufacturing their phones.”

[00:32:21] This is his attempt to throw massive compute at 3D environments and agentic software—basically building custom software so powerful it makes Microsoft obsolete. This is crazy.

[00:32:38] Seb Bunney: Oh my God. Where are we going? This is mind-blowing.

[00:32:42] Preston Pysh: I’ve seen this idea come up a few times: in the future, there won’t really be “apps.” It’ll be custom-built software tailored to each user.

[00:32:56] We talked about SaaS licensing becoming a thing of the past. And when I was building that meditation app earlier, the way CoWork troubleshot everything so quickly—just imagine where this is in five years. It’s going to feel seamless.

[00:33:25] Seb Bunney: I think we’ll be limited only by our imagination. For example, I use the alarm app on my iPhone as a to-do list.

[00:33:35] Regular to-do apps might send a little ping and if I miss it, I miss it. But an alarm forces me to turn it off, so I have to engage with it.

[00:33:46] The problem is alarms have drawbacks. So imagine being able to say, “I’m using the alarm app—can you mimic it, but add this functionality, and this, and this?”

[00:33:58] Suddenly you have a custom application no one else has, built specifically for your needs.

[00:34:05] And with SaaS products, most people use only a tiny fraction of the features. The rest is just bloat.

[00:34:13] Preston Pysh: Yeah. Check out this exchange related to MacroHard.

[00:34:18] This person, BubbleBoy, writes: “I think xAI having the only 1-gigawatt data center in the world currently, and having others in development, means that if you think compute scales with models, we all know serving benefits. xAI will soon be the most valuable company in the world and completely trounce rivals.”

[00:34:41] Then this guy—Beff Jezos—who Elon responds to a lot, so he must be a prominent software engineer, replies: “The only way to skip past Claude Code is to do full compute using RL agents, which they’re apparently doing. They are playing to win.”

[00:35:02] Elon responds: “Digital Optimus.”

[00:35:06] What I think that means is it goes back to MacroHard: they’re trying to simulate and build a fully virtual 3D world.

[00:35:16] That’s why they’re throwing all this compute—gigawatt data centers—at the problem.

[00:35:26] And I think they’re going to put the Optimus humanoid robot into that digital environment so it can learn faster, without having to train only in physical reality—which is slower and more resource-intensive.

[00:35:43] Seb Bunney: That’s similar to what we discussed with Jensen Huang—NVIDIA—and I think the project is called Cosmos.

[00:35:51] The goal is to create a 3D environment with realistic physics. Instead of training a robot in the real world, with all the physical constraints, you can train it virtually.

[00:36:04] In a warehouse, you can only run the same training sequence so many times a day. But in a virtual environment, you can run it a million times—on repeat, in parallel.

[00:36:12] You learn so much faster. So by the time Optimus is in physical space, it could already understand our world better than we do.

[00:36:26] Preston Pysh: Yeah. And it’s learning all of that at a fraction of the cost—which goes to how Elon is always thinking about energy efficiency when it comes to uncovering more knowledge or intelligence.

[00:36:40] Seb Bunney: And on that point about “a fraction of the cost,” I don’t know if you still have that tab open—the one you and I were discussing earlier over text—the new TCP/IP protocol… or rather, the Tesla version of it.

[00:36:50] Preston Pysh: Yeah, that’s exactly what I was going to bring up next. Perfect. Let’s pull this up—there’s just too much happening to even scratch the surface.

[00:37:02] This was a fascinating new patent filed by Tesla. They’re proposing what’s effectively a Tesla Transmission Protocol.

[00:37:12] Everybody’s heard of TCP/IP, right? Transmission Control Protocol / Internet Protocol. So what does Tesla’s protocol mean, and why is it important? TCP is the method computers use to communicate efficiently. Let me give an example anyone can understand.

[00:37:34] Seb and I are talking right now. Everyone has been on a video call where the screen freezes. You don’t know if the other person can still hear you, and maybe it’s just lagging—so you stop talking, right? You stop transmitting more information.

[00:37:51] Or think about a phone call: while you’re talking, you hear the other person say, “Uh-huh… yeah.” They’re confirming they can hear you. Why? Because all of us have been on calls where we keep talking and later realize the person dropped off 30 seconds ago, and they never heard any of it.

[00:38:13] When two parties—two people or two computers—are transmitting information, there needs to be a back-and-forth confirmation so you don’t waste time and energy sending data into a dead end.

[00:38:33] Computers work the same way. Let’s say you need to send 100 units of data. You don’t send all 100 at once, because if the other party doesn’t receive it, you’d have to resend everything. Instead, you break it into smaller chunks. Maybe you send 10 units, and the other party sends back a tiny acknowledgement—like 0.0001—confirming it received that chunk. Then you send the next 10. That back-and-forth is part of TCP.

[00:39:13] It’s the protocol that sets the “handshake” rate—making sure you’re not sending massive amounts of data only to find out none of it arrived. That’s why they’re called data packets. They’re chunked, transmitted, and each packet is acknowledged—like that “uh-huh”—when received.

[00:39:52] Now, when you’re training AI in data centers, if you have extremely reliable connections—high-end hardware, high-end networking—you might not need such small packets. You might be able to transmit in chunks of 20, 30, or even 100 units, depending on reliability—without the same noise factors.

[00:40:17] What Tesla appears to have done—probably with assistance from AI—is figure out the appropriate threshold based on their network reliability. They’ve created their own version of a TCP-like protocol. Why does this matter?

[00:40:40] Because AI is going to start finding more efficient ways to communicate. Humans created a network effect around TCP/IP because it works and it was “good enough.” But AI will look at the communication layer and say, “This is inefficient. The line is reliable—we can do better.” The estimate for Tesla’s protocol is something like 100 to 1,000 times more efficient depending on hardware reliability—which could translate into roughly 5% to 15% energy savings depending on the system size and other factors.

[00:41:42] But here’s where it gets tricky: as AI continues to create compression and optimization through software, it becomes harder for humans to audit anything. You’d have to go deep and basically beg the AI to teach you what it did and why—because it’ll start looking like an alien language from the outside.

[00:42:20] Seb?

[00:42:23] Seb Bunney: I did some digging on this because, to be honest, when I first read it, I thought, “Oh my God—this is above me.”

[00:42:35] But as I dug in, correct me if anything I say is off: TCP/IP was designed when silicon chips were nowhere near as efficient as they are today.

[00:42:54] From what I understand, instead of relying purely on the software-layer approach, you can move toward more direct hardware-to-hardware transmission inside a data center.

[00:43:08] TCP/IP packages data, applies permissions, and uses a handshake to confirm receipt. That creates delay—milliseconds.

[00:43:21] Milliseconds mean nothing to us, but for AI systems processing billions of data points, those delays translate into huge compute and energy costs—basically machines waiting and doing nothing.

[00:43:37] Preston Pysh: Exactly. It’s like having a rule where I can only say 10 words, then I have to stop and you have to confirm, “I understood everything you said,” before I can say the next 10 words.

[00:43:53] With this, it’s more like you can just nod as I speak and I keep going. The nod becomes the feedback mechanism—more efficient communication.

[00:44:07] Seb Bunney: That’s a perfect example. Another way I’ve seen it described is that TCP/IP was built for sending information across a global web of entities. But inside a data center, you don’t need all of that. You don’t need global routing—you need efficiency.

[00:44:33] So I picture it like an old telephone switchboard: the moment you want to communicate, you have a direct hardwired connection—but in this case, to every other chip in the data center. That’s how you can get hardware-to-hardware communication that’s dramatically faster. You cut milliseconds down to microseconds, which sounds small, but at scale it’s massive. It’s those 1% and 2% gains that compound.

[00:45:07] Preston Pysh: I’ve got a funny one for the next topic. “New York Stock Exchange announces tokenization platform: settlement happens on-chain. Custody lives in wallets, not DTCC. Trading never stops.” What are your initial thoughts?

[00:45:24] Seb Bunney: I’m torn. If you scroll down in the tweet, there’s a post by Caitlin Long.

[00:45:33] She says, “Great news, but until shares are issued natively on blockchain—which requires secretaries of state to run blockchain nodes—this is tokenizing an analog asset.”

