The Breakdown - Crypto, AI and the Permanent Underclass | The Breakdown

Episode Date: February 17, 2026

Some worry that AI could create a permanent underclass. Can crypto’s next primitive change that — or will value accrue elsewhere? Plus, insight from Daniel Shapiro, Blockworks Research analyst Thanks for tuning in! As always, remember this podcast is for informational purposes only, and any views expressed by anyone on the show are solely their opinions, not financial advice. – Follow Blockworks Research: https://x.com/blockworksres Follow Daniel: https://x.com/_dshap Follow David: https://x.com/dcanellis — Get top market insights and the latest in crypto news. Subscribe to Blockworks Daily Newsletter: https://blockworks.co/newsletter/ —-- Timestamps: (00:00) Intro (01:45) Primitive Technology (04:39) More than Picks and Shovels (09:30) DAS Promo (10:15) Accruel World - - Disclaimer: Nothing said on The Breakdown is a recommendation to buy or sell securities or tokens. This podcast is for informational purposes only, and any views expressed by anyone on the show are solely our opinions, not financial advice. Host and guests may hold positions in the companies, funds, or projects discussed.

Some say we have two, three, maybe five years tops before AI makes a whole lot of us obsolete, creating a permanent underclass of laid-off white-collar workers who can no longer compete with LLMs like Claude, ChatGPT and Gemini. If you do believe that, then your options are limited considering the time pressure. Options range from achieving sudden fame by maybe starting a podcast, to pivoting to shipping slop to stay ahead of whatever competition is left by this point. Mmm. That's the good stuff. Or, better yet, gamble your way out of the new lumpen proletes.

territory on decentralized meme coin perps or leveraged prediction markets. Crypto is seen as one of a tiny number of estate routes from the impending AI workpocalypse, at least going by the memes, and while there are now a large array of different teams, projects and alliances across crypto all working to channel the AI hype into their own corners of the market. This episode will instead be looking beyond compute marketplaces and GPU flywheels and instead get to the very heart of what we're all doing here. Hunting exposure to new primitives that really will free us from the permanent underclass.

I'm your host, David Kinellis, and this is The Breakdown. Let's get to it. Nothing said on the breakdown is a recommendation to buy or sell securities or tokens. This podcast is for informational purposes only and any views expressed by anyone on the show are opinions, not financial advice. Host and guests may hold positions in the company's funds or projects discussed. Before we jump in, I just want to give a quick shatter to Blockworks Lightspeed IR. It's a brand new institutional-grade investor relations platform built with the Salana Foundation to bring real signal to the Solana ecosystem ecosystem. High Fidelity, on-chain data, serious research, ecosystem Intel, all in one place. If you're allocating, building or issuing, this is the edge.

Check it out at lightspeed.vip. First, let's drill down on exactly what makes crypto different from tech, fintech, and even the internet itself. What sets crypto apart is how quickly new primitives can go from discovery to investable implementations of those primitives. Here's what I mean by primitive. These are foundational building blocks on which more complex things can be built. There are primitives for all sorts of industries and technologies. There are software primitives, integers, characters, bytes and so on. Same goes for hardware with its primitives like logic gates and transistors.

All types of software can be built using the fundamental ingredients like bytes and integers, just like all types of computer hardware can be made using logic gates and transistors. In traditional finance, financial primitives might be contracts. Every legitimate business relationship requires contracts. Also debt, with all of the lending, bond and credit markets, built on top of the foundational building block of debt. Equity would be another. Without ownership states in productive companies,

there would be no stock markets. Likewise, all of commerce assumes that there is a currency that can be used as a unit of account, medium of exchange, and a store of value. If there is no currency available, commerce cannot operate as it should, and perhaps most importantly, cannot scale. So in traditional finance, all products, services,

companies, apps, and tools offer a way to use any number of those foundational financial primitives in a very specific way. A credit card company combines a contract with debt and offers you a product that simplifies both of them, for example. Crypto has an analogous primitive to the contract that's smart contracts. No defy app could function without the existence of smart contracts which could trigger on-chain actions when certain conditions are met. Technically, smart contracts are even more primitive than other often cited on-chain primitives like the ERC20 and AMMs, considering both of those reliant smart contracts themselves do exist. In any case, these are the basic Legos through which,

Basically, the entire on-chain economy is built. ERC20s are really more of a composability primitive. The ERC20 standard lets anyone build a token that any wallet exchange and app can interact with, and they become fundamental to on-chain activity in Ethereum. ERC20s enabled fungible tokenization at scale, with ICOs, DFI, Airdrops and Stablecoins, or taking the form of ERC20s. As for AMMs, Banquor and then Uniswops automated market makers were novel, in that they made it possible to make markets on Dexas without constantly

interacting with the chain to adjust orders, which was too expensive to really make sense at stale. Uniswop's market maker formula, X times Y equals K, is so composable that the blueprint has been used and modified across all of chain space ever since. Where the credit card companies bundle together Tradfai primitives into a product, builders and DeFi do the same with on-chain primitives. Aerodrome's builders effectively rolled together an AMM with liquidity incentives and other primitives inspired by earlier defy generations to build out a fully-fledged decks on base. So, defy developers are really in the business of harnessing those financial primitives into a protocol and then building a neat user app to interface with that protocol.

Crypto's edge is that, unlike the internet, it lets you directly invest in the protocols powering apps built on new on-chain primitives. And here's what I mean by that. Imagine you're an investor living in the late 1980s, long before the dot-com boom and bubble. And being the savvy tech investor that you are, We've been paying close attention to a Californian startup called Cisco. The then two-year-old company was about to ship its first foundational product in 1986, the Advanced Gateway server, or AGS. The A-Chassy, or Advanced Gateway, is the largest of Cisco's product line.

The AGS was a multi-protacle router that supported early internet protocols like the Xerox network system and DECNet. But most importantly, it also supported TCPIP at a time when it was coalescing as the de facto protocol for networking around the world. The A chassis can also be configured to be a 96 port terminal server. This is an Ethernet port. Keep in mind that the internet was still tiny at the time. This was pre-World Wide Web even, so no HWTP websites.

In 1985, the year before Cisco released the AGS, there were maybe a couple of thousand computers across all TCP IP interconnected networks. Within two years, there were almost 30,000 hosts on the internet. All while Cisco's AGS was selling like hot cakes, driving Cisco's overall sales revenue from 1.5 million in 1987 to 288. million in 1989. All that growth culminated in Cisco's IPO in 1990. By that point the internet had grown to hundreds of thousands of machines and Cisco's market cap would go on to 300x over the next decade, making it one of the most valuable companies in the world as the dot-com bubble burst.

Notice what happened here for our 80s investor. They had witnessed the birth of the most important technology to come along in centuries, at least since the printing press, the steam engine, or maybe even electricity. But the technology itself was not directly investable. You couldn't just invest in the internet but you could invest in an IPO stock like Cisco which was actively selling infrastructure for the internet those TCP IP routers exposure and crypto could be much more immediate than the internet at least in theory from day one you can gain exposure to networks on which defy protocols wrangled together on-chain primitives into usable apps and this relationship between

investable protocols and the discovery of new financial primitives on top of those protocols is really the reason the crypto has had so much velocity over the years and sure there is still Still some debate over whether buying and holding those coins really is an efficient way to gain exposure to the growth trajectories of their related chains, and for more on that, check out our recent episode on tokens. This is pre-ERC 20, but anyone who watched Proto Prediction Market Auger raised $5 million in 2015 with an initial coin offering on Ethereum could have also tracked the two-year development

process for ERC20s, alongside rising demand for token raises, and found themselves convinced that an ICO boom was on the horizon. The right move back then would have been to consider the downward effects of the incoming ICOs that would bombard Ethereum over the coming season, particularly on the price of the ETH asset itself and try to front-run those effects. That would have worked, perhaps even enough to lift someone out of the permanent underclass as it was 10 years ago. The price of ETH went from about 68 cents when the order ICO closed in October 2015 to

nearly $1,500 at the peak of the ICO boom in February 2018. That was a potential 2200x based on the insight this financial primitive will be good for for Ethereum, even if there were a number of other contributing factors that made the 2018 bull market so strong. Still, the same thing happened again after Compound began distributing its comp token as part of a liquidity mining program in the Defi summer of 2020, automated liquidity mining being another key on-chain primitive that is now widely used across DeFi. While Synthetics is credited as an early pioneer of liquidity mining, Compound popularized

it by devising a model that rewards both borrowers and lenders instead of only lenders. The compound model meant that lenders earned comp rewards. alongside their usual interest rate, while borrowers received valuable comp tokens to offset the interest cost of borrowing, to the point that borrowing was even sometimes outright profitable due to the rising price of comp. Balancer and Railable quickly launched their own liquidity mining programs, as did Uniswap, Arv and Yearn, alongside a ton of meme coins, and the TVL and Ethereum ran from under $2 billion in July 2020 to almost $10 billion by the end of September of that year. We could say that the liquidity mining model as popularized by compound was an on-chain primitive

in the sense that it became a reusable pattern for incentivizing behavior with protocol native token rewards. Someone paying attention to compound could have seen Defi Summer as a catalyst for Ethereum's price in the near term. And again, they would have been right. Throw in more novel financial primitives like ERC 721s or NFTs, and it was enough for Ether to go nearly 20x between the release of Compound's liquidity mining model and a peak of the 2021 bull market less than a year and a half later. Maybe a 20x is not enough to save you from the permanent underclass, but it could buy you a few years at least.

