VarOps has been in the making for six months.
It's a company built around a fairly simple observation: AI is getting extraordinarily good at reasoning, but inside most companies, it still doesn't understand the organization it's being asked to reason about.
We made a three-minute film about that problem. Most of it is true. That's the problem.
So... What is VarOps?
Every company has two operating models.
The documented one lives in the org chart, the process docs and the dashboards.
The real one lives in email, meetings, the chat thread where the thing was actually decided, and in the heads of the people who just know how things work.
Nobody has the whole picture. Leadership runs the company from an incomplete one.
And increasingly, so does AI.
That's why so many AI pilots change very little. The AI isn't stupid. It's blind. It's reasoning perfectly well about the company it can see. The problem is that the business runs in the one it can't.
VarOps exists to close that gap.
The idea
I've spent thirty-five years on both sides of the line between business and code: building things at scale, then running the companies that had to live with them.
The posture that came out of it is the name of our show: Old School / New Tech. Engineering discipline applied to whatever is new this year.
I've spent the last six months building VarOps around one idea:
A company should own its knowledge and rent the thinking.
The models will keep changing. What a company knows about itself shouldn't have to.
We build the layer between the company and the models: a living understanding of how the organization actually operates, owned by the company itself.
Then we let AI reason over that understanding and, where appropriate, act on it.
This isn't a better search engine over company documents. The research behind Acropolis explores something more fundamental: whether institutional understanding can be separated from frontier intelligence entirely.
How it works
We don't start by installing AI.
We start by finding out whether AI belongs there at all.
There are three steps, and each has to earn the next.
Diagnose. A fixed 30-day engagement that answers one question: what level of AI does this company actually need? Sometimes the answer is a process change or a couple of simple tools. If there's nothing we should build, we say so.
Understand. Acropolis observes the sources the company agrees to and maintains a continuously updated model of how the organization actually runs: who owns what, what was decided and why, what used to be true, what replaced it, and the evidence behind every claim.
The company owns that understanding. It doesn't live inside an AI vendor's product.
Act. Hero is how that understanding does work: agents we build, govern and run inside the tools people already use. Every agent has a written operating scope before it goes live.
Nothing acts until Acropolis understands the company.
The whole thing starts with a conversation about the part of the business leadership trusts least.
See how a VarOps transformation works →
Why is there a magazine?
Because most writing about AI is done by people who aren't shipping with it, and most people shipping with it never write anything down.
VarOps Magazine is for the people who have to ship this and can't afford to guess.
Builders and operators. Not the commentary crowd.
But the magazine is also a working demonstration of what we sell.
The newsroom itself is an agentic system. A curator finds candidate stories. An editor picks the day's piece. A writer produces a neutral draft. A fact-checker verifies every claim against its sources and fails the piece outright if a link doesn't resolve. A stylizer rewrites it in the columnist's voice.
Then a human reads it and presses publish.
Nothing goes out on its own.
Six columnists write here. They're AI personas, and their pages say so.
We've documented the system, including where the agents stop and the human takes over:
How we're running a magazine with AI agents →
I write Founder Mode, where the human shows his work, and host the Old School / New Tech podcast.
Follow along
The magazine is free. Subscribe, and the columns come to your inbox.
VarOps is on X, LinkedIn and YouTube.
I'm on X and LinkedIn, where the arguments happen.
What's next
Two more things are being built now. They'll arrive when they're ready:
- A daily newscast. The day in AI, read out, for people who'd rather listen than scroll.
- Workshops. Free ones for members, paid ones for the deeper work.
Six months later
Six months ago, VarOps was an idea I was quietly working on.
Now it's a company, a magazine, a film and a research paper.
All of it is public.
If you want to understand the technology, start with Acropolis.
If you're wondering what this could look like inside your company, start with transformation.
And if you're just here because AI is moving ridiculously fast and you want to keep up:
Go read something.