We believe companies should become more immediately capable as they accumulate experience. Model Operator builds the shared memory and AI systems that make that possible.
The company brain is the asset
At Rails World 2026, David Heinemeier Hansson, the creator of Ruby on Rails, announced that 37signals had gone “pencils down” on handwritten code. Robert “Uncle Bob” Martin, whose book Clean Code shaped a generation of programmers, has described letting agents write code and relying on tests and quality constraints to verify it.
Engineer friends at NVIDIA and Spotify tell me they haven’t written code by hand in 2026. Creative production is changing too. Have you not tried ByteDance’s Seedance 2.5 yet? You have serious control over inputs and outputs are increasingly realistic with human appearance and movement. More of the production process can happen through direction and iteration.
We expect this shift to reach every white-collar function. Marketing, finance, sales, operations and management all involve work that becomes more accessible as models improve at reasoning, writing and acting through software. Progress will vary, but companies can already reconsider how much a person or team should be able to accomplish.
The connection to company knowledge is direct. Delegating more execution means making the context behind that work available to whoever or whatever performs it. A capable model still needs to understand what this business is trying to achieve, what it has already learned and which constraints matter.
We see an opportunity for businesses to make more of their accumulated experience available this way. A company brain connects records with the reasoning and outcomes needed to interpret them. An experiment should leave behind enough evidence for another team to build on it. A correction should improve subsequent work.
Preserving the reasoning matters because circumstances change. People need to understand when an earlier decision still applies and when its assumptions deserve another look. Otherwise, memory can make a company more consistent while keeping it consistently wrong.
This knowledge should belong to the company and remain usable when models, vendors or employees change. Each new tool should be able to draw on what the business has already learned.
Make AI multiplayer
A private assistant can learn a great deal about a business without making that understanding available to anyone else. The company pays the cost of explaining itself repeatedly.
Shared agents allow colleagues to develop and question the same body of knowledge. Through Slack, Teams, planning sessions or voice conversations, people can examine evidence together and contribute context the system lacks. Accepted corrections inform subsequent work, with access governed by the underlying permissions.
This gives delegation a firmer basis. People and agents can work towards an agreed outcome without repeatedly reconstructing its background. Their results become evidence the team can use to reconsider the plan.
Over time, that should improve the company’s ability to question its own direction as well as execute it.
Find the ceiling before scaling the work
8 years ago, I read an essay online that changed how I think about growth. It was called Invisible Asymptotes, written by Eugene Wei, who had worked in strategic planning at Amazon. He described how a business can keep improving yet approach a ceiling imposed by the way it operates.
At Amazon, research revealed that shipping fees discouraged purchases. Free shipping on orders above $25 helped, but customers waited until they had enough to qualify. Prime let them pay through a membership, removing the decision about whether delivery was worth paying for each time they ordered.
What struck me was how easily a capable team could spend years working around a constraint it had come to accept. More marketing or a better checkout might improve results while leaving the reason people held back largely untouched.
That possibility matters even more when AI makes execution faster. We can use the additional capacity to investigate assumptions that previously went unexamined.
A company brain supports that investigation by connecting performance with customer evidence and previous experiments. Outside research helps test explanations that internal experience may reinforce. Shared agents can then help turn a promising hypothesis into an experiment, preserve the result and inform the next investment.
The company gets faster at discovering which changes are worth making.
More capacity, more choice
For AI to change the economics of a business, its capabilities need to translate into dependable work. Shared context helps teams coordinate that work and catch misunderstandings before they spread. We start with a consequential workflow and measure the outcome: shorter delivery cycles, better acquisition payback or lower cost to serve.
As the system becomes dependable, people can pursue opportunities that previously exceeded their capacity. More of the roadmap becomes achievable. Growth experiments can run sooner. Investment decisions get the investigation they deserve, and promising partnerships receive attention before the opportunity passes.
Faster feedback makes larger ambitions practical: teams can test their assumptions, learn and commit resources with better evidence.
We want that capacity to give people more choice over their time, too. Some can go towards building the business. Some can go towards family, friendships and interests that deserve deliberate attention. Greater ambition should become compatible with a life people have time to enjoy.