One founder.
An entire company.

01 / THE NEW UNIT

A founder with
more ways to act.

A company begins with someone choosing a problem, seeing a possibility, and deciding that it is worth pursuing. Turning that decision into something real takes many kinds of work. The challenge is keeping those kinds of work moving toward the same result.

The unit I’m exploring is a human with direction, supported by a system of intelligences that can take on different parts of the work. A founder can hold the product in mind while agents investigate an idea, develop an interface, work through implementation, and examine what might be wrong.

That changes what I can attempt. A question can become a prototype sooner. A prototype can create new questions. Several experiments can develop at once, while one person keeps the purpose coherent.

The opportunity I care about is the distance between an idea and a useful thing in the world. I want that distance to get shorter without losing the judgment that makes the thing worth building.

02 / THE SWARM

Different jobs.
One shared direction.

For me, a swarm means a coordinated set of agents with defined jobs, shared context, and a way to bring their work back together. An agent needs to know the question it is answering, the material it can rely on, and what a good result would look like.

Specialization makes that coordination easier to reason about. Research establishes what is known. Design explores an experience. Engineering makes it work. Strategy examines the choices ahead. Music develops another creative language, and automation connects repeated work into useful flows. Review runs through the whole process. Those responsibilities can change from project to project.

The real coordination happens in the handoffs. A beautiful design that ignores a constraint, or code that solves the wrong problem, creates more work for everyone. I want the swarm to make context easier to carry forward and disagreements easier to resolve.

Codex and Claude are the current building tools in this experiment. The operating model should keep evolving as I learn which tasks benefit from parallel work and which need a slower, more deliberate pass.

03 / THE HUMAN

The founder’s job
gets sharper.

I still decide what matters. Who is this for? What should exist? What deserves another week of work? Which idea should end? Those are choices about purpose, taste, and consequences.

Agents can bring options, arguments, drafts, and implementations. I need to understand enough of the work to make a decision I can stand behind. More capacity makes that responsibility more important: it becomes easier to produce things before asking whether they should be produced.

The founder also owns the relationship with the outside world. A customer’s trust, a creative promise, a release, and the resources spent getting there remain real. I want AI to expand my ability to act while preserving a clear person responsible for the outcome.

The human holds the vision.
The swarm expands the capacity.
The work has to earn its place.

04 / THE OPERATING LOOP

Build a loop
that learns.

The process I’m working toward is a repeatable loop. Each pass should leave the project with something useful and the next pass with better context.

1. Define the outcome.

   Write down the intended result, the constraints, and how it will be judged. A task needs an end that someone can recognize.

2. Split the work.

   Give agents bounded responsibilities. Identify what can happen in parallel and what depends on an earlier decision.

3. Bring the work together.

   Resolve conflicting assumