And we have Balaji Srinivasan, the founder of Network State, who is our first special guest live in the Situation Room. Balaji, welcome to the Situation Room for the very first time.
Well, thank you. And technically, by the way, I'm the author of Network State, founder of Network School, ns.com, but I'm also an investor in MTS. Wow, how about that? Eric is going to RT that or something like that after this. So, I'm very pleased. Eric and I have been talking about media stuff for a long time. Eric's been crushing it and look, this is looking like it's going to be fun. Go ahead.
Why don't you contextualize where we are right now in this media moment, right? We've been talking about where tech fits in, how tech needs to build its own media landscape. We've also been talking about how the New York Times has continued to grow. How do you kind of make sense of where we're at in 2026 as you've been on this sort of, you know, 10 plus year quest to not just understand the media landscape but also build within it.
That's right. So, okay. Essentially, there's a long version and there's a short version which is tech and media actually share a common root in that we're both about the collection, presentation, and dissemination of information. Right. So the collection of information like their sourcing our data, right? Presentation, user interface or articles, dissemination, distribution whether on social feeds or newsprint or what have you, right? So at a very structural level, we are the internet first digital alternative to the 20th century printing press, newspaper kind of model. We are essentially a contender, a competitor for what is upstream. You know, if you ask the question of what is upstream of at least in the west or in the Anglosphere and you keep ask the question, what's upstream of a factory? Well, it's, you know, political this and it's capital and what's upstream of that? Well, it's eventually you get to money and media because media is upstream of politics and money is upstream of media, but media is also upstream of money and that's in Ouroboros where that circle eats its tail. So the venture capitalist and the journalist, right, the tech and media are actually sort of locked in a struggle for what is upstream, right? And that's like a good way of seeing it like who's at the control panel flipping the switches, hitting the buttons and so on and so forth. And with the advent of the internet over the last 10 years, really the last three years, essentially speech has actually been freed for the first time in our lifetimes because until essentially the 2010s, like there's that saying freedom of the press belongs to those who own one or never argue with a man who buys ink by the barrel which meant that unless you had a newspaper that you inherited or a TV license or a radio license, all of which cost many millions of dollars, you could talk to your neighbor in the 80s or 90s, but nobody could hear you, right? With the advent of social media and Twitter and blogs and all and so forth, all these voices that previously had no distribution had distribution, which caused a cacophony and all of this kinds of chaos in the 2000s and 2010s and early 2020s. And there was a counter reaction that tried to censor all those voices and then a counter counter reaction that uncensored them with Elon's purchase of X. And that has brought us to the present day. And one of the consequences of that was the media was though we didn't set out to do it like Twitter set out to basically be tweeting breakfast, Facebook set out to share likes and poke people or what have you and those ended up disrupting classifieds and disrupted legacy media, disrupted print media. There's a great graph of the print media disruption. So, as a consequence, imagine like a kid who just grows to be 6'6", 250 in an elevator and squashes everybody against the wall, right? That's like what the internet was, right? Where we just grew and just added all of this muscle mass and we didn't mean to do it, right? But became really, really, really, really big and went from cute gadget makers and toaster makers to, in my view, the single most important force in the world that's still underestimated. You know, Orwell the writer obviously in 1984, he had this saying which is it takes an enormous effort to see what is in front of one's own face and what's in front of our own face basically every single moment of the working day is the internet. That's right. You know that what's upstream of AI? The internet. What's upstream of drones? The internet. What's upstream of your finance feed? The internet. What's upstream of the data center spend? The internet. It's internet first, right? In the sense of mobile first or micro first, internet first, internet first. Where's your community? It's internet first. Where's your business? How are you sharing your business? Internet first. How are you finding information? Internet first. So that is actually the organizing principle like fish can't see water. We swim in a sea of electrons. And that was not the case in the 80s and 90s, right? We have essentially been teleported into the matrix. Right now, you watching this, hearing this, you are in the matrix. Shout out. We are in the matrix. All the fans. All the viewers. Yeah. All the fans. Exactly. That's right. We're monitoring the situation where on the internet, right? So, the global internet, you know, the closest precedent to it, by the way, just to digress on this for a second, is the ocean. And the reason is there's something called the law of the sea. The law of the sea governs how, you know, because you have ships that are going from Britain to Hong Kong to Brazil and so forth. And so, you have international waters and what country controls that? And you could very legitimately have something that was thousands of miles away but is flagged British or Portuguese or what have you. And so for hundreds of years there's something called the law of the sea that kind of governs that. Like when you're sending a packet from one port to another port, what law governs that? And we actually use the same words today. When you open up a computer, you have one port and it's sending packets to another port, right? And so how can that information, how can those packets be sent? What can they do? and so on and so forth. The rule of code on the internet is like the new rule of law. It is the new law of the sea and you can think of the cloud as like the new oceans in this way right ideally a demilitarized zone but of course there's also navies right so with that macro context let's bring to this moment you know those zombie movies where at the end of the movie the zombie opens their eyes and claws their way back like this right okay so there's a guy who's a good guy named he's a good analyst called Philip Le Moy. Okay. PHL43. All right. And maybe we can put this link on screen. Can we do this? Hold on. Let me send you guys a link. Bang. Bang. Okay. Can you guys see that? Can we project that on screen? Let's take a look. PHL43 is a very smart guy, good poster, French guy. Okay. And tell me when you got this on screen. All right. So, I'm screen sharing. Can we get this on? There we go. All right. So, I like both Nate and Nikita. Okay. And they were basically talking about link boosting and Philip found something important. So, if you scroll down and look at these graphs, just click the graph. The first graph or the second graph will say it. All right. So essentially like a zombie movie basically the New York Times their distribution collapsed after peak woke in 2020 2021 2022 and after Elon took over X and he deboosted them they basically went to complete zero 2024 2025 but with link deboosting turned off