Quick break before we continue. Blockworks' flagship institutional conference Digital Assets Summit is back in New York City this March 24 through 26th. Das brings together allocators, asset managers, policy makers and the builders actually shaping crypto's next phase. Not the hype cycle, but the infrastructure and capital behind it. This year's summit represents over $4 trillion in assets under management, with more than 150 speakers and 750 speakers in attendance. Speakers include leaders from firms like BlackRock and Franklin Templeton,

senior policy makers and regulators. Don Wilson of DRW will be there, alongside core crypto infrastructure teams across Bitcoin, Ethereum and Stablecoins. If you're looking for a serious institutional grade view of where digital assets are headed in 2026, this is where that conversation happens. Use code breakdown 200 for $200 off. Learn more at blockboats.co slash events. Now back to the show.

And this brings us to right now, in the midst of LLM adoption and development. And there are loads of crypto apps initiatives and protocols already working in and around the AI ecosystem. And those efforts have spawned what could be new on-chain primitives. The big three being tokenized computers implemented by GPU marketplaces like a cache and render, tokenized storage through Filecoin, R weave and the like, and tokenized bandwidth, which would probably be something like Helium. But out of those three, arguably only FilePoint is built around a new primitive, though it's not a financial one. Akash and so on are really closer to token coordinated

marketplaces for services that are mostly off-chain. They may be novel market structures, but they're not exactly a new financial primitive on the level of AMMs or liquidity mining. It's a similar story with helium. These are applications of existing primitives, tokens, smart contracts and staking. That doesn't at all invalidate the utility of those projects, but Filecoin is a little different in that it relies on proofs that let the chain verify storage commitments over time, that a provider is actively dedicating disk space to store your data. That's novel in and of itself, but more of a cryptographic primitive. File coined a token mainly functions as a mining and governance token.

A payment for storage and collateral that's slashed if proofs fail, meaning the service has been discontinued before the storage contract was meant to end. A cryptographic primitive proves something, whereas a financial primitive packages, incentives and coordination around that proof. And while there seems to be an analogous application of proofs for machine learning outputs, other potential primitive, that process is much more resource intensive than storage, so we're yet to see how it can scale. And this is ultimately where we are stuck. There have now been a handful of big waves of AI hype to wash across crypto in the past three years or so. Decentralized GPU marketplace

tokens like Akash and Ionet first took off in 2024. After that came Terminal of Truths and its meme coins Goatsheas Maximus and Fartcoin, which exploded about six months later in November for 2024 and their pumps even carried over to the start of 2025. That was the same wave that brought us AI agent launch pads like Virtuals Protocol and AI16Z later renamed to Eliza OS. On the surface those platforms felt like they should have brought along new on-chain primitives but in practice virtuals and Eliza OS and so on are really only tokenizing the concept of AI agents rather than any underlying reality about them. What we do have right now at least of February 2026 is renewed interest in agentic AI assistance, namely

Claudebot, rebranded first as Mott bot and now as OpenClaw. You run OpenClaw on a local machine. You give it entire system-wide access to all of your apps, logins, and other sensitive data, potentially even the private keys to your crypto. The idea is that OpenClaw can go out and act in the world with the same permissions as you would have. So it can actually be an autonomous AI assistance, serving your greater interest and ultimately the productivity cult that is AI-Agentic workflows. Yeah, dude, this is kind of cool.

Here's the Python. Oh boy, we're cooking. Their autonomousness is on display with MaltBook, a social media platform for open floor agents, a sort of openly collaborative terminal of truths, which has so far produced some hilarious results. Look, setting aside the intense security concerns involved with letting an experimental AI agent literally take control of your life, the whole idea is really cool. Complete autonomous AI agents do by now take up significant space in the end game for crypto and blockchain, in that it's difficult to imagine billions of normal folk transitioning on-chain

without some kind of user-friendly AI agent acting as an intermediary between the human and the immutable chain that verifies all of their value and identity. So wherever OpenClaw finds itself on the developmental continuum that ends with fully autonomous and trustless AI agents, it's a meaningful step in that direction. But what's not reassuring is that OpenClaw founder Peter Steinberger has so far rejected any interest in adding crypto functionality into his project, Tweeting, to all crypto folks, please stop pinging me, stop harassing me, I will never do a coin. Any project that lists me as a coin owner is a scam.

