Agents Are the Commodity. The Network Is the Moat.

Ronnie Reynolds · 2026-02-24 · 5 min read

The race to build the personal agent is over. Most companies just don't know it yet.

Open-source agents run on consumer hardware. A $500 Mac mini. No platform. No cloud dependency. No lock-in. Hundreds of thousands of people are already using agents that don't just answer questions — they browse, decide, and buy.

Apple is still iterating toward this. Microsoft is still building toward it. The open-source community shipped it first. On Apple's own hardware. Without their permission.

The personal agent isn't a product anymore. It's a commodity. Just like web browsers commoditized in the 2000s. Just like mobile operating systems commoditized after Android. The technology that every major company thought would be their next great lock-in is already available to everyone, for free.

Within 18 months, every device you own will have a personal agent that acts on your behalf. Apple will ship it. Microsoft will ship it. Google will ship it. OpenAI, Anthropic, and every frontier model builder will ship it. Open-source projects will keep shipping it faster than all of them. The agent layer is a utility. The Billion-Agent Question

When every person has a personal agent, the first thing they'll ask it to do is buy something. "Order more coffee." "Find me running shoes under $150." "Get a birthday gift for my wife." Shopping is the most natural, highest-frequency use case for an agent that acts.

The shift from "I search for X" to "I need X and it happens" is already underway. It will be normal consumer behavior within a year.

Which raises the question nobody in commerce is asking loudly enough:

When billions of agents try to buy something, where do they go? The Network, Not the Agent

Every one of those agents needs to hit a store when it wants to make a purchase. Search a catalog. Compare products. Add to a cart. Complete a checkout.

And right now, 77% of stores are invisible to them. No protocol support. No MCP server. No structured endpoint. The agent can't find them. It buys from someone else. The merchant never even knows a customer was looking.

The gap isn't on the agent side — that's solved, commoditized, shipping everywhere. The gap is on the merchant side. The stores agents can actually reach.

Visa didn't need to convince consumers to use credit cards — the banks handled that. Visa built the merchant network. The infrastructure that every terminal connected to so that when someone swiped, it worked. The more terminals on the network, the more useful every card became. The more cards in circulation, the more every terminal needed to be connected.

The agents are the cards. They're being distributed to every person on earth, for free.

The merchant network is the moat. And it doesn't exist yet — not at scale, not for the millions of small and mid-sized businesses that make up the backbone of commerce. The Pattern Is Familiar

This has happened before. Every major technology shift follows the same arc:

1990s: The operating system was the platform. Microsoft owned it. Then the browser commoditized access to the internet, and the value moved up the stack — to search, to web services, to platforms built on top of the browser. Microsoft's "own the gateway" strategy collapsed.

2010s: Mobile was the platform. Apple and Google owned it. Then apps commoditized and the value moved to cloud services, to APIs, to infrastructure. The platform mattered less than what ran on it.

2026: The AI agent is the platform. Everyone is building it. And it's already commoditizing — open source, consumer hardware, no lock-in. The value will move to the networks agents connect to. The infrastructure that makes agent commerce actually work.

The generator doesn't matter when electricity is everywhere. The grid does. What We Built

I've been in this industry for fifty years. Every platform shift has the same shape: the new technology arrives, the interface changes, but the infrastructure underneath is what determines who participates and who gets left behind.

Four months ago, I started building the commerce network for agents that didn't exist in the market yet. A Federation Gateway that any AI agent — using any protocol — can connect to and shop any store on the network. One connection. Every merchant. Standardized tools for search, catalog, cart, checkout, orders, inventory. The plumbing that makes agentic commerce work.

Five AI commerce protocols. 56 microservices. A multi-tenant gateway designed so that when agents arrive at scale, the stores are ready.

Then, this month, the agents arrived. The Window

The agents are coming. They'll be free, everywhere, on every device, within 18 months. Every major technology company on earth is racing to put one in every person's hands.

When those agents try to buy something — and they will, constantly, because shopping is the most natural thing an agent does — they'll connect to the stores they can find. The ones on the network. The ones with protocol support. The ones that are visible.

The stores that aren't connected won't get a second chance. The agent won't come back and check later. It buys from whoever is reachable and moves on. The merchant never knows what they lost.

Network effects don't wait. The merchants who connect first build permanent structural advantage — more agent relationships, more data, more visibility. The merchants who connect later compete for whatever's left.

This is the moment. Not next quarter. Not next year. Right now. The infrastructure exists. The agents are arriving. The only question is whether your store is on the network when they get here.