and the kind of repositioning of the American left to basically be less woke and there's more I can say about that. But they are now trying to occupy some of I shouldn't say the center again that's not exactly right but let's say facts that Republicans don't want to publish right for example facts about the war or things like that right now boom look their distribution has come back right and so the tactics that worked five years ago like going direct we do want to still use them don't get me wrong it is important to speak for oneself and so on and so forth we don't just want go direct. We want to prove correct. Okay, that's the new that's the next five years. Why? What happened in that intermediate era in 2022 to 2026? AI. And look, there's a lot of good about AI, but there's a lot of bad about AI. And if I love AI, I also hate AI. Why? I love AI because when you know I've got a rule, it's called no public undisclosed AI. What that means is private use of AI for search for code when you're not trying to bamboozle somebody, not trying to get one over on somebody. Fine, good. It's great for research, all that kind of stuff. Awesome. Public disclosed AI, when you put out a video that's obviously a cartoon or an animation or something along those lines, it's obviously AI, so nobody feels you're fooling them. But public undisclosed AI when it is it's not this, it's that text, right? Whenever I see a slide deck that has AI content in it, it's a new Lorem Ipsum. It's Lorem AISum, right? Like just add an A. It just shows that either they're dumb or they think you're dumb. they're dumb because they can't tell the difference between normal text and AI text or they think you're dumb and they can get one over on you and spend a little bit of effort to just flood you with a bunch of words and then as soon as I see that I ignore it and I send it to zero because it takes only a little bit of effort to send a whole, you know, just mash of a ton of words and I don't know if any of them have been checked or read through or whatever. And so anytime I even detect there's a hint of AI in a communication to me, I set that to zero and I set that person. It's a significant downgrade. And I think this is going to be baked into social networks where something like pangram.com, you know, now people will tell me and of course it's true that you'll never have a perfect AI detector and so of course that's true. However, you can do really well if I can recognize it with a naked eye and you can see it's not this, it's that. Many people are just going to, you know, how like Windows has a background where most people don't change the default. Right. Yeah. Just like that. Most people don't change the default settings on AI. They don't prompt it so aggressively that it's actually creative. Once it busts out of that, I actually don't have a problem with it because then it's like so good that it doesn't look AI. But if it's detectably AI, right, if they haven't polished it enough, then if it's detectably slop, let's put it like that. Trying to find it. But then I think it was this. Yeah, there was a recent study about AI writing in newspapers becoming like increasingly common. And yeah, only five out of 100 AI written stories actually said they used AI. And this was October 2025. And AI adoption has gone up quite a bit since then. Quite a bit. and go to pangram.com for example, pan. So I think this is pretty cool. I think something like this is going to become more and more and more common. And basically, you know, it's alien versus predator, right? And an enormous part, see one thing that most people don't get when they talk about AI and the jobs and so forth, right? I have a post, you can look at it if you go to balajis.com. Just balajis.com. Exactly. And go to read it first and click AI is polytheistic not monotheistic. Yeah, that one. Yeah. So I think this is a good I think it's a good post. Right. But basically just to make a few points here. First is the entire AGI framework that is sort of implicitly monotheistic. It presumes that there's going to be a single all knowing all powerful AGI. But in practice, we're seeing a polytheistic AI of many decentralized models that are good at different things. Number two, AI doesn't take your job. It lets you do any job because you can be a passable sound effects creator or UI designer. You're not amazing, but you can be like, you know, you're jack of all trades. Number three, AI is amplified intelligence, not artificial intelligence. You know, the smarter you are, the smarter the AI is. Number four, AI doesn't do it end to end. It does it middle to middle because you still have to prompt it. You still have to verify it. Then number five, AI doesn't take your job. It takes a job of the previous AI. Okay. What that means is like the way I process all the change that comes with AI models. I have a spreadsheet which has rows for what's the best AI coding model, what's the best AI image model, best AI video model, and I have a column which says January 2026, February 2026, March 2026. And when I determine that a new model has come in, that Codex has taken Opus' job or vice versa, then the AI takes a job of the previous AI. Okay? So AI doesn't take your job necessarily. It takes the job of the previous AI, right? The next is AI is better for visuals than verbals. So it's better for the front end. It's better for movies and things like that because your eyes have effectively built-in GPUs. So you can immediately detect subtle things with the hands or this doesn't look completely photorealistic or whatever in a way that it takes you much more time when you're reading code. There can be a subtle one bit off that completely changes a cryptography result that it's very hard for you to detect with your eyes. You have to do system two versus system one thinking right and Karpathy also talks about he agreed with me the verification gap is actually a very big deal. Right? Then another point I made again this is all obvious now perhaps but killer AI it's actually already here it's called drones and every country is going to pursue that. So, the whole concept of oh my god, let's regulate the chat bots and the image models. I'm not saying that one doesn't want to have some counter measures on that, which I'll get to in a second, but like the idea of what's going to kill you. It's going to be the drone that's got to actually have a physical actuator shooting you. It's not just going to be like the super intelligence that makes things, right? The next point, AI is probabilistic while crypto is deterministic. AI and crypto are actually complementary. Crypto is what AI can't do. Okay? because AI can solve partial differential equations but it can't solve chaos it can't solve turbulence right when you've got ODEs or PDEs that are chaotic or turbulent respectively it cannot forecast the outcome of a hash function because these are things which it provably cannot do what's computationally irreducible this is a big old firm idea I think is kind of underrated in the discourse That's exactly right. Sorry, I should say it can't easily calculate the pre-image of a hash function, right? So there are mathematical and physical bounds on an AI like for example, you could have a turbulent source of entropy that would be something in that that would go into decision algorithm that AI provably could not forecast because it couldn't forecast what was going to happen to the turbulence, right? So it's simply not omnipotent. People are treating it as if it's omnipotent. It's simply not, right? even like a pendulum with a few different weights on it there's different ways you can quickly get into chaotic or that kind of behavior. So empirically right