AI Doesn’t Have an Operating System Yet

For the past forty years, we’ve built software on operating systems.

Windows. macOS. Linux. iOS.

They all solved the same problem: creating a trusted layer between people, applications, and hardware.

AI has no equivalent.

Today, AI is scattered across chat windows, APIs, cloud providers, and disconnected tools. Every conversation starts over. Every application reinvents the same plumbing. Every company is forced to decide where its data lives and who ultimately controls it.

The industry has focused almost entirely on building better models.

I think that’s solving the wrong problem.

Models will continue to improve. They will become faster, cheaper, and more capable every year. That matters, but it won’t define the next decade.

The next decade will be defined by how AI is operated.

Who owns it.

Where it runs.

How it remembers.

How it is governed.

How it moves between your laptop, your organization, and the infrastructure you control.

That’s the layer I’m interested in.

I believe AI needs an operating system.

Not another agent.

Not another model.

Not another workflow builder.

An operating system.

One that gives people and organizations control over where intelligence runs, what it remembers, which models it uses, how it connects to other systems, and how it evolves over time.

I started building because I couldn’t find that layer.

I wanted AI that could begin on my own computer, extend to more powerful infrastructure when needed, and scale into enterprise environments without forcing people to give up ownership along the way.

I wanted AI that could move with its owner instead of being trapped inside a single vendor’s ecosystem.

I wanted something that treated memory, governance, security, and execution as fundamental parts of the platform—not features added later.

That belief became .

This publication is where I’ll share what I’m learning while building it.

Some posts will be technical.

Some will be about infrastructure.

Some will challenge the assumptions driving today’s AI industry.

Some will simply document what works and what doesn’t as we build in public.

My goal isn’t to convince everyone to agree with me.

It’s to contribute to the conversation about what AI should become before today’s assumptions harden into tomorrow’s defaults.

We’re still in the earliest days of this transition.

The operating system era of AI hasn’t started yet.

I think it’s about to.