I'd like to start this conversation with a few what me and my co-host Ryan call Balaji-isms. Balaji-isms about kind of where we are in the world, just talking about the tailwinds, the large macro tailwinds, like all kind of setting the stage for technology in the future. And I can give you a few prompts to see where you want to go with this. Right now, the United States is buying companies that are critical for national security, chip companies, precious metals. That's new. China is moving from being a copycat to pushing the frontier. There's a right-word political wave happening in Latin America and a rise of socialism and straight-up communism on the far left in America. Just a few examples. What other big tailwinds do you see that you think will kind of define the rest of this decade and into the 2030s, especially as it relates to AI, privacy and crypto?
Well, I think there's actually, you know, people talk about the singularity, but there's several singularities. In fact, you can make a deck which is just going through like 50, 70. So for example, rates are going through the moon as you know, the sovereign debt crisis is happening, this gigantic important thing. All like meeting and dating in the West is just all internet now. Every other form of meeting people has gone to zero from like meeting in bars, restaurants. This enormous change in like how humans, the next generation is formed. Obviously, you've got AI adoption, right? You've got like, you know, what's happening in war with like it's all turning into drones, right? So human infantry is obsolete. The future's already here. It just hasn't been evenly distributed yet. And so many militaries are actually obsolete and that hasn't been propagated out. You have a hundred unsolved math problems or whatever the number was that was just solved yesterday, right? And you have, you know, like things like the New York Times that don't even report on that because like beyond their ability to even understand what's going on or let alone like package it for an audience. Um, you have...
But it's also like social media becomes now the source of news, right? Like that's where it propagates and it becomes available to everyone at the same time.
Yes, exactly. And so what I think is happening is upstream of all of this is the internet. And what the internet does, it takes things to their logical conclusion very quickly. It's like a binary switch where you're like if A then B, if B then C, and you get to Z in the same day. And the logical conclusion is not always the practically feasible conclusion because there are some frictional steps somewhere around step D, but they go all the way to Z very quickly, exhaust it, it crashes into some obstacle, or maybe it works, and then they're onto the next thing. And so all kinds of things are just going flipping up, flipping down, like binary switches flipping up, flipping down. And we have to develop, I think, a robustness to what I call the post-internet world. You know, you had the pre, you know, the post-war order, right? The post-war order, like 1945 to 2020, 2025, will be conceptualized as the pre-internet order.
And I mean, I would argue this is a post-AGI world. Like it's a world where intelligence is extremely cheap and information propagates instantly, and you effectively like a lot of systems in our world have built on friction, right? I mean, starting from banking, right? The banking rates are effectively a friction system. But this kind of is true across almost everything.
So this is why I think what's going to happen is digital borders are going to become on par with physical borders. And that's why there's only really two groups that really take the internet seriously enough, which are China and crypto, because the great firewall and the blockchain are both defensive structures on the internet, right? And once you think about the digital as actually being on par with and in fact upstream of the physical, then the friction that gets reintroduced is you've got digital border control and the only agents that are able to peek in are those that you have cryptographically authenticated. And actually this is the part where friction I think is going to get reintroduced. What agents are doing now is they're spamming everybody, they're trying scams, they're trying malware, they're trying hacks. And very quickly the drawbridges are going to go up and people will only share information within the trusted tribe. And because with, you know, like within the trusted tribe you share all your code and you can learn a lot. Between trusted tribes you're just, AI spam is just going like this. If you've seen GitHub, it's got so much downtime because anything that is not armored up, the entire concept like the trusted internet is going to go away. All trust in the internet, it becomes untrusted, you know, area. And so it's going to be crypto tribes are going to become, going to have digital borders and that's going to be where what hardness is on the...
You disagree?
I would disagree with the conclusion. I agree with the premise. I think, I mean, as a blockchain, we've actually been building, I mean, kind of permissionless tribes maybe the way to do it. And you know, with NEAR Intent, we've been building a road between these tribes and kind of super highways.