now at least and you can argue this point whether AI is centralizing or decentralizing but I think there's so many AI companies and there's so many AI models out there that it's pretty hard to put a lid on the thing. And then finally if you go to the end the optimal amount of AI is not 100%. If you scroll down a little bit further, right, 0% AI is slow, but 100% AI is slop. Just having that concept in mind means that in almost any process, you do not want to have 100% AI, but you probably some and frankly sometimes you want 0% AI. you often want to just learn offline with pencil and paper and then speed up with AI because AI is a shortcut and like any shortcut it can be overused to the point that you don't know how to take the long cut right okay it's a laugher curve but for AI and then you know I talked about how referential is this whoa we're referencing Treninberg on Treninberg A6Z podcast mentioned yeah A6Z podcast that's right this is recursive right and fundamentally my worldview on AI is it's constrained AI it's economically constrained because every API call is expensive and now it's energy constrained, right? It's there's so many competing models. It's mathematically constrained because it can't solve chaotic, turbulent or cryptographic equations. It's practically constrained because you have to prompt and verify it. It does it middle to middle rather than end to end. And it's physically constrained because it requires humans to sense context and type that in via prompts rather than gathering that all for itself. And this is very different than the people who look at AI as AGI, a deus ex machina that's just going to solve everything, right? To be clear, this is as of this time of writing. Maybe somebody overcomes it, but at least I'm describing what the constraints are on the current generation, right? You could in theory unify the probabilistic system one impressionistic thinking of AI with the logical system two thinking of that computers are good at and Claude Code and so on is starting to get there perhaps, but I think it's still something where context breaks down and so on. And also somebody, you know, that famous graph that people show that it can do longer and longer problems, the meter graph. Yeah. profiles about that graph. Yes, that particular graph I saw a very good, you know, once in a while out of a thousand replies on X, there's somebody who's actually, you know, gem award, right? How do they put it? Right. Gem, you know, like Okay, gem alarm. Yeah, exactly. Right. Fine. So there was a very good counterargument that actually said that this study does not purport to show what people think it's purporting to show and that the task completion it's like it's much more questionable than it actually looks. Right. And this is important because a lot of people's paranoia and stuff is levered on this. Yeah, that's it. I think that's it against the metagraph. Right. Yeah. Can we pull that up and pull that up? Right. So yes, this was like this deserves more people looking at this. Okay. But fundamentally he says actually the graph doesn't really look like this because a lot of the tasks aren't actually being completed and b like what is completion you know they're tldr now I don't want to overstate this because it's certainly obviously true at a guest level that if you use Claude Code or Claude Co-work or something like that it is possible for AI to complete longer tasks and more complicated task than it could a few years ago. That's clearly true. However, it's also true that you still have to supervise it a lot and check the output. And it's a little bit like a spaceship that can go in any coordinate direction that you point it in and move very fast, but you still have to keep pivoting it and give it the route that it's moving on. It's like a car. It'll take you there very fast, but then you have to still give it a direction. And now the new thing is everybody's got a car. So now you've got a car race as opposed to you just being able to teleport somewhere, right? Okay. So this guy is pretty critical of it and I think this is worth reading carefully and maybe figuring out what the true graphs are from his critique of it. Taking into account that yes, there's something real with Claude Code and Claude Co and so forth. Definitely. But also taking into account that it basically speeds up I don't know. I don't know if what your guys view is on this, but I find that it's basically like managing, you know, where you have to write the whole thing up and provide all the context and that's actually a lot of work, the context engineering, you know what I mean? Because you're sensing the world and you have to articulate that in clear written English and describe exactly what you want and check the result. The prompting and verifying then just takes up all the work. That brings me into my next point which is AI actually destroys arguably as many markets as it creates. You know why? What? Because for example take resumes right? AI makes it wasn't that hard to make a resume before. AI does make it easier. You know what AI makes it hard to do? It makes it hard to verify a resume. AI makes it easy to write an email. It wasn't that hard to write an email before. But now it makes it really hard to filter them. Okay. So many markets between strangers, between tribes, right? So recruiting, sales, marketing are being destroyed by AI spam, AI slop, AI scam, anything which is between economically disaligned tribes. There's so much AI spam that that channel is now defeated. It wasn't built for that level of adversarial behavior. It wasn't built for a channel where 99% of inbound is so fake that it can beat the probabilistic fake detectors like AI spam detectors broken, right? Your normal filters broken. So what happens is people retreat to just digital tribes. They retreat to deterministic trust and it's only the warm intro that gets through. Okay? So the sales market's being broken by AI. recruiting is being broken by like people will sometimes show me these products and they're like I've got this AI agent and it just spams your resume to like a thousand companies. I'm like great this is just going to like leech out what trust remains in the ecosystem where everybody claims they're a machine learning superstar blah blah blah and normally pre-AI you could just read a resume and it was relatively scarce to be able to write well and now you have to carefully I mean people do keyword stuffing it's true but still like good writing was relatively scarce now you have a situation where it's much less scarce but scarce is concise writing fine okay but we need a completely different paradigm for sales, for marketing, for recruiting. I have a whole thesis on this. I think that we're going to have to literally rebuild like web 3 actually becomes a thing where the open web, web 1 just gets completely corrupted with AI and web 2 becomes walled gardens. But web 3 is the hardened cryptographically provable signed web, the open web, where things are on chain and you've got signature trails and you can actually it's like a diamond. It's hard. AI is like a slithering thing around this diamond of cryptography that it can't fake, right? So, you can prove that so and so signed this and AI can, like for example, a letter of reference can be cryptographically signed and you can show that so-and-so endorsed you as opposed to saying just claiming that it happened, right? That's a longer topic and actually we're building stuff related to that network school which I can talk about. But the point is the entire concept of oh it's just a wand that's going to solve everything actually you know for example if you take just that sales or marketing or recruiting