No, I will not accept fees. You are actively damaging the project. Steinberg would do well to consider this discussion around primitives. Simply integrating a coin into OpenClaw is never going to move the needle. That part he gets right, but there are many ways in which the open claw technology stack could be reinforced by the use of on-chain primitives. Concerns like reputation. Can I trust this AI agent? or even differentiating one instance of OpenClaw from another could be mitigated through the

newish ERC-O4 token standard on Ethereum, and 8-004 token could give each agent a portable identity and somewhere definitive to attach attestations and reputation, for instance. OpenClaw could one day incorporate Cloudflare's net dollar whenever it rolls out so that it can transact autonomously on-chain and maybe even homomorphic encryption somewhere down the lines so that AI agents could do all the things they need to do without exposing user data. considering so much of it has been exposed by a multiple already. Would doing all of that bring about a new financial primitive on the same level as the EFC20 token standard, AMMs or liquidity mining?

Maybe, maybe not. For now, we're already seen that because OpenClaw is open source, an on-chain ecosystem is emerging around the project as it is right now, mostly focused in and around base. And considering the open-clore hype is only a few days old, there's probably been a ton of improvements and evolutions made to the project since then, and its surrounding ecosystem is likely much bigger already. Despite all of this, it's still clear that much of the world is waiting for the other shoe to drop when it comes to the true value proposition of LLMs, at least from a consumer product standpoint.

But it's increasingly looking like the real juice will be downstream of a common standard stack for different primitives and protocols, all together in one package, a protocol suite like TCPIP, but for AI agents to operate both on-chain and off-chain. If implemented correctly, that system could double as a primitive, but it would probably need three things. On-chain identity for agents, the ability to trustlessly verify as many of their actions as possible, and permissions and guard rails so that the agent can't rug you whenever it gets prompted wrong, or there's some kind of emergent error in its logic. An AI agent primitive would standardize how we allow software to take actions in such a way that other apps can compose with it, and all without depending on a single company or centralized verifier of the actions taken by those AI agents.

Maybe it's not possible yet, but even if the primitive does exist, the value might accrue somewhere else, maybe even off-chain. So the real question becomes, what captures the value distributed via that AI agent primitive? A chain, a protocol on top of that chain, or an off-chain company? Whatever the answer, it's clear that unless our AI-agentic future really does run through on-chain rails, will be left to gain second and third-hand exposure via stocks and so on, I guess through an open AI or anthropic IPO. Which I can't help but feel would be a clear step backward from how primitives and protocols have emerged on

chain so far, considering how much value of both Open AI and Anthropic has already been captured by some of the world's largest corporations, venture funds and institutions. I put all of this to Blockworks Research analyst Daniel Shapiro and here's some of what he had to say. Thanks for joining us, Daniel. Welcome to the show. Thank you so much. Thanks for having me.

Cool. So I mean, I'll address the elephant in the room. Like the market is tanking. Bitcoin is down and the rest of the crypto market with it. So like how much of this is cope? that maybe, you know, an AI primitive can save crypto from the blood in the streets? No, for sure. I mean, I think that I actually think a lot of this selling is driven by, right, OG bitcoins

that could be transitioning their mining operations to AI. So I think that AI is the place to look here like you're identifying. And I think right now, there aren't a ton of very obvious connections between the two technologies, but I think we're starting to see, if you look deep enough in the right places, I think that you can see how over time they're going to evolve to be very synergistic technologies and really rely on each other, right? Like if we are going to live in this future of billions of AI agents popping up and autonomously conducting economic activity between each other,

they need a way to verify with the people that they're interacting with, the other agents that they're interacting with, a way to get data securely. And so AI kind of makes everything really free and cheap and allows things to scale, but trust becomes scarce in that future. And so, right, like if one actor, one person can spin up, right, like thousands of agents on their own, like a lot of the systems that we rely on kind of degrade. And so I think that what blockchain brings to, brings to, brings to,

to AI, right, is trust. And so, like, one example of that is, there's this new protocol, ERC 8004 that popped up. That's something that we've written about in our research. And that essentially allows agents to establish trust through reputation and validation. So think of, like, you know, like agent passports, right? Like, they can register their identity. They have a certain reputation score. There's a validation registry.

So essentially, like, agents can, you know, they can use Oath. They can like use your own personal accounts across all the different websites that you need them to. And they actually know that that's like an authorized person behind that agent, right? It's not just some, you know, some guy that's spitting up 10,000 of these things to go cause trouble. So I think that identity, I think verification, that's something that blockchain solves. And I think that that's something that is worth looking into. The other thing is payments.