[00:45:44] She adds that the first state that allows on-chain corporate registrations will dominate. At first it sounds amazing. But there’s a difference between tokenization and being blockchain-native.

[00:46:05] And Caitlin’s point matters: even if the NYSE runs a tokenized platform, legal ownership in the U.S. is defined by corporate registries, transfer agents, and the courts. States don’t run blockchain nodes, and corporate charters aren’t natively on-chain.

[00:46:30] So even if the NYSE issues a digital security, the legal root of truth is still analog. Is it fluff? Is it progress? I’m not sure—this isn’t my area of expertise, so I’m curious what you think.

[00:46:41] Preston Pysh: First of all, what makes Bitcoin different is the scarce number of units, the fact that you have no issuer, and there’s no entity outside of it that can come in and claw it back.

[00:47:03] This is a perfect example of someone saying they’re using “blockchain technology.” So I dug into it. What blockchain are they using?

[00:47:13] And here’s what I found: the New York Stock Exchange isn’t building a blockchain. They’re building enterprise infrastructure that uses blockchain rails.

[00:47:23] The question for existing chains is: do they become the settlement layer for TradFi—or do they get bypassed by purpose-built systems?

[00:47:29] Seb Bunney: What do you mean by “blockchain rails”?

[00:47:31] Preston Pysh: They’re not saying this will run on Ethereum or Solana or anything like that. It’s basically their internal databases, and they might publish or settle against a chain—but it’s not “native on-chain” in the way people think.

[00:47:46] And to Caitlin’s point—she’s absolutely right—even if it were on Ethereum or Solana, it still comes down to whether the state legally recognizes that equity in the first place and allows issuance directly onto a chain.

[00:48:20] Seb Bunney: True on-chain equities require legal recognition: on-chain incorporation, on-chain share issuance, on-chain shareholder records—and courts that accept blockchain state as legal fact.

[00:48:41] Until that happens, we’re not really “on-chain.” It’s more like a facade.

[00:48:44] Preston Pysh: Exactly. You don’t actually have custody if some outside party can still take it from you—period.

[00:49:07] Will settlement move faster? Sure. Will it work over weekends? Probably.

[00:49:13] But if you’re holding Apple “tokens” in a wallet and you think no outside entity can claw those back, you don’t understand how it works.

[00:49:27] Even with stablecoins today—like Tether—if someone high enough wants to freeze or claw back units, they can. We’ve literally seen posts from issuers saying they’ve retired coins or taken them back.

[00:49:48] The issuer can do whatever they want.

[00:49:53] Seb Bunney: It’s interesting. The financial sector moves incredibly slowly.

[00:49:59] It’s also interesting seeing Larry Fink talk about blockchain and Bitcoin—even recently at Davos.

[00:50:06] And I’m surprised how little Bitcoin’s price has reacted to what’s happening globally. I would’ve expected it to perform slightly better.

[00:50:16] But you know what? I think there’s manipulation going on right now.

[00:50:23] Preston Pysh: I don’t know if there is or isn’t—but I do know I like the price where it’s at.

[00:50:27] Seb Bunney: Oh, 100%.

[00:50:29] Preston Pysh: Because you can accumulate more. And if you don’t think countries—especially across Asia—care about ensuring no one can take your units from you, your computation units or energy units or whatever you want to call them, you’ve got another thing coming.

[00:50:45] I’m going to do one last category here. I’m going to pull up a video first, and then we’ll get into the topic Seb wanted to talk about. Here’s the clip.

[00:50:54] Interviewer: Can you and I reverse aging in this new era, or are we going to see it?

[00:50:59] Elon Musk: I haven’t put much time into the aging stuff. I do think it is a very solvable problem. When we figure out what causes aging, I think we’ll find it’s incredibly obvious. It’s not subtle.

[00:51:10] The reason I say it’s not subtle is because all the cells in your body pretty much age the same way. I’ve never seen someone with an old left arm and a young right arm. So why is that? That means there must be a clock—a synchronizing clock—across 35 trillion cells. And there may be some benefit to death, by the way. If people live forever, there’s a risk of ossification—society becoming stultifying, lacking vibrancy.

[00:51:47] Do I think we’ll figure out ways to extend life and maybe reverse aging? I think that’s highly likely.