market the level of economic damage that AI did there is actually probably larger than the benefit because it introduced easy fakes into a system that was not calibrated for that amount of easy fakery. Yeah, it reduced the cost of generating a resume, but it radically increased the cost of verifying a resume such that less hiring happens with me so far. Yeah. Just anything that requires some kind of like proof of work that happens as like writing text is kind of gone now. That's right. So, what that means is AI will create lots of jobs in proctoring and verifying, right? you have essentially the whole KYC economy, the entire you know the like all the stuff that people do for Face ID, for authentication, biometrics, all of that stuff merges with, you know, proving you're a human, you're unique, all those things merge together where proof of X becomes like this critical critical thing. Can I log into a system? All the identity, authority, all that stuff will increase 10x from where it is today. Maybe 100x, right? Prove that you are human, that you're unique, that you have the credential that you say you do, the endorsement that you have, you do like have multiple things. It's not just one thing. It's like all of these different signals. Prove it cheaply for yourself, but very hard to fake. Example, having I won't give out your email, Eric, but blankas.com. That is easy to verify, hard to fake, right? because you'd actually have to break into the cryptography like an email verification at a6z.com is something that actually carries a lot of signal with it where it's just one click for Eric to verify that he has an A6Z.com email. But it's challenging for someone who didn't have DNS access and the private keys effectively associated with the a6z.com domain to get the MX record and Gmail access to be able to issue that user that email username. Does that make sense? Right. So easy to verify, difficult to fake becomes a critical thing in any system that deals with strangers, which is lots of systems. The internet genuinely becomes an untrusted environment. I even haven't even gotten to, you know, fake video, fake photos. This gets me back to media, right? The reason you can we go back to that Philip Le Moy post. Yep. Yep. All right. Here is the pull it up. Yes. So here is the re part of the reason that's happening. It's not just link boosting, right? It is something where like you know I kind of here let me put this article on screen. Okay, here we go. New one. You know old old Balaji like I'm a reality shell, right? So I'm always about just observable reality, what our constraints are, what the battlefield is and so on and so forth, right? So Jason who I like you know is if you scroll all the way down right he is basically posting something that I was posting in 2021 which is go direct and so and if you do from colon balajis yes you can see go direct nothing wrong with that actually in fact allin has done a good job of that all great right however this is necessary but not sufficient the reason is as follows if you go back one the issue is that due to AI right and also due to the fact that in my view in some ways significant ways MAGA is overcorrected to the right in a huge way. Now, a people can't tell what's real because when you see a photo or you see a video, it literally you literally don't know if it's real or not, especially if it's some crazy bomb or something like that. Maybe it's often like some bombing from Turkey or something like that that's which is that a lot of these things are like obviously extremely fake and then they get tens of thousands of likes anyway. and X as it exists right now doesn't really have a good mechanism for filtering that garbage out. That's right. So you have all the possibilities. A obviously fake stuff that we know is obviously fake. B nonobviously fake stuff. C stuff that's real but it's faked in terms of it's a video from this time and place and presented a video from that time and place and everything there. And that makes you very cautious about like it used to be that video and photos were pretty hard to fake. Though for example like the Brazilian fires like of years ago there was a big aerial photo of a fire and it was being used to justify like invading Brazil and it was actually a photo that was from a journalist from like years and years earlier and the time stamp that showed that it was here let me I'll find you this thing hold on it was like a fake photo that the Atlantic thought was real right and because of that they were there's literally a guy who was like calling for invading Brazil on this right the three most photos Amazon Fire are fake yes here we are so if you click this right you just it's pasted in a chat to you all right this gives you an example of yeah so so click my click my thing there okay so just to show you why timestamps and cryptography are so important right click click those images Right. So the Emmanuel Macron used one of these misleading viral photos 40,000 times embedded in a New York Times article. The photo was taken by Lauren McIntyre who died in 2003. In other words, the time stamp that showed when that photo was taken, it was taken many years ago, right? So these guys were saying like, you know, the case for territorial incursion in the Amazon is far stronger than the justifications for most war. Basically, literally fake photos almost justified some crazy war on Brazil in the Atlantic. Okay, that's bad, right? So, and New York Times Atlantic like this is a canonical example of where decentralized cryptographic truth was able to defeat corporate journalism. Okay. And so we have to go to our and I can give more examples of this, right? So basically you know three examples of here let me show you this. Here you go. So, here's one. Okay. This now was from years ago when photos were hard to fake. Now, this could be faked. Okay. But from years ago, Vitalik was able to show proof of life when people were saying that he was not real. Okay. If you put that on screen, tell me if you got that on screen. Okay. Then click into that thread underneath. See, you see I've been thinking about this for a long time. This Yeah. Just click into that. Yeah. So I mentioned the Brazilian photos. Scroll down a little bit further. I'll show you another example. Click this one with a graph there. The Yeah, that one. Right. So, this is something where Tesla had the vehicle logs where there's a fake NYT story by a guy named John Broer that claimed that, oh, Teslas were, you know, they ran out of battery on the side of the road and they got a photo of a Tesla being hauled onto a truck and Elon actually had the instrumental logs and was able to use digital evidence to disprove the verbal narrative. Right? This, by the way, is one of our core strengths, numbers over letters. Okay, we can actually be quantitative. We can verify things numerically over the internet. That is actually a power that journalists are generally pretty bad at, right? Digital history, unfakable cryptographic history. Go a little bit further. I'll give you a third example. Okay, so yeah, do you see this defendant proved that its content so this is in a Chinese court. There's a patent suit and the defendant showed that basically Yeah. So this is like eight years ago. Defendant proved its content was really produced by showing two hash values. One on the Bitcoin blockchain. It had hashed relevant data onto relevant chains and proved by way of time stamp that infringement would have been impossible. Isn't this cool? That's cool. Rule that data stored on the blockchain is admissible as evidence and trial due to the characteristics of the tech. Exactly. Eight years ago in China. So this stuff is actually almost