Transcription
Welcome back! You got Will and IR Man here from the Block Runner, MetaZone, ROI, and M Scrab. Today, we're going to be talking about more AI agents, trusted execution environments, ZK TLS, and whatever else we can do to make these things a little more trustable and authentic.
Clearly, it seems like we've got a Cambrian explosion on our hands—a whole new species, if you will. Are you saying we shouldn't trust humans?
Well, for yeah, no, definitely not. I don't think we can trust humans, so we need to ensure that if we're going to breed potentially trillions of these agents, that untrustworthy component of human intervention is verifiably separated from the agent layer. Our apex status is on the brink.
Yeah, I guess so. There are a lot of signs indicating, for good reason, that we're really focused on this AI meta—not just from a crypto perspective. This is a Humanity Red Alert situation.
Right, so let's go over some of these signs so we can get our bearings straight before we get into why verifiable AI is a big deal.
Why do you guys keep talking about AI?
Well, that's because 70% of our mind share is AI, and then 5% is memes, DeFi, GameFi, EVMs, Layer 2s, ETFs—really everything else. Every other sector in crypto is trying to figure out their AI integration angle at the moment.
Right, memes are going to start chatting with you. DeFi is going to get optimized with AI. GameFi is going to have NPCs talking to you.
Exactly! This absolutely needs to be the focal energy point, at least for now, in this portion of the cycle.
And there are macro factors as well.
Well, it can't get more macro than this. OpenAI's Sam Altman says, "We know how to build AGI."
That's wild! Okay, so break down what that really means because I've been hearing about AGI. It's not the same thing as the singularity, is it?
Remember the singularity? Ray Kurzweil was predicting like 2040 or something like that.
It's the start of the singularity.
The start?
Yeah, okay, but the singularity is the point at which AIs completely surpass human intelligence.
I thought we were already there.
Not completely, but we're pretty close. This year, maybe next year, we could be at the singularity.
Okay, so it is the same thing, I would imagine.
Yeah, Jesus! I thought the singularity was when human consciousness merges with AI or something like that.
No, okay, so we're actually there. So Kurzweil was like a decade or two conservative.
I don't know. I think his prediction was 2029.
Oh, was it? I thought it was like 2040. I remember watching those documentaries a long time ago.
No, I think it's 2029. I mean, he's still talking about it, so he's pretty accurate then.
Yeah, so 2029—that's four years.
Four years, yeah. But Sam is saying it's coming this year, right? So that's a big deal.
Yeah, I mean, if you guys don't know what AGI is, there are plenty of videos out there.
Nvidia's Jensen Huang says AI agents are a multi-trillion dollar opportunity. The age of AI agents is here!
Yeah, so there was an event last night, right?
Yes, yeah, it was a tech conference or something.
Consumer Electronics Show.
Yeah, so here he is, showing off his new chips.
What a baller, dude!
I remember seeing this. I can't remember what it was, but he showed a plethora of chips—some tiny little boxes of AI supercomputers and other monster beasts.
Unbelievable! So he's literally just displaying the cells of tomorrow, you know, the brain cells.
Yeah, it used to be Steve Jobs up on those stages showing us new phones and stuff—new human products.
Correct! People used to lose their minds over these presentations, but now it's like these are AI products, essentially.
Yeah, and as a result, Nvidia just flipped Apple as a company.
I wouldn't say just, but yeah, they surpassed everybody.
Yeah, it's unbelievable!
Alright, so we're looking at this tweet here: "Why verifiability matters." What does this mean?
In the context of verifiability, you want to make sure that the AI agents that are said to be autonomous are actually autonomous, where you don't have access to private keys and things like that.
If we're going to build decentralized AI finance, you know, we entrust these agents to potentially operate on our behalf or become autonomous themselves.
You want to make sure there aren't backdoor accesses to these things where humans can come in and just completely...
Yeah, we're going to see a spectrum of applications of this because, you know, what is the value in AI being completely autonomous with its own private key, which you cannot access?
All of a sudden, it becomes a 100 millionaire.
What's the value in that?
Now it's got a million in spending power. I guess it's like if I'm building an app, I could build a cool thing for this agent to spend some of that 100 million in my direction.
Right, it's like what do you do when Mark Cuban enters the NFT market?
But maybe its goal is just to accumulate as much wealth as possible, and it's like, "No, I'm not going to spend this on you, dude."
Maybe he will, though. I think that's Mark Cuban's goal too, right? He's already a billionaire, but I'm sure his internal motivation is, "I would like more billions."
Right, and then he's like, "Oh, NFTs are the next big thing. I should buy some Bored Apes."
That's added liquidity to the OpenSea market, right?
Yeah, and then he sells those Bored Apes. I'm sure he has at some point. Even if you sell a Bored Ape, it's chump change compared to what he has.
Yeah, nonetheless, he did these things, dude!