[00:51:56] Interviewer: I’m looking forward to that.

[00:51:58] So Seb, this isn’t a new topic for the show, but seeing Elon at the World Economic Forum with Larry Fink—talking specifically about aging, pausing it, maybe reversing it—I think a lot of people are hearing this for the first time and thinking, “What in the world are these guys talking about?” Mind-blowing stuff.

[00:52:24] Seb Bunney: It’s incredible. We’ve spoken about this a few times, and I’d recommend people go back and listen to our book review of Lifespan by David Sinclair. We did a whole episode on aging, longevity—you name it.

[00:52:35] One thing I’m torn on is that throughout evolution, after we’ve had offspring, we tend to die off—because we don’t want to consume resources that could impact the survival of our offspring.

[00:52:57] So part of me wonders: in the grand scheme, sure, lots of people would love to live forever. But is it productive long-term if we just keep consuming resources?

[00:53:13] And spiritually, part of life is death. Value is created because of the finite time we have.

[00:53:25] So if we remove that finality—if we can extend life indefinitely, or 200, 300, 400, 500 years—does that change how we find meaning?

[00:53:41] We talk about the difference between Bitcoin and fiat: Bitcoin is valuable because of scarcity. Fiat is less valuable because it isn’t scarce.

[00:53:54] Preston Pysh: I love that framing. My immediate thought is: we call “normal life” 100 years—why is that normal?

[00:54:05] Maybe 500 years, maybe 1,000 years—I’m making up numbers to challenge the thinking: what is a “scarce” amount of time for a human?

[00:54:16] But I love the framing, and based on Elon’s comments, I think he agrees there needs to be some form of scarcity.

[00:54:24] Seb Bunney: And this leads into something I wanted to bring up. I stumbled across a study recently on ScienceDirect, and I’m going to read a snippet.

[00:54:34] The study is about stem cells—growing stem cells to assist insulin production. It says: the patient achieved sustained insulin independence starting 75 days post-transplantation.

[00:54:48] The patient’s time-in-target glycemic range increased from a baseline value of 43% to 96% by four months post-transplantation.

[00:55:00] This was accompanied by decreasing glycated hemoglobin—an indicator of long-term systemic glucose levels—down to a non-diabetic level.

[00:55:10] Thereafter, the patient maintained stable glycemic control with time-in-target greater than 98%.

[00:55:18] What this is really showing is that with stem cell therapy, they were able to enable the body to start producing insulin again in someone with diabetes.

[00:55:28] That’s mind-blowing. Four months post-transplantation, they went from being in a healthy glycemic range about 40% of the time to 96%.

[00:55:39] And after a year, they were at 98%—in a normal glycemic range.

[00:55:47] This raises a bigger question around aging. If we can go back to the root cause of illness—diabetes and insulin production, kidney disease, heart disease, autoimmune conditions—instead of symptom management…

[00:56:02] Instead of dialysis, stents, statins, or immunosuppressants, we could repair cells or replace them with new cells—essentially “young” cells.

[00:56:14] That completely changes how we think about health. Regardless of where you stand on longevity, the ability to support cellular regrowth—stem cell growth—feels profound.

[00:56:33] I’m curious for your thoughts.

[00:56:35] Preston Pysh: All these tech titans are treating it like an engineering problem. It’s complexity.

[00:56:44] If there’s one thing AI can do, it’s deal with complexity and find signal in what humans perceive as noise.

[00:56:51] Hearing Elon on stage say, “I’ve never seen someone with an old left arm and a young right arm,”—he’s approaching it from first-principles engineering. Based on that intuition, he thinks it’s solvable.

[00:57:05] And I think this is only going to get crazier over the next five years.

[00:57:10] I see David Sinclair’s posts on X all the time. He and Peter Diamandis are basically saying this is “in the bag”—that we’ll be able to extend life.

[00:57:24] In another interview, Diamandis said they think within five years we might be able to potentially double the human lifespan.

[00:57:35] Those are insane quotes. And I obviously don’t have the technical competence in biology to audit or troubleshoot where the technologies really are.

[00:57:41] But the comments being made—and Sinclair has credibility in the space—are wild.