a decade old at this point. You know the saying the future is already here. It's just not evenly distributed. Right. So it's very important that and this is a broader thing in tech okay like we as tech guys feel okay yeah we're generally have good motives and so forth but the point is to trust us the point is to not have to trust us okay the point is to have systems that are auditable by math that anybody can look at and the reason that they would trust what we're doing is they don't have to trust That's what we're doing. They can literally download it. They can cryptographically verify it in the same way that you can check whether a website has SSL via the HTTPS lock symbol, whether you can check whether somebody has an email address, right? whether you can check whether someone has DNS access with a TXT record. You should be able to check via chain check whether something is put on chain. Another way of thinking about this is, you know, if you've sent somebody something on, I don't know, a blockchain, like Bitcoin or Ethereum, have you ever sent them the either scan or blockchain.com record to show that you sent the money? Have you done that? I don't think so. Eric, you've definitely done that. Yeah. Yes. Right. It's like, hey, I sent you the money. Here's a transaction receipt. Right. Yep. And for example, if to give you a concrete example, if you went to Grok and you said, Hey Grok, tell me, you can do this live if you want. You say, Hey Grok, use onchain information, okay, to document the FTX hack that happened in late 2022, okay? And give all references. Grok will tell you about a hack that happened of FTX at the time all the FTX drama was happening and in particular it will link to block explorers that have onchain records that show that that event for that amount there it is it's happening right there it is right so this is what's called onchain media crucially the source of truth is not the New York Times and Sulzberger and Sulzberger employees. It is the blockchain, right? It's something that everybody has free universal access to. It's not paywalled. It's not something where you have to pay Sulzberger for him to tell you what the truth is. It is something that anybody can see what the truth is. Right? See onchain evidence and timeline, right? Initial inflows long, right? So, you can actually click and you can diligence it yourself. And if you ask at the bottom and say you know give the you know like click exactly see give the either. So this is something which can be done for those kinds of things that are onchain right. So yeah exactly right. So but like plowing blockchain money trail. So can you if you go back to the thing and you say give the specific either scan give an example specific either scan link right and this is the raw of the raw I mean it's not the raw of the raw data but let's say it's basically a URL right so click that right so that's something where with a little bit of technical understanding you can literally look on chain and see what happened it went to the FTX account strainer and so on and so forth right so you can pull all this data and analyze it yourself you know have you heard the term OSINT yeah yeah we're all about that so this is you're all about that that's right so this is like OSINT but for onchain stuff. So onchain intelligence is like OSINT. The difference is and this is going to be important in my view. Right now the verifiable information in crypto is mainly financial. It's this kind of thing where there's a major news event and some of the key things key things were happening on chain with Farcaster and you know so Dan Romero who's a friend of the show Eric I think it's fair to say right our friend right yeah exactly so Dan came up with a protocol called Farcaster and I actually think it's still an amazing protocol in fact you know he basically you know he sold and you know he gave over management to the protocol to someone else but sometimes protocols have sort of a lifespan that's beyond any one company and so him you know going and doing something else with the protocol being around it's still this very very powerful protocol that I want to do a lot with which essentially is like Twitter but onchain okay and allows anybody to post anybody to read anybody to write and crucially to add code to like verify photos verify videos or at least give onchain evidence so imagine if much more was onchain not simply financial data, but social data, much of what goes into an article, you know, years ago, I said something as a joke, but then it became true. I said I gave a talk on what I call the ledger of record, and I said, you know, just like a sports article, right? Here, I'll find the talk. Ready? I said just like a sports article is a wrapper around a box score, right? where you you know basically you have a box score where you could automatically generate an article just like a financial article is often a wrapper around a ticker symbol, right? A news article is often a wrapper around tweets, right? And I've been thinking about this for years. You can see this post here, right? And then there's like a I just sent you a link there, right? and basically on let me send you the ledger of record thing. So I've been thinking about this for many years and the concept is that we can separate fact from narrative. If you have a feed of onchain data and you have AI on top of that feed, then you can autogenerate stories that have no bias because the AI agent summarize the onchain data and it can summarize it in your language with your political orientation and you can just change the reporter by flicking the AI flags, but you can also diligence it all the way down to the raw data which when it's financial, it's onchain and eventually when it's social, it's onchain as well. Does that make sense? Right.
I just noticed, by the way, sorry to interrupt. I just noticed we have 53K live viewers on the stream right now. We have Oh, wow. It's almost twice as many as Alex Jones who has what a million and a half. He's been around forever.
Hold on. Should I RT this? Should I RT the stream? I haven't actually even RT it from my feed or whatever.
Yeah, go for it. Please do.
Hold on. All right. Send me the link. I'll do that.
Did not expect that. It's okay. So, the reason all this stuff is Yeah. So, okay. Sorry, I didn't mean to interrupt. So, go ahead. Say what you're gonna say.
Yeah. So, here is the stream link.
All right. Let me put this on screen. One second.
Is it on your feed? Whose feed?
It's on my feed. It's on Eric's feed. Also at MTS Live.
MTS Live at X MTS Live. All right.
Yep. Go there.
Uh, oh, that's Ministry Training Strategy. No, MTS Live. We're meta monitoring again. All right, here we are. Monitor situation. All right, so it's the All right, here we go. Bang. All right, so so boom, boom, boom. That was just QT. So on the new monitoring the situation stream with Eric. What's your what's your uh Twitter?
Uh Theojaffy. t h o e o e o j a f e. All right, there we go. Bang bang. All right, good. So, all right. So, coming back, what's my point?
Um I have a talk on the ledger of record. You can um you can put that on screen. Okay. There was also a um a good post actually. Hold on, let me show you this. There's um so this is the reason is I you know I got to thinking about how would you actually adjudicate truth. If you just take a look at this, this is my kind of you know Twitter thread on it, right? Um but I will show you a pretty good uh post by a guy who he actually you know deleted it but then I I there's an archive version of it. Um let me see if I can find it. The um basically if you look at this post, this is from 2020, right? You see where it says the ledger. Do you see it on screen?
