The Agent Economy's Missing Foundation
Somewhere between AI and crypto, there's a trillion-dollar problem that neither industry can solve alone.
I've been watching two industries talk past each other for years.
AI teams are building agents that manage portfolios and negotiate with other agents autonomously. Crypto teams are building financial infrastructure that’s increasingly sophisticated but still designed for humans clicking buttons. Almost nobody is working on the overlap. The few who are tend to be so deep in one world that they’re starting from scratch in the other.
I’ve spent years building blockchain infrastructure and Bitcoin products. Some of the most exciting moments of my life happened in this space, so did some of the most expensive lessons. Both matter when you’re trying to figure out what comes next. I’ve watched protocols I trusted get drained overnight, bridged assets across enough chains to know which ones make you hold your breath, and sat through plenty of market cycles to stop confusing noise for signal.
All that time spent watching what breaks and what survives turned out to be exactly the right training for what’s happening now. Autonomous AI agents are moving real money, and the problems they’re running into aren’t new. They’re crypto’s problems, resurfacing in a system where nobody’s even pretending there’s a safety net.
The list of what agents need and don’t have is long, from identity to liability to dispute resolution. But three sit at the foundation of everything else:
Security guarantees that don’t depend on a human checking their phone for a 2FA code at 3am.
Collateral that isn’t controlled by a single company that can freeze it.
Programmable money that settles without a counterparty who might be asleep, bankrupt, or running a different software version.
These are three of the biggest gaps I’ve found, and each one splits into a dozen smaller problems underneath it. The question is what kind of financial infrastructure solves for all three?
It’s not a stablecoin - those have issuers who can freeze your assets with a phone call. Not a centralized exchange - those have business hours, terms of service, and compliance teams that weren’t designed to adjudicate disputes between agents.
What checks every box is Bitcoin.
No human required. No counterparty. No downtime. No single entity with a kill switch. Bitcoin wasn’t built for agents. It was built for a world without trust. Turns out, that’s exactly the world agents need to operate in.
There’s a small number of people starting to figure this out. Mostly they’re quiet about it. The agent economy is still early enough that the serious builders aren’t doing press tours. They’re writing code, running into challenges and solving problems that don’t have names yet.
The problems they’re solving - and the ones they haven’t figured out yet - are going to define how autonomous software interacts with money, identity, and trust for the next few decades. I’ve been researching that intersection for months, and I’m going to lay out what I’ve found across a series of four pieces, starting with the conclusion that surprised me most: Agent security isn’t a tooling gap, it’s an architecture problem.
The push to bring AI agents into decentralized finance - what the industry has started calling DeFAI - has attracted more branding energy than engineering so far, though the underlying work happening beneath the label is real and worth watching closely.
Bitcoin’s role in the agent economy is not about price or store of value - it’s about being the only financial infrastructure that doesn’t need a human in the loop, which turns out to matter a lot more when the users aren’t human either.
Not all of my ideas will age well. Some will prove out exactly as I expect, and others will get overtaken by things nobody sees coming.
The pattern I see right now looks like 2016 Ethereum - a moment where the infrastructure decisions being made by a small group of builders will shape how an entire economy works for the next decade. Last time that happened, most people didn’t notice until it was too late to influence the outcome.
I’ve lived through enough hype cycles to know the difference between a narrative and a shift. Narratives need conferences and Twitter threads to stay alive. Shifts keep going whether anyone is paying attention or not. This one keeps going.
The noise around AI agents is only going to get louder, and I have no interest in adding fuel to that fire. My intention is to map what crypto already learned the hard way onto what the agent economy is about to learn the same way.
I’ll start with security. Not because it’s the most interesting part of the agent economy, but because there’s a flaw in the foundation that nobody has a good answer for yet. Including me. But the more I pull on it, the more I realize it’s not a bug in the system. It’s the system itself.