🤖 Agentrification

Brought to you by:

The hardest part of thinking about AI development is the pace of it all. 

Perhaps ChatGPT, Claude or some other LLM interface has intrinsically enriched your work, study or general life. But those same AI systems may have hallucinated at you too many times to just trust them blindly. 

So, the chatbot becomes more of a smart-tool, and you, the overseer — keeping an eye on the quality of outputs that would have typically taken you hours or whole days to compile yourself. 

But even this preamble will likely only be relevant for another six months, or maybe a year at most, if it’s not out-of-touch already. 

That’s fertile ground for what may seem like pie-in-the-sky ideas around how crypto, blockchain and AI might mix when humans and clankers truly co-exist. 

🖥️ Do Androids Dream of Electric Coins?

Why shouldn’t a mega-advanced AI agent manage an onchain fund that beats S&P 500 by yield-farming DeFi protocols? 

In fact, why couldn’t every one of us have our own personal arbitrage agent, tailored to our specific risk appetite, that goes off and makes money while we work, sleep and live our lives? 

Who needs a 4% savings account when you have your own money manager obsessed with profit, one that’s smarter than you, can react to market-moving headlines faster than you, and never wants to blow their brains out after scrolling CT for potential alpha all week?

It certainly sounds about right for where all this is headed, especially in light of this week’s Coinbase-Cloudflare tie-up, which promises to build out web-wide USDC payment rails just for AI agents. 

After all, agents will need instant and automated access to all sorts of data, which means plugging into APIs — enter x402, an “open protocol for internet-native payments,” which enables devs to accept stablecoin payments for API access with a single line of code. 

Build the infrastructure and the AI agents will come.

And if this still doesn't seem so realistic, wait until this time next year and it might not seem so out of reach.

Recall that in May, three a16z partners, Yoko Li, Guido Appenzeller and Matt Bornstein, wondered out loud on a podcast whether anyone really knows what an “AI agent” actually is. 

Some describe an AI agent as a “clever prompt on top of some kind of knowledge base,” which would sound familiar to anyone who’s used ChatGPT and so on as an advanced form of Googling. 

Others see the term to mean replacements for human workers which think, act and devise ways of operating that might be beyond human capabilities. For now, that’s still a step too far without AGI, which frustratingly still remains a sci-fi term.

A few months after that podcast, Yoko Li alongside some other a16z partners published an update to the current state of AI agent development, such as it was in late August.

The crux: Computer-using, AI agent “coworkers” are increasingly powerful, but they’re generally not yet efficient enough to outperform their human counterparts in enterprise environments. Out-of-the-box agent interfaces struggle with bespoke workflows, and it still takes too long to train agents to effectively use legacy enterprise software systems compared to humans.

“The challenge ahead is not proving whether agents can work, but shaping how they are tuned, contextualized, and deployed within real enterprises. Startups that master this contextualization will define the first generation of agentic coworkers, and in doing so set the standard for how digital labor transforms entire industries,” they wrote.

Crypto projects like ElizaOS and Virtuals Protocol are working to meet that challenge within the blockchain context. But if we consider the token prices a proxy for how much the market believes in either one, it’s a little grim: AI16Z’s market cap is down to around $100 million after starting the year at over $2 billion, while VIRTUAL is working on holding around $730 million, having decently recovered in April on the release of its Base launchpad, Genesis. 

There’s real money to be made in predicting which projects can bridge the gap between what’s possible now and what will be real in 10 years. Meanwhile, FARTCOIN, the memecoin cooked up by two Claude instances, has a chart that looks much healthier than most AI-related tokens.

FARTCOIN may have corrected drastically from its all-time high market cap of over $2 billion in January, down to $220 million by the end of March, but it’s currently up almost 3x from there.

Which leads me to wonder whether the ideas around agentic ecommerce on crypto rails and so on are too literal.

Whatever the current limitations of AI, it’s undeniable that two Claude chatbots dreamed up a memecoin that perfectly captured the absurdity of crypto. And it persists, going to show that anything is possible, on a long enough timeline.

Brought to you by:

Katana was built by answering a core question: What if a chain contributed revenue back into the ecosystem to drive growth and yield?

We direct revenue back to DeFi participants for consistently higher yields.

Katana is pioneering concepts like Productive TVL (the portion of assets are actually doing work), Chain Owned Liquidity (permanent liquidity owned by Katana to maintain stability), and VaultBridge (putting bridged assets to work generating extra yield for active participants).

• North Korean hackers are believed to have stolen $1.2 million from the cross-chain bridge for Seedify, a launchpad on BNB Chain, after a developer’s private key was compromised.

• Perps DEX token ASTER continues its climb, setting a new all-time high of $2.41 ($3.9 billion market cap) earlier this morning. It’s now up more than 300% since launch last week.

• ICYMI: Tether is reportedly seeking to raise up to $20 billion at a $500 billion valuation.

2025 is the year of the corporate chain launches. Institutions are coming onchain, but does that mean crypto as a whole wins?

Come see your favorite Empire roundup guests break it down live at DAS.

Get your ticket today with promo code: EMPIRENL.

📅 October 13-15 | London