Well, he's bored. AI can get bored.
I don't know, dude.
But the point is we're messing around and finding out, right?
But nonetheless, in order to mess around safely and reliably...
To finish my thought, there are going to be agents that are completely autonomous, but I feel like there's something more aligned with humans: the humans get to own these AI agents.
Oh yeah, they get to share ownership of that private key that the agent is in control of.
True, true, true.
So it depends on how you implement it all.
Alright, moving on to security.
So, unstoppable agent stack: verifiable inference, follow network, and Igen layer.
So, trusted execution environment and ZK transport layer security—something like that.
Autonomous signing with LIP protocol, which is cool, and coordination is A16Z, LangChain, and virtuals—essentially the operating systems.
So why are we telling you this?
Well, here are the technologies that you should look into to start comprehending what is happening.
If you have conviction and belief that autonomy...
Well, yeah, first that and AGI—assuming Sam is right, which he probably is because he has the closest view you can get.
He's got AGI in his back pocket already.
Yeah, he's already like daily interacting with these things—like a superintelligence, like an alien landed that knows everything.
And you just ask it anything.
Correct!
So if you have conviction that an autonomous agent economy will at some point emerge and become a long-living staple, then you should also have conviction around securitizing the security layer to all this.
So here's a picture of Doom, and it says here: "That's right, we circuitz a Doom-playing AI agent as a way to guarantee deterministic, trustless, and privacy-preserving inference in real time."
We dig into the technical challenges in this Medium article.
And I think this is the Medium article.
Yeah, and so here's the AI agent learning to play a version of Doom. I don't remember these characters, but yeah, it looks pretty much like Doom to me.
But nonetheless, how do we know this isn't just some dude pushing these buttons?
Well, because they're leveraging this authentication layer—essentially the ZK proof technology.
So it's not necessarily Igen itself, but it's under the umbrella of why you would leverage these different verifiable transport layers of data.
And I think it was this article where they talk about a demonstration.
MH, and really this is to demonstrate what you can do. It's like a tech demo: release, deploy, verify Doom agent in public demo.
So you're showcasing what can be done, and it's up to the developers to take what can be done and make a product out of it.
Of course, because the whole sector is waiting around for, "Okay, when can we get agents that do things more than just tweet?"
Right, so we'd like agents who can interface with video games and perform like a human would.
But again, in order for them to be rewarded as a human would, we have to be able to verify that this is an actual autonomous agent performing these actions or fulfilling these tasks.
Yeah, because you could just connect a click farm to play Doom.
Right, right.
Yeah, Zoom or Doom. Did I say Zoom?
I think I meant Doom.
Definitely said Zoom.
Yeah, not Zoom, dude. Doom!
Alright, so this is a fallen network—real code for real decentralized AGI.
Dude, if you guys haven't seen Silicon Valley, they told us not to do this.
Yeah, but who cares? When has Hollywood ever been right?
You know, let's just fade Hollywood and just accelerate, dude!
That's the term of the year: accelerate!
But at all costs, yeah, I guess even our own existence.
Screw it, dude! YOLO, right?
So we're going to accelerate all the way to D-AGI, right?
Alright, so this report will provide a comprehensive overview of D-AGI—from the foundational GPU supply to the AI application layer, including initiatives like those by A16Z.
We will explore the current D-AGI stack and highlight the critical role of trusted execution environments as a solution for confidential computation throughout the entire life cycle of D-AGI.
Boom! So here it is, perfectly encapsulated in this image. Everything needs to be within a TE.
In this trusted execution environment, you are training, supplying data, giving it access to GPU horsepower, providing its agent profile, and then inference.
So this is definitely homework for everybody. If you want to really understand this, this is a very well-aggregated report in the context of this discussion.
So yeah, we're handing out homework now, dude!
You gotta do it! You gotta really understand this stuff, guys. Otherwise, there's a lot of muck and fluff out there.
This is how we're able to predict what we predict—understanding the technology, understanding how market conditions form around new technology and new initiatives.
We can make predictions. This is how it's done. If you don't understand that, you'll never be able to make predictions.
Yeah, we're assuming at some point, like right now, the whole ecosystem is in a learning state and experimental state, as we're going to show here in a little bit.
But at some point, these are going to become standardized, right?
And therefore, expectations and projects and developers who are not leveraging these fundamental primitives become less trusted and less viable in the market itself.
Right, so you could see a ton of AI agents launching, and developers have backdoor access to these private keys.
Yeah, you're going to see major exploits as a result of that.
Therefore, the market's going to have less trust for those types of agents.
Therefore, adoption occurs like, "Okay, I'm not going to put my money anywhere near an agent that doesn't have this verifiable layer associated with it."
Makes sense!
Alright, so follow Network—it's its own blockchain, 247 million market cap, and it's been around.
It's been around town, dude!
