"The most important thing is that we actually have things to spend bitcoin on. I remember when Lightning came out - the sentiment was: OK here is my lightning wallet - what can I actually do with it?"
-Andre Neves, Co-Founder & CTO of ZBD
Last week I was at Bitcoin Takeover at Bitcoin Park Austin. Andre spoke on a panel about the intersection of AI, bitcoin and gaming. His words stuck with me.
There has been great discussion about L402, X402 and other AI payment standards brought forth by a lot of smart people in bitcoin.
Agentic payments and the rails they use represent a trillion dollar opportunity.
But Andre's question is the one that is really most important. In a world with marginal differences between agentic payment rails, the one that wins WILL be the one with the best network of applications.
Not which protocol wins. Not which rail is marginally better. Just:
what can you actually do with it?
We’ve seen this already. Lightning shipped. Wallets worked. Payments were fast. And everyone had the same reaction:
cool… now what?
That is still the problem.
Fast forward. We’re heading into a world where agents operate autonomously, call tools, and pay to get things done. That’s not a niche. That’s trillions in future commerce.
The question of apps vs rails is as clear as it is prescient. The rails show up first, and people assume adoption is inevitable. It’s not. Applications decide everything.
So the real fork in the road isn’t technical. It’s simple: do we build things agents actually need to pay for, or do we hand the entire stack to systems that already have distribution?
Bitcoin is up against incumbents that control distribution and default behavior, which means if we don’t align on building real applications, we lose by default.
In this post, I’m making a simple argument:
Bitcoin doesn’t win because it’s better. It wins if Bitcoiners build things for agents that make using it unavoidable.
The good news is it’s never been easier to build. The bad news is no one else is going to do it for us. Let’s look at history.
History Doesn’t Say We WinHistory Doesn’t Say We Win
It says we might... if we build the right things.
There’s a pattern here.
The systems that win aren’t the ones that are most correct. They’re the ones that get used and dominate with overwhelming mind share and adoption.
Email Wars -> Developer Experience Wins Over Better SpecEmail Wars -> Developer Experience Wins Over Better Spec
In the mid 80s to early 90s, X.400 was a better-designed email system. SMTP was simple and easy to build with. SMTP won because developers could actually use it. But that alone wasn’t enough. It won because people built email.
FTP vs HTTP -> Human Emotion Beats Pure FunctionalityFTP vs HTTP -> Human Emotion Beats Pure Functionality
By the 90s, FTP already worked. You could move files. HTTP didn’t add a new capability - it created an experience. Pages, links, browsing. The internet became a place, not just a pipe. That’s when usage exploded. When it appealed to real human emotion and demand.
Browser Wars -> Default IE Distro Dominates Better ProductBrowser Wars -> Default IE Distro Dominates Better Product
Then there’s Netscape. Early, strong, technically solid. It still lost. Distribution beat it.
Same pattern, three angles:
- if builders can’t use it, it doesn’t spread
- if there’s nothing to do, no one cares
- if incumbents control distribution, you still lose
This is exactly where Bitcoin sits.
We risk overbuilding the “correct” standard instead of the usable one. L402 is already drifting in that direction. You don’t need a bloated spec to enable agent payments. The core loop is simple: request an invoice, pay it, verify the preimage. That’s enough to build real things.
But even that’s not the main point.
Lightning already works. Payments are fast. The rails exist. And yet it still feels like FTP.
There is no default experience. No obvious thing to do. No place agents go where paying is just part of getting something done.
And while we figure it out, incumbents aren’t standing still. They already have distribution, integration, and default status. If they ship something that’s easier and “good enough,” most builders will choose it.
If you're reading this and you're vibe coding an app: Bitcoin needs you. Bitcoiners need to make something that agents and users absolutely have to have to get the job done.
A Wallet With Nothing to Buy Is a Bridge To NowhereA Wallet With Nothing to Buy Is a Bridge To Nowhere
No one wakes up wanting to make a payment. They wake up wanting something. The payment is just the unlock. Agents reveal this truth more ruthlessly than any human ever could. They have zero sentiment about rails. They don't care that Lightning is fast or that the protocol is elegant. They care about what the rails get them access to.
So the question was never "can we move money fast?" We can. The question is: what's on the other side worth paying for?
What Will Agents Actually Pay For?What Will Agents Actually Pay For?
Agents are ruthlessly efficient. If something helps them get a job done faster, better, or cheaper, they'll use it. But right now, when an agent can't find what it needs, it hallucinates, stalls, or falls back to a closed API. That's the gap and that's the opportunity.
Not generic APIs. Not primitives. Capabilities:
Context (give me the right information). An agent that can buy the right context on demand doesn't need to be pre-loaded with everything. It gets smarter as it goes, cheaply. Example: "find every time X was discussed and summarize it."
Decision-making (tell me what to do). An agent that can purchase a reliable recommendation skips the reasoning overhead entirely. Example: "which supplier is the best choice given cost, reliability, and delivery time."
Execution (actually do it). An agent that can delegate a task end-to-end and pay for the outcome is more valuable than one that just plans. Example: "book the flight, pay for it, and send me the confirmation."
Picture an agent trying to complete a real job. It needs context it wasn't trained on. It hits a decision it can't resolve alone. It needs something done in the world. At every one of those moments, there is a product waiting to be built and a payment waiting to be made.
Most APIs being built right now are primitives dressed up as products. An agent doesn't want a primitive. It wants something that reliably solves the problem and charges accordingly. That gap is the opportunity. Build for it.
Think I'm wrong? Good. Build it yourself and prove it. And if you've already built something, make it callable. Add an OpenAPI spec. Expose it as a skill. Let agents find it and use it.
Closing Thoughts: The Opportunity of the MomentClosing Thoughts: The Opportunity of the Moment
We are not early on rails. We are early on what you can actually do.
If we get this right, Bitcoin looks like SMTP or HTTP: simple, widely used, driven by real applications. If we get it wrong, it looks like X.400 or Netscape: technically solid, respected, mostly irrelevant.
Right now Lightning feels like FTP. It works. It moves money. But it’s not a place, not an experience, not something you return to.
What we need is the HTTP moment. The browser moment. The thing that makes people say: “this is what I do with my AI and Bitcoin.”
That does not come from better rails. It comes from applications that are useful, repeatable, and hard to ignore.
And this matters even more in a world of AI agents. These systems will decide what tools to use, what services to call, and where money flows. If those flows run through closed, permissioned systems, the entire economy becomes mediated and controlled.
I don’t want that world for my children.
I want them to grow up in a world where they and the agents they rely on can transact freely, build without permission, and operate without constant surveillance. Where commerce is open, private, and global by default.
That future does not happen automatically. It has to be built into the systems agents use.
We do not win by default. We win if we build things people and agents come back to. We win when payments disappear into the experience.
That’s the bar. Build for it.
P.S. - What I'm building
This isn't just an argument I'm making from the sidelines. Jamie is my attempt to build one of these applications. To humans it looks like podcast search. To an agent it's paid context retrieval across millions of podcast moments. Ask a question, get structured answers, pay for it. No fluff, just the thing you needed.
If you want to see what making a product agent-callable looks like in practice, check out how Jamie is listed on Clawhub. That's the kind of network we need more of.