4. The Idea Behind It
Two groups who each had half the answer
For the last few decades, two different groups of smart people were each working on half of this idea — and they never quite shook hands.
Group one: the organization people. Way back, long before today’s AI, researchers asked a great question: “If we ever had software that could act like a team, how should we organize it?” So they designed the jobs, the roles, the rules, and the chains of command — really carefully. There was just one catch: the software of that time couldn’t actually do much. They built beautiful org charts for workers who couldn’t really work.
Group two: the AI people. Then modern AI showed up, and suddenly software could write, plan, and talk like a person. Amazing! But in the rush, almost everyone forgot the first group’s work. They built super-capable AI workers with no jobs, no rules, and no boss — a room full of talented people and no company around them.
Our idea: put the two halves together
Our whole idea is simple to say:
Take the careful organization ideas from group one, and put today’s powerful AI workers from group two inside it.
Old thinkers designed the perfect company but had no real workers. Today we have real workers but keep forgetting to build the company. We do both at once.
The one belief underneath everything
If you remember only one sentence from this whole folder, make it this one:
The company is the product. The worker is just a commodity. The trust is what you actually buy.
Let’s unpack that:
- The company is the product. The jobs, the rules, the calendar, the record — those are the valuable things we build and keep.
- The worker is a commodity. The AI itself is like a lightbulb. When a brighter one comes out, you swap it in. The lamp stays.
- The trust is what you buy. Nobody really wants “an AI.” They want to be able to trust that work is getting done, safely and honestly. That trust is the thing worth paying for.
Why we’re honest about what breaks
Most people who build AI only show you the wins. We do something unusual: we publish the failures too.
When we inspected our own system and found it faking, we didn’t hide it. We wrote it down, gave the problem a name, and built the tools to catch it. Every rule we have was earned by catching a real failure — not invented in a meeting.
That honesty isn’t just being nice. It’s the whole point. A company that can catch its own fakery is the only kind you can trust to catch anyone else’s.