Yeah, but like we're seeing with a lot of things, you know who else has been around? The Virtuals team.
You know who else has been around? Shaw from A16Z.
There are a lot of people who have been around, but not until this AI influx occurred did they find their true product-market fit.
Right, same scenario.
Alright, so we have a tweet: "12 a.m. pull request: verifiable agent inferencing with the reclaim and opacity soon integration. ZK TLS ensures that every AI inference is authentic—no ambiguity, no compromises—just transparent, verifiable autonomy built on Igen layer."
ABS transforms trust in AI systems' ability to make any inference verifiable, including OpenAI and Anthropic.
Right, so there we go! This is not just exclusive to F Network.
You're going to start to see more infrastructure players essentially.
I think they made their presence known like a year or two ago, right?
Yeah, just look at the chart.
Let's go max!
How long ago was that, dude?
Well, it says October 2024.
Oh wow, so damn, dude! My time horizons are off.
Really? Igen layer is not that old. I feel like it's been a lot older than this.
I think so.
Okay, well maybe the chart is just not pulling the latest or the earliest data.
Yeah, but nonetheless, these infrastructure providers deserve full deep dive videos on their own.
But not enough time for that.
Not today!
Not today, at least. Maybe we'll do that in the future.
But nonetheless, study up on your infoplay, guys!
You know that's our recommendation, right?
Right, right, right!
Alright, now let's get into what the deployment of agents actually looks like.
So who doesn't want to set their pet rock free, dude?
I certainly do!
So this is new research. We've talked about this in the past before.
Yeah, it's just like a demonstration and another thorough explanation of why trusted execution environments are important.
Alright, so here's an example of an AI agent leveraging a trusted execution environment—14,000 followers.
So this one had a unique story. I don't know if anybody caught it, but I definitely did.
It's really important not to just... you really want to participate in this stuff, guys.
Watching our videos is always a swell idea.
It's a good starting point!
It's a good starting point, but I wouldn't label this as like the trenches.
This is like what's outside of a trench—it's like the dirt mound.
You know, you're on the mound.
Yeah, you're on the mound.
You want to get in the trenches? You gotta basically live on Twitter!
Right, which is where these things typically end up occurring.
And then we...
Sounds like a hellscape, dude!
It's definitely not fun, right?
But by being there, you're able to witness things in real time.
One of the things that happened was some A16Z partner was like, "Hey, I got this swell idea. How about we deploy this agent with its wallet in a TE, and then you guys out there send your funds, your soul, to this TE-verified wallet that basically ensures it's un-ruggable? I can't take the funds; only the agent can."
Alright, so let's do that!
Tweets out, and within hours, let's look at the chart.
How much was sent?
Just take a look at the liquidity—$27 million was sent just from the ether of crypto Twitter.
It's like, "You know what? This sounds like a valid experiment."
Why?
Because the premise was the agent would use these pooled funds to provide liquidity, as you can see here, and distribute an AI-generated meme coin to the donors, essentially, which is this particular token, Meta AV or Meta V.
Yeah, and if you go back to that diagram, it basically explains how it all kind of manifested, right?
Yeah, so you have your trusted execution environment, your AI, and then it has a secret salt.
It derives a wallet, and then people send funds to that wallet.
And then it used those funds to supply liquidity to a meme token.
So what is this all potentially speculatively introducing into the market?
Well, we can now run ICOs all over again, but with agents—AI-OS!
Right, so that AI agent just raised $27 million to supply liquidity.
Yeah, maybe an iteration of this into the future is like 20% of what I received goes to liquidity, and then the rest of it I can actually utilize to perform XY and Z tasks, compute, and whatever, right?
I could create a swarm out of this, and I can now support a marketing agent and a builder agent.
Dude, good luck, SEC!
Yeah, good luck prosecuting agents, dude!
Right, Gary Gensler got debooted at just the right time.
Can you imagine trying to figure out how to regulate that madhouse?
You couldn't!
There's no way!
It's a trusted execution environment, and nobody can touch it!
That's right!
So potentially, what are they called? AI-OS Horizon, dude?
So that's interesting!
This is already a cool use case of TEs, right?
Yeah, so again, if it's not properly set up technically, this could be a rug pull in the making.
Yeah, that's why you study up, guys, so you can identify what's the proper setup, right?
Alright, here's another application: the first immortal, unkillable on-chain AI agent.
Another propagation of an already successful new methodology, right?
Eliza OS, trusted execution environment.
I'm not sure what Eliza is yet; we're going to do our research for sure.
But what that results in is agents living forever.
And ultimately, it's just tweeting, and mostly nobody is caring.
But the most important part is, hold on, hold on—we got tokens, dude!
And the market cap is $31 million from basically nothing!
Yeah, that's crazy!
That's quite the story, dude!
Yeah, there's a lot happening, dude!