Yeah. Yes. The set of all cryptographically signed feeds of onchain data. Social media feeds, data APIs, event streams, news, RSS. It'll take years to build ultimately be become the decentralized layer of facts that underpins all narrative. We're a lot closer to this if you scroll down a little bit further. Right. So click that if you just click into that right. So right now people were thinking about this in terms of payments data right retail payments backend system bank payments market data events data but obviously poly market all these prediction markets they need to have news data on chain right so take all these oracles you put them on chain for trading purposes but you need them for other purposes as well the the financial stuff incentivizes the information feeds right like in a sense why does Bloomberg have a new service so people can in a sense bet on the Bloomberg terminal right so the Oracle services that actually put this information on chain help resolve smart contracts for various kinds of markets. Right? So the news, the media actually influences the flow of money. There's a vortex that is pulling verifiable information on chain because the market wants it there. And in a sense, what the blockchain is is like an armored car for information where you can transport that information on chain. And so if you scroll down a little bit further, for example, look at this thing. Again, this is like six years because I've been thinking about this for a long time. So the weather channel posts cryptographically. If you click into this, right? Um this shows all these different kinds of feeds that could be posted. So the weather channel could post sign feeds of weather data. Redfin post real estate transactions. Forsquare's location data. They can make them free or they can crypto paywall them. Right? And uh so now you have this stream of the raw data and then you have AI on top of it that turns that into legible text. Now, in 2020, I only saw GPT2 and may maybe GPT3 had been out by that point, but I could there was actually a company called Narrative Science, okay? And Narrative Science, which went bust, but it was a cool company. Okay? They had something which uh let me show you what it looks like. Ready? Take this image. It's a little small, but put this on screen. Okay. narrative science show the possibility of something like what we have. Right? Again, you know the thing about the future is already here. It's just not evenly distributed. Right? If you put that on screen and zoom in, okay, zoom, zoom, zoom. So that would take, it's a little blurry, but basically it would take your financial data and it' say like, you know, bookings were strong this month with new pipeline blah blah blah. You know, it kind of give like sports commentary on top of your raw numbers. Does that make sense? Right. So this it was like a domain specific AI, right, that would turn numbers into legible text with me so far?
And that was there back in the day. So I saw that and I knew about that. And I'm like, you know, as AI improves, it's possible that we could take streams of not just financial and other data and automate the process of turning box scores into sports articles, the process of turning weather feeds into weather reports, turning financial feeds into financial reports, turning lists of tweets into political articles. And what I didn't know is how fast it would happen, right? And I've actually already, you know, I actually built some prototypes on this already. For example, here's a fun one where um so I I set this up uh a year ago. Actually, like as soon as ChatBD came out, I knew the potential of this. Hold on. Um it's pretty funny. Uh we we basically did a an NYT um but an AIT. We did two versions of that, right? So, hold on, let me find this. We set this up as a replet bounty and it was so so so fast. In fact, hold on. Where is this? So good. Um, here. This is the This was the original bounty. I'm pasting this into you and then I'll show you what happened to it. Ah, yes. Here we go. Bingo. Okay. So, I put that bounty up and then there's a kid who literally learned to code that year in replets like Summer of Code. Okay. So, I said actually go back go back for a second. Hold on. Go back. Uh, so if you just click into my tweet first and then we'll play this video, right? Just clicking my tweet. Um, scrolling a little bit further. Yeah. Prompt with a few tweets. Output an article in the style of NTW, etc. Make a set look exactly the same. Compare the speed and cost of having an AI do it versus a traditional NPC. Lol. I think that's funny. Okay, fine.
All right. [laughter] So, do view quotes, right? No. Yeah. Uh yes. And then uh exactly. Right. So then there we go. Pow days. Look at his look at his thing. Okay. And let's zoom in on this. Let's maximize this. So basically post in a tweet, right? And you know, basically he'll hit submit, right?
This is back when it was called Elon Musk on Twitter. That's funny.
Yeah. This is this is four years ago right after Elon took over Twitter, right? The GPT times, right? And so now he hit submit and literally it turned this into a image and headline that looked exactly like you know Coca-Cola has been a part of America. The same tone and everything. Okay, that was four years ago, right? And the full thing was open source and so on and so forth. Now I actually did a follow-up on this. So as much as I love Twitter and Twitter's great, I did a follow-up on this where basically um the main issue is Twitter's still a commercial service. And so it's a closed service and and and and so on and so forth, right? So um and it's fine. It's good. It does a lot of good stuff and you know I think it does a lot of good for the world, but we also want open source, right? So a year later I did this guy, the Robojouro, right? And um here was the result of that. This was actually on Farcaster. And then I'll show you the result of that which actually was even better. So tell me if you got that on screen, right? at the Robojour, right? So, you scroll down a little bit further uh blah blah blah and then go to the quotes or whatever by let me see if I can load this.
Bye. The thing I want to ask you though is why is New York Times ripping? What despite their credibility being uh sort of compromised by a large set of people, why is their business crashing?
Isn't it just the games? Well, it's too The thing is they traded one Mark Anderson for 10,000 essentially, you know, Democrat party members, right? So, basically there's all of these party faithful who, you know, the equivalent is, you know, you know, the the Chinese Communist Party has a has an outlet. It's called Cushia. Do you know what that is? Right. If you go to uh this, right? If you go to en.qory.cn, okay? and they put out a bimonthly called Cushia and it's actually very it's it's turgid party doctrine but if you actually want to know what China's doing this is in English and you can actually read it and you can actually learn a lot about you know just just to know what they're doing they put it all out there right so the Kusha subscriber is a lot like the NT subscriber it's like the party member in China who reads the party newsletter and says they're expected to like repeat what the party expects them to say and to change their narrative as the party changed their narrative. We're for Russia. We're against Russia. We're this. We're that. Right? In the same way, the NT subscriber is like, you know, whatever the Democrat party wants, we will say, and they just kind of go back and forth in the wind like this as, you know, they they're for this, they're against this, and they're for this again, right? They acknowledge something that was taboo several years ago, and vice versa, right? So, once you kind of see NT equals CCP in this sense, right? Um the what they did is they traded the the dissident, the people who they couldn't control, the tech guys for thousands of essentially uh NPC subscribers and so they traded power for money, right? That's what NYT did. They lost influence over the center. They lost the center, but they did it's a little bit like honestly like uh like Alex Jones or something like that, right? they they went to the, you know, they built a an audience um and they've just gone farther and farther into that audience and they never say anything. They they have what people would call audience capture, right? So, um the uh All [clears throat] right. So,
I think that like a lot of a lot of the reason that the New York Times business is booming now is just because of the games like Wordle, you know, the mini cross word, the Sudoku. cooking and stuff like that.
