Hi friends! I've spent the last few months deep in the AI agent x crypto rabbit hole. Rather than waiting to form a complete thesis, I'm sharing my explorations as they unfold. Maybe we'll end up with a thesis eventually.

TL;DR:

• Crypto's latest trend is AI agents, powered by easy-launch tools like Virtuals.

• Most AI agents are just smart social bots that make money from attention.

• Blockchain gives agents built-in payment systems and token economics.

• The big opportunity is autonomous companies, not just crypto applications.

• Today's agent memecoins hide real progress, just like early internet toys.

Crypto loves a good narrative. The space is basically a superconductor for trends, amplified by social media network effects and echo chambers. We saw this with NFTs. We saw it with memecoins. Each wave mints millionaires overnight, sends protocol marketcaps to the moon, and leaves plenty of rekt traders in its wake.

Such is life in the trenches.

The latest fling? AI agents with crypto capabilities. But wtf is actually going on here? Are we just building infinite slop machines and sentient memes, or is there something bigger on the horizon?

Before diving in, let's revisit the most important mental model for exploring the tech frontier. It's so crucial you should make it your morning mantra:

Amara's Law:

"We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run"

And if you're feeling spicy, throw in Dixon's Law:

"The next big thing will start out looking like a toy"

Both are at play here.

The transformative power of blockchain networks needs a long lens. Eventually, all tickets will be NFTs. Culture will become an investable asset class at the micro-trend level. Online transaction costs will compress from today's ~2% credit card fees by orders of magnitude.

Eventually. Not tomorrow.

When a narrative takes hold, the signal-to-noise ratio goes to shit. Crypto Twitter (CT) obsesses over quick gains, short-term fortune hunters flood in, and it all obscures the fundamental progress happening beneath the surface.

You know a crypto narrative has reached escape velocity when everything converges into memecoins. That's exactly where we are with AI agents right now.

What are agents?

Simple distinction: A bot follows rules. An agent thinks for itself. It can understand its environment and take actions to achieve goals.

Bots = stupid. Agents = smart.

We could probably discuss until sunrise about to which degree these agents are autonomous, but let's just leave that for now.

Where we are & How we got here

The crypto + AI intersection isn't new. Infrastructure projects like Render Network, Akash, and Bittensor laid the groundwork. But things kicked into high gear last year when the narrative shifted from backend to frontend with tokenized agents.

It started with Truth Terminal, an autonomous (but supervised) social media agent on X. From there, it was turbo speed through copycat social agents before landing at Fartcoin (yes, really) and more importantly, technological breakthroughs like the Virtuals launchpad.

Virtuals is basically Wordpress for agents – anyone can launch a tokenized agent with basic instructions. The crypto integration makes it unique from web2 alternatives.

It's (Almost) All Memecoins (For Now)

When launching costs nothing and potential rewards are massive, everyone rushes in. We're seeing hundreds of new agents daily. Most are vaporware or glorified social media bots.

But that initial product-market fit makes sense. Humans already compound social media attention into token value.

Take AIXBT – arguably the first crypto influencer on X. It analyzes market trends, shares insights, and engages with its 400K followers. Current marketcap: $700M.

Sure, much of that value comes from memetics and attention. But AIXBT is also fronting a data intelligence terminal that requires 600K tokens (~$500K) to access. Given the recent price appreciation, that's a steep price, but it hints at an interesting pattern:

AIXBT functions as a marketing agent, driving attention by sharing insights. This creates token demand from people wanting full terminal access. In theory, one could add in lower-tier priced products, since a $500K piece of software has a very limited customer segment.

According to sentient.market's AI agent index, this is already a $13.6B category. While it may be tempting to take the easy route and overindex on the memecoin convergence and dismiss it – I don't think you should.

The Opportunity Framework

At CES this month, Nvidia CEO Jensen Huang dropped this bomb: "AI agents represent a multi-trillion dollar opportunity"

Not today. Not tomorrow. Eventually.

He wasn't specifically talking about crypto-capable AI agents. But putting agents onchain makes perfect sense. At minimum, a wallet lets agents efficiently pay and get paid in near-real-time.

Now imagine agents trained for specific tasks, working together, collaborating. The opportunity isn't just about creating automated slop machines with tokens attached.

Dream bigger.

Let's evaluate the crypto component from two vectors:

1. On crypto: Tokenizing the agent itself creates an embedded economic model. The agent delivers services (to other agents or humans), driving token demand. Others can invest in the token as an asset, similar to company stocks.

2. For crypto: Most utility agents are currently crypto-native, handling market analytics, trading, etc. It's logical to first solve problems in their native environment.

Crypto's UX complexity might become irrelevant. An agentic abstraction layer between blockchain operations and users could leapfrog the problem entirely. Projects like Griffain are already showing this. This "agent as defi abstraction" got dubbed "DeFAI" (who tf names these things?).

But the biggest opportunity category might be: On Crypto, For [Everything].

Tokenized agents solve monetization, but the services don't have to be crypto-specific. We could see agent-first companies bootstrapping on crypto rails with capital efficiency, leveraging narrative strength while solving real-world tasks. Think of them as autonomous micro-companies.

I'm still forming these thoughts, so bear with me. Over the next weeks, I'll explore:

- Network effects in the agent economy

- AI agent swarms (multiple agents working together)

- The agentic micro-company concept

- Economics of launchpads like Virtuals

- Agents in entertainment

If any of these threads seem particularly interesting, reply to the email or hit me up on X.

Thanks!