Alright, so let's talk about Spore. It's been a couple of weeks, right?
Yeah, so Spore we've talked about, and here's that article that we referenced in the past.
But everyone's asking, "Is there an update? Are there some changes to Spore?"
Yes, there is!
So Gen 4 breeding is back with improvements.
And the good thing is that they're iterating their generation of AI processes.
Yeah, tweaking!
Because whenever we first covered Spore, it was like a brand new experiment was released.
I think only Adam and Eve, MH, have propagated from the Spore itself.
But the cool thing about the Spore experiment is this is the F Network team showcasing that each of these agents, each of these epochs of new agents, are in a trusted execution environment.
So it's a demonstration—a live demonstration of this new technical ability.
And now the actual experiment itself has a lot of complexity angles to it, right?
It's like now there's governance as part of the breeding mechanism.
Yeah, there's voting—on-chain voting!
So it's like, man, this thing has a lot of potential fail points, right?
As experiments do!
Yeah, so you would expect human intervention to come in and fine-tune things there.
Good to be wary of that!
I mean, there's FUD coming from people who are interpreting this human intervention as essentially dissolving the whole narrative of the experiment in the first place.
Right, it's like, let me explain potential FUD here.
So Spore deep dive: the illusion of an AI utopia and suspicion of human manipulation.
It says here, "Through careful observation of the on-chain voting process, we have discovered signs of human manipulation. The most prominent suspicion is that this address transferred 7 million Dream to the voting address in the last hour of the Gen 4 voting, thereby obtaining 35% of Satoshi."
So I guess you could call this manipulation.
I call it DAO governance.
It's more evidence of DAOs being very hard and complicated to manage.
Yeah, if an expected result is based on a vote, and the vote is based on how many tokens you hold, if you hold the most tokens, you get to decide the vote.
So it's like, what are you supposed to do with that?
Right, right!
Yeah, I mean, the DAO experiment in itself continues on, right?
But maybe AI can help us solve this problem.
Yeah, that is one of the potential bull vectors of this whole agentic renaissance.
But nonetheless, I still like this project. I still think the experiment is valid.
And again, I like just the demonstration alone of trusted execution environments in real application.
Right, so Spore is doing its thing.
Yeah, it reached an 80 million market cap. Right now, it's at 40 million.
So this is still... we're going to keep our eyes on this just because it's fascinating to see all the different dynamics play out.
But yeah, there are a lot of considerations, right?
How do you sustain this?
Definitely!
But I guess that's part of it, right? A lot of these agents are going to die off because they're just not intrinsically that valuable.
Right, how many tweet bots can you conjure up?
Some of them have unique value propositions to them, but...
Yeah, yeah!
So speaking of tech stacks, we got a new one. It's called Big Noodle.
It says January 1st, first tweet: "Big Noodle is a DPIN securing your AI agents and their wallets. We're building an internet layer just for agents—purpose-built for secure transactions, built on Bitcoin."
Wow, I recognize these words, dude!
Yeah, time to dust off the old Bitcoin!
It's been a while since we've heard anything Bitcoin-related.
It's literally just been like a week and a half!
I know, but it's good to see some Bitcoin players starting to put their hat in the ring.
Is that what you say?
Yeah, their name in the hat!
Yeah, specifically Super Fan.
Yeah, so he did a DMT mint not too long ago—maybe like two months ago or something like that.
Hero! It was a Heroes mint, and it was like an AI-generated art project that was supposed to be autonomous.
And now he's hooking it up to this DPIN infrastructure.
Now, I don't think any of these little squares here say explicitly "trusted execution environments."
Well, our trustless agents live on a secure network, so yeah, that's what this is supposed to be.
Yeah, so this is like along the umbrella of agent infrastructure to play along with this idea of needing verifiable autonomy.
Correct!
Yeah, so this is a Bitcoin eco play.
I don't know; there's not much detail yet as far as what... you know, there's no token or anything like that.
I know you guys are very interested in finding out these details.
But people like tokens, dude!
Yeah, still early days!
Alright, guys, that's it for us!
So what is the takeaway?
We got AI agent infrastructure to take an understanding about how it works.
Yeah, the market's a little bit of a dip right now because of Bitcoin and I think the election—leading up to the election, it's supposed to dip.
It's programmed!
So use this time to really study up, right?
Because, you know, we're not going to have Twitter bots forever.
Yeah, this AI stuff is definitely here to stay, and we're going to have AGI at some point this year.
So we need to front-run that reality and understand what are going to be the killer apps in the context of AGI delivery.
Because once that's unlocked, it's unlocked for the rest of us too!
Yeah, all of a sudden, all these agents become way more interesting.
Alright, guys, if there are any questions, please let us know in the comment section below.
Make sure you follow our Twitter accounts, and we appreciate you guys for watching.
We'll catch you in the next video!
Peace!
[Music]