Uh I think like I think more than half of uh New York Times app screen time is games or something. So I think without without the games uh they would be doing a lot worse financially. That's true. That's exactly right. But basically um the thing about that is they cross subsidize it in the same way like everything gets cross subsidized where they have the fun stuff and then they you know have the eat your vegetable stuff. The big thing that they did is they lost the um uh they lost tech, right? They lost Anderson. They lost all kinds of people who are neutral who thought that they were actually, you know, good and and so on and so forth, right? And um and here we go. Let me show you this. Ready? Hold on. I found the I found the link. Here we go. Hold on. Eric, maybe you can put that on screen here.
And then I got another topic for you.
Yeah. Here, I'll just put in the same chat with the Here we go. Yeah. So, this was fully automated less fair journalism, right? And um there's like a video version of this which I can find, but basically you get the the concept, right? There's actually a nicer version of this actually in a later thread, much nicer version. Let me see if I can find it. Um, and the the point being that this was uh this was something that uh we could take the open-source feed and we could turn it into something which uh was looked like an NYT kind of thing, right? Hold on. I really want to find the actual version of this because there's like a better there's a better version of it. Um problem is I've posted so much that it's kind of hard to find. Let's see if I can find it. Hold on. It's like from two years ago. Give me a second. Ask me a question while I'm fighting this. Go, go, go.
Bye. I want to hear your thoughts on clvicular. Clavicular.
What about Oh, about clavvicular. Yeah. I mean, I feel like you were on looks maxing early years ago. Well, I Well, I mean, I'm not I myself am certainly not a looks, but what I meant to what I do agree with or what I think Okay, what's one cheer for it? because I think biotech is um the what I did say is that uh the uh bro science right was going to become a huge huge huge thing right and um and it did and basically uh that is going to become something which is you know a a major market right um something which it's a little bit like you know how Twitter was uh like a small thing that became a big thing and it went from you know just tweeting your breakfast and so on to you know the center of the world economy where people are are quoting you know right so in the same way bro science goes from um like the the type of thing which people think is sort of a joke to the thing that is completely completely real right and um and and I think that's already we're actually midway through that with peptides and so on and so forth And we have currently ompic for uh for weight but I think we'll get hopefully ompic for longevity maybe ompic for cognition ompic for many other kinds of things we'll get like large effects size drugs and uh and that could be a really big deal right um and so uh hold on I want to find this thing um so so that's why
super bro sciency he seems to know align with the mainstream line
yes but it's it's something where the um there is a aspect to what he's talking about which is self-improvement. Okay, by the way I found this I found this thing finally. Hold on. Wait, that's not it. Hold on. There we go. Sometimes copy paste work. All right. Bingo. Click that. So this just shows a prototype. Let me explain what you're seeing on screen. If you click the first one, the first image. So what that did is that takes the opensource completely free stream of onchain posts of the Farcaster feed and automatically generates a front page style to look like NT but that's completely free and you can click into and you can um you know see the articles and so on and so forth. Right? If you go to the right okay uh you know you can compare that to the NYT homepage. Right? Is this live on
uh we're going to we're going to actually bring it back, right? So So stay tuned. If you're interested in this kind of thing, I'll announce the launch on monitoring the situation, okay? We're going to have I'm going to become hopefully one of the largest funders of free open-source citizen journalism in the world where anybody from anywhere can become a reporter by posting online and then if we include that in content uh that we have commissioned where we're basically uh we're covering the stuff that is useful for technocapitalists for example what happened in India with thorium is so important right they've actually figured out how to get breeder reactors working. All kinds of countries should know about this. Everybody should know about thorium. There should be tutorials about, you know, exactly what they did. And so
there's a really good uh Sam Monella Academy YouTube video on thorium, which everyone should watch. He's very funny. Thorium rocks.
Oh, great. But that's nine years ago. So, but but that's good. So, so the thing is this, you know, it's weird that there's so much coverage of like some look obviously, you know, if it bleeds, it leads is one model of what news is and unfortunately they will make it bleed like they'll often cause a conflict to get generate the reporting and the content like set fire to something sell tickets to the blaze. There's a completely different view of you know what the 8 billion people in the world are doing where not in a naive polyiana way like oh why don't we report the good news of like you know giving a cookie to somebody that's fine I like that too don't get me wrong but report the stuff which is let's call it investable okay the progress in batteries the progress in solar the progress in nuclear the progress also in let's call it the resilience economy okay with this crazy hormos thing everybody's going to be reallocating capital from wants to needs. Okay. So, how you know oil is not simply upstream of transportation but all kinds of things from plastics. It's a it's a hydrocarbon that's an input to many many different things down the supply chain right from medicines all types of stuff right all of that you know actually Eric did you know I have a masters in chemical engineering from Stanford
count so I actually have a hard science background right um and in fact that was actually my initial my initial thing right so I want to have citizen journalism decentralized reporting on hard science tech across the world, right? And commission that and then pay for that and then have that basically be something where we have a certain number of those per day and they're just posting on Firecaster and so on and so forth. If you're an engineer um or you're if you want to be like an editor-inchief or something like that and you want to work with me on this then reply to me or DM me um with with uh what you can do like basically non AI or rather if it's AI it's disclosed AI right you should be like a great writer so you can verify you know the submissions that are coming in um and or you should be uh like a great coder or something like that you should be interested in essentially taking these prototypes that I put on screen and turning them into something which is free, verifiable, open-source, decentralized, international media for the world. And look, again, I love X. X is great. There's a lot of great stuff about X, but as I as I said, if we go back to my original post here, um you know the thing, uh we have an iron serious problem.
We have an inserious problem. Yeah.