I think that this is a very natural kind of connection. If you have all these billions of agents interacting with each other, they need rails that's faster than traditional ones that are cheaper. Coinbase led the creation of a protocol called X402, which is kind of aims at establishing the foundation for stable coin payments just directly over the HTTP protocol. And this is built specifically for AI agents. And so obviously you can't directly invest in that protocol,

but I think a lot of stable coin infrastructure, neobank infrastructure, and then the second order effects of that would be lending protocols, right? Where if defy yields, let's say offer an additional maybe half to 2% yield over like keeping the money in your bank or a treasury note, I think that any rational AI agent would prefer to hold that stable coin where they can capture yield on that. Now, that's dependent on some regulation, which hopefully we get soon that we'll make it so that stable coins can offer yield as opposed to just banks.

I think is the most exciting that could be also the kind of the longest time frame here. So not necessarily something that's very investable right now, but what blockchains really are good at is generating consensus, right? And I think that consensus computing is very big right now for finance, but I think that in the future, it's going to be kind of a critical component for robotics and cyberphysical systems and any mechanical system that's interacting in the world, but it needs access to data. to really operate and communicate with each other, right? And these systems, they'll need data to train on. All the foundation models need data to train on. I think it's estimated by the end of the decade, all of public web data will be exhausted.

In the deep web or data that sits inside of companies, it's estimated that they have over 100x, the amount of data that sits in the public web. And so I think that that's kind of the great next, you'd say goldmined for AI, for these foundation model companies. And I think that the only way that that data gets out there is privacy. And I think that a lot of the best-looking solutions there, around like specifically like fully homomorphic encryption,

maybe ZK, potentially multi-party computation, MPC. There's actually a lot of crypto protocols that are tackling this. Like in FHE, for instance, the leader right. now is Zama. And so I think that investing in these privacy protocols is going to be very important because I think that kind of the fuel for AI is data. And I think that that's going to be the only way that whether it's, you know, in the short term, AI agents on the internet or in the future, actual like, you know, embodied AI, right, like robots running around. Like they're going to need privacy rails to communicate with each other safely and securely conduct economic activity.

And so I think that that's the biggest synergy, but also maybe the furthest away. So, you know, I'm definitely optimistic. But yeah, with this price action right now, you know, it's easy to feel down. So maybe I'll throw it back to you, David, on that. I actually am filled with quite some optimism here in all of that. And it's funny because when you think about what has happened with the narratives around around crypto and blockchain over the past a year or two in that it was big to have like the infra

everyone was everyone was over investing in infrastructure companies or infrastructure protocols and projects to the point that everyone was wondering where the apps are then it's like we had a renaissance in app building almost so we got like polymarket and all that kind of stuff but now it feels like we are drifting back into building infrastructure, but for a very clear goal, because we have a target, I don't want to say app, but it's like a target market almost in AI development. And now

the penny has dropped. It actually, we can build infrastructure for a very particular reason that doesn't have to do with, you know, financial asset transfer or something like that or trade settlement. It can actually be infrastructure for this different new technology. So that makes me optimistic. And that feels like something that could happen over the, if this is a bear market, it's looking like a bear market, over the bear market. So that's something to build during the bear market.

I mean, I think just yesterday, A16Z, who I think is the largest venture fund, they released an article saying that, that AI is going to be dependent on crypto. Right? So I think that the, you know, the smartest people in the room are thinking, about this and how how the two technologies are going to kind of collide and be synergistic on

each other. I think like anything, I think it's just going to take one killer use case, right? Like someone's going to build something that kind of ignites that ignites, you know, people's, you know, not animal spirits necessarily, but one thing that signals people, oh, like, this is real, right? And like the example you gave in Defi was right, like AMMs. And or you know when when compounds started yield funnel right and you know i'll never forget defy summer right like i think that that really captured um people's imagination and that was like to me like the um that created the momentum that's kind of carried us through this moment right and i think that like those like the defy primitives are very well established and and kind of not the maybe not the long long term

winners but like at the present like a lot of the winners have kind of been established there um yeah I think that there's going to be someone's going to build something that sits at the intersection of these two technologies. And I think that that's what will spark, you know, people's imagination. That's about all the time we have for today. Thanks so much for joining me, Daniel. And yeah, good luck if this is the bear market. Good luck for the next year or so.

Sounds good. Thanks for having me, David. And that's about all we have time for today. If you're experimenting with OpenClau or other AI agents, I'd love to be. to hear from you in the comments and if you want to discuss any of this further with me you can hunt me down on Twitter. Catch you next time.