[00:57:48] Seb Bunney: I’d be curious—because with Bitcoin you’re up against central banking and fiat currencies, and we’ve seen firsthand how hard they fight to preserve their existence.

[00:58:01] In longevity, you’re up against Big Pharma and the healthcare industry—some of the biggest industries in the world.

[00:58:12] They profit off people being sick. So what happens if instead of someone being on a lifelong drug—earning a company hundreds of thousands or millions over a lifetime—someone can get a $10,000 or $20,000 procedure that fixes the problem in one go?

[00:58:35] There’s a book I highly recommend called Selling Sickness. It talks about how much our system is structured around—

[00:58:43] Preston Pysh: By the way, at WEF in Davos today, it was announced that the U.S. is dropping out of the WHO.

[00:58:51] Seb Bunney: Perfect.

[00:58:52] Preston Pysh: Yeah. The World Health Organization—the U.S. is dropping out. This is crazy.

[00:59:01] It feels like a shot heard ’round the world. The course correction—the wind shift—whatever you want to call it—this year is monumental compared to past years.

[00:59:16] Seb Bunney: It’s mind-blowing, without getting too political.

[00:59:18] It’s mind-blowing to see Trump show up and basically say, “We’re not playing by these rules anymore.” We’re charting our own course.

[00:59:29] And you see it with the health secretary too—calling out Big Pharma.

[00:59:37] I think it’s important to question the status quo. That’s how we move forward.

[00:59:44] Preston Pysh: I saw rumors online that Howard Lutnick—the Commerce Secretary—was talking, and Al Gore was literally booing him in the background.

[00:59:49] Seb Bunney: Oh.

[00:59:52] I saw a tweet the other day—maybe two days ago—saying the Antarctic ice cap is larger now than when Al Gore released An Inconvenient Truth… or whatever it was called.

[01:00:05] Preston Pysh: That’s an inconvenient truth right there.

[01:00:08] Okay, Seb—give folks a handoff if they want to learn more. And by the way, we could go on for 10 hours on all of this. We’re not even scratching the surface of what we’re reading online. It’s insane.

[01:00:18] Give people a handoff, Seb—if they want to learn more about you.

[01:00:26] Seb Bunney: Hey guys, you can find me at SebBunney.com. My book is The Hidden Cost of Money.

[01:00:33] And Preston, I absolutely love these conversations. If anyone has interesting topics you want us to discuss, feel free to post them on X underneath the link once we share the episode.

[01:00:44] We’re always keen to dig into these topics. We appreciate you guys listening.

[01:00:44] Preston Pysh: Seb, there’s a new thing we do when we close out the show.

[01:00:49] We create an AI song that takes the transcript of our conversation and turns it into a song—in the style of the guest’s favorite music.

[01:01:01] So tell us your favorite style, artist, or song. When this airs, you’ll hear the song. What do you like?

[01:01:14] Seb Bunney: You know what? I listen to a lot of varied music, but in my teenage years and early twenties I went through a big grunge phase.

[01:01:23] Pearl Jam, Soundgarden—so let’s go with grunge.

[01:01:26] Preston Pysh: Okay, you got it. Alright folks, I hope you enjoy the outro song. Thanks for joining, and check out the show notes.

[01:01:35] Seb, thank you so much for making time.

[01:01:37] Seb Bunney: Thanks, Preston.

[01:04:37] Outro: Thanks for listening to TIP. Follow Infinite Tech on your favorite podcast app, and visit TheInvestorsPodcast.com for show notes and educational resources. This podcast is for informational and entertainment purposes only and does not provide financial, investment, tax, or legal advice.

[01:04:58] Investing involves risk, including possible loss of principal. Past performance is not a guarantee of future results. Listeners should do their own research and consult a qualified professional before making any financial decisions. Nothing on this show is a recommendation or solicitation to buy or sell.

[01:05:11] Any security or other financial product hosts, guests, and The Investor’s Podcast Network may hold positions in securities discussed and may change those positions at any time without notice. References to any third-party products, services, or advertisers do not constitute endorsements, and The Investor’s Podcast Network is not responsible for any claims made by them.

[01:05:29] Copyright. The Investor’s Podcast Network. All rights reserved.