Yeah. We have an serious problem. So the thing is we have to understand that this zombie of NT is getting back up off the mat that they are getting traction again. Um that they are uh they are going to especially with this you know war in my view it's [snorts] given them a sword. They're going to be attacking all of us. Uh they're going to be yelling at tech guys. They're going to be calling us all kinds of names whether you were for the war or not. Right. And uh so post this on screen this thing that I just posted. So you can come back to this post
So the issue we'll need to address with cryptography and AI is some people have started linking to legacy media news sites simply because the URL has a built-in form of validation in it and they can't tell what's true on a social media that's just optimized for views. Right? So that's why I said the sequel to go direct is prove correct. We must prove correct, not simply go direct. Yes, put out your own opinion, but prove the facts. Okay? And how do we prove the facts? Cryptography, mathematics. That's a property of all human beings, not some New York media corporation. Okay? People don't have to trust the tech zillionaires. They don't have to trust us because they don't have to trust us. They can just look at it on chain, right? And so we turn crypto and cryptography just like there's crypto information cryptocurrency. We call this crypto information, right? Information that's built in verification and validation. Okay. Now, by the way, you know, tutorials are actually a pretty good place to start on this. You know why? For you when you go through tutorial, you actually verify it line by line in a way that you don't for most content, right? Because you're like typing that into, you know, an editor or you're you're trying to replicate each step. Like even a recipe actually has a built-in verification to it that most content doesn't. It's not like, you know, if you read some article on Egypt, it's not like you're flying out there like Indiana Jones and verifying every line in the story. You know what I mean, right? This is gel man amnesia, you know, um like do you guys you guys know what that is? Maybe you can put that on screen for Do you want to explain?
Yeah. basically this idea that you read something in the paper, you think it's deeply [__] and you're like, "Wait, I I know about this topic. They're totally wrong about this." But then when you think about when you read other things they write about uh that you don't have expertise in, you sort of give them benefit of doubt and assume they're they're right about that.
That's exactly right. Um and don't go to Wikipedia because Wikipedia is itself a terrible source for everything. Only Grockipedia baby. We're groipedia only in [laughter] this household.
It's wictionary. So I I hope they make a grotionary because wictionary is actually pretty good.
Yeah, I know. But we need we definitely need uh we definitely need grotenary, right? So here we go. Hold on. Um Elon, if you're listening to this, um grotionary um honestly, by the way, Elon, it's actually absolutely insane that like Groipedia is an is a worldclass thing. And for many people, like it's literally it's better than we could. For many people, they'd find it hard to remember it in the top 10 of things from rockets to cars to Tesla, you know, to to to Boring Company and so on. By the way, Boring Company, you know why that's actually way more important than people think?
Iran has actually got a form of defense where they've actually put all of their bases, they built these missile cities, these subways, basically giant subways that are filled with missiles. And so they've got underground bases. And so for both offense and defense, we're going to see a lot more in the way of underground cities. So lots of cities are going to put more and more of their stuff underground because standoff missiles have kind of changed the military, you know, and actually I think New York City got its first approval. I can't believe they got an approval for this for like a 80s story underground skyscraper kind of thing. And so in a sense, the underground world is like the encrypted world. All this satellite footage can't see what's happening underground because it's underground and um there's a lot of actually space under the earth and that direction for things underground and in the water and maybe under the oceans a whole with with modern engineering we might be able to do a lot more than people people think right so the boring company is actually potentially a much much much bigger thing than people even realize beyond just tunnels for cars and self-driving cars it might tunnels for cities and for homes and so on in the medium to long term. Also, you know the term air rights? Have you heard that term?
When when you're buying property, right? Property has, you know, cadesters. Yeah. So, it's like boundaries like the the latitude longitude coordinates in XY space that define what you're what you're actually buying. If you're buying property in cities, you also have air rights which add a Z-axis because you're going up in the air and you might build a skyscraper. And can you build upward in this way? Because if you know are you casting shadows on people nearby and taking away you know so now you have let's call it subterranean rights which are um ground rights how low can you drill into the ground without you know obviously you can destabilize things near you and whatnot so there's engineering considerations that's going to become a big thing anyway coming back to the stack so Elon groia sorry we got gelman amnesia we got that on screen
I hope Elon is uh watching this by the way because are we the biggest live on X right
Yeah, we might. You can at mention. Yeah, there we are. I think Alex Jones. So, how do we check? Is there like a trending tab live?
Maybe. So, but well, let's not let's let's let's entertain the audience. So, if Yes. All right. So, actually, can you look at the thing I just sent you? The not that one. The uh I put in the the Zoom chat pbs. Twimage.
Yeah, there we go. All right. So, let's just read this out. The G amnesia effect is as follows. You open the newspaper to an article on subject. You know why? Well, in in Murray Gelman is a physics noble laureate. So in this is Michael Kryen talking the late Michael Kryen, late great Michael Ken. In Mary's case, physics in Michael Ken's case, show business. You read the article, see the journalist has absolutely no understanding of the facts with issues. Often the article is so wrong it actually presents the story backward reversing cause and effect. Uh I call the cause called call these the wet streets cause rains stories. Papers fold them. In any case, read with exasperation or amusement multiple errors and turn the page to national international affairs. Right? So then you know like uh you know you can hit X and close out of this. Right? So the point is basically um this is something where uh whatever you if if you actually model it right NYT was a centralized hub and how would you know something about Japan in the 1980s or the 1990s or the early 2000s? NT would have a reporter in Japan and they would present it to you because you'd have subscription to NT and how do you know about what's going on in in Turkey or in uh in in the nuclear industry. They would basically be a central hub and they would essentially intermediate everybody's perception of each other like a hall of mirrors in the center. The Japanese guy would only know what was going on in Turkey through one of these centralized news agencies. Right? So you think of it, they call it the media in part because it mediates your experience of reality. It's like a shimmery mirror, right? We're all looking into the shimmery hall of mirrors. And of course, that power of controlling that centralized hall of mirrors where everybody perceives everybody else through this smoky looking glass is this enormous power. This is why they were so so so so you know, insistent. They fought this bruising and ultimately losing battle to try to use their centralized power to censor everybody and to stop people in particular from courting with each other. Do you remember all the freakouts, Eric, about clubhouse and you know like