But I think my point here is like I think the blockchain, the crypto thinking applied to AI, applied to social infrastructure as well. I think like one of the points I've talked a lot about is like AI is great to like effectively DDoS all of the legal system, all of medical system. I said it like three years ago. Now we actually see this happening, right? Like all the insurance companies are effectively drowning in a bunch of claims which are legitimate claims that you know, but like a lot of minuscule things. And so you need to apply the civil resistance. You need to apply all of the systems that we've built to create kind of a permissionless, but kind of aligned experience. And I think that's the post-AGI internet and the world we need to build.
That's right. I think to first order, crypto is what AI can't do. Like AI makes everything fake, crypto makes it real again, because right? And so you have, you can fake any image and so and so forth, but you can't fake the solution to a cryptographic equation.
Not yet, right? So yeah, unless P equals NP and we're going to, we may see. One thing was interesting, somebody observed this today, you know, talking about AI safety. In all these results that OpenAI dropped, there weren't any cryptography papers. And people noticed, oh, they did result 44 and then result 46, but result 45 was held back, right? So there may be some zero days on who knows, ECDSA or something like that out there, right?
Well, I mean, I actually have this in my presentation, but some of the lower bounds that were effectively thrown, they kind of have implications for cryptography. Now, they're not practical yet, like we went from, you know, it's going to take more than life of universe to, you know, half of the time of universe, but the fact that they lowered that is effectively where the risks are coming from. Like it's like, well, how much more we can lower it now? And so that's I think opening up. I mean, we've seen, I mean, Justin Drake, right, just opened up this kind of can of worms with like, which right now practically don't do anything, don't screw things up, but in theory there is like a factor like, hey, look, we can actually continue crunching and actually lowering bounds on cryptography as well.
Yeah, one of the things I've been thinking about is there's an interesting figure eight where AI agents will need to buy coins to, you know, for civil resistance, to prove, you know, for example, the insurance companies in a few years may not accept a submission unless you pay a fee, like an application fee, which means they don't get DDoS with things, right? So AI agents will need to buy crypto. But, or so, let's call it, let's make it very symmetric. AI agents will need to buy crypto tokens, but I'm also starting to think that crypto protocols will need to buy AI tokens for the following reason. Every crypto protocol, you are going to need more defense than there is offense against that protocol. And very crudely, your defense budget in tokens should be larger than the attacker budget, so that as soon as a new model is available on OpenRouter or whatever, it's almost like high-frequency trading. Your, you know, your all the nodes that are running, they maybe have a local model running, or they have something that can ping OpenRouter or something like it. They instantly get the latest model. They start prompting it with your codebase 15 different ways trying to hack it with a new model. And you're spending your defense budget on your own protocol faster than the guys can attack it because you care more, right? And I think that's going to be the new thing where you're trying to hack your own thing. You're red teaming your own thing automatically, and you actually allocate some of your coin towards red teaming. And so you have this circular economy where the agents buy the coins to prove they're human, and then the coins buy the agents to stop the hacks.
This is dancing around one of my favorite Balaji-isms, which is AI makes everything fake and crypto makes it real. And I think that that Balaji-ism is actually kind of manifesting in NEAR because NEAR has products like verifiable inference, which is to simply secure the inference. First things first, maybe Illia, you can kind of talk about the role that NEAR can play in some of the world that Balaji just painted for us.
Yeah, I mean, we exactly are this eight. Because we have the inference side where you can, including with NEAR tokens, you can actually have a consistent flow of AI tokens to feed back into your protocols, into your systems, into your agents. And then on the other side, indeed, like you have civil resistance through the NEAR token to effectively prove that you're kind of able to participate in the economy, in the intents, in the markets. And so we're kind of really building this, like the way I usually structure this, like AI is a front end, blockchain, like blockchain markets are the back end, but the reality is, yeah, this kind of a continuous loop that's connected.
There's some fun way of saying it like AI buys crypto tokens to prove they're human, crypto buys AI tokens to show they're not hacked. Something, something like that, right?
We can get a future-ology-ism is being cooked right now.
I'm refining it. I'm, you know, like distilling it. I'll tweet it after.
There is like worth noting formal verification is one way to get out of this arms race with new models.
Yes. It doesn't solve all the problems, but it does...
