GitHub for the Rest of the Economy

"One of the things we don't have is real-time data. I gotta constantly go in, pull it out, manipulate it."

That was the CEO of a staffing business, ten million a year in revenue, on a discovery call a few weeks ago. Not a startup, not a hobby. A real company, paying real people, generating real revenue. And he described his own visibility into his own business the way someone describes a room they can only see by lighting matches. Every answer was a project. Every number he quoted had a silent question mark behind it.

This is what half of GDP looks like right now.

Before 2008, code lived in places. Your code, my code, their code, somewhere in a folder or on an FTP or on a drive that crashed last Tuesday. You could ship software this way. People did. But the work was lonely, the artifacts were opaque, and everything ended at the boundary of one person's workstation.

GitHub changed the shape of that. Not by replacing the editor, not by telling anyone to switch languages, not by adding features to the IDE. It did something lower. It made code legible, composable, and collaborative. It gave every piece of code a URL, every change a history, every repo a network. And once that layer existed, everything on top of it got built. Pull requests. CI. Dependency graphs. Copilot. The entire developer experience we take for granted in 2026 runs on a substrate that didn't exist before a private company put it there.

Microsoft paid $7.5 billion for GitHub, and it was the best acquisition in their history. The trillion-dollar software industry we have now is mostly sitting on top of that substrate.

So what does that have to do with the staffing CEO?

The first attempt at solving his problem was automation. Zapier, Make, n8n. The idea was reasonable. If your tools don't talk, wire them together. Trigger fires here, action happens there, data flows, problem solved. And for a while, that shape of solution fit the shape of the problem well enough.

It doesn't anymore. Not because wiring is hard, but because wiring is the wrong layer.

Agents broke it. An agent doesn't want a declared flow from point A to point B. An agent wants to read the state of your business, reason about it, and act. You cannot reason about state that lives scattered across fourteen tools in fourteen shapes. You cannot act on data that doesn't agree with itself between the dashboard, the chat, and the report. Automation's job was to move information. That job is mostly done. The next job is to make the information itself legible, and automation doesn't know how to do that.

This is why every automation company you've heard of is quietly trying to rebuild from the bottom up right now. They know. The category is dying.

Automation's job was to move information. That job is mostly done. The next job is to make the information itself legible.

Here's the part I think about constantly. GitHub didn't win because it built a better text editor. It won because it recognized that the unit of progress in software was code, and that code needed to be legible to collaborate on. Once you treat code as a first-class, versioned, linkable, composable thing, the entire industry reorganizes itself.

Business state is the next unit. The ledger, the pipeline, the inventory, the customer history, the open tasks, the closed deals, the conversations, the outcomes. Right now it lives in whatever the vendor decided it should live in, in whatever shape the vendor decided it should have. No two small businesses have the same stack. No stack agrees with itself across tools. You cannot compose on this, you cannot collaborate on this, you cannot even read it without a full-time human ops person translating.

So you normalize it. You give every piece of business state a canonical shape. You give every metric a single source of truth. You give every action a trace. You give every tenant a memory. You make it legible.

And once that layer exists, things get built on top that couldn't exist before. Agents that actually work because they have a real substrate to reason over. Autonomous loops that run hourly across every tenant and surface problems the owner hasn't noticed yet. Cross-tenant learning that makes every business on the network smarter without ever leaking any single business's data. A marketplace of composable skills that any operator can drop in. The flywheel that built the developer economy, running one layer lower, on the half of the economy that got skipped.

That's what we're building. Liaison is GitHub for the rest of the economy. Any tool stack. Normalized business state. Composable agents on top. Autonomous operators running hourly on every tenant, every day. Shared learning across the network, privacy-preserving by design. Not a dashboard you check. An operator that acts.

Ten years from now, the companies worth the most money in software are not going to be the ones with the prettiest UI for a single vertical. They're going to be the ones who own the substrate that every other business tool has to sit on. GitHub sat under the developer economy. Someone is going to sit under the operator economy. The first trillion-dollar outcome built for the half of GDP that never got software is on its way, and it does not look like another SaaS tool.

The rest of the economy deserves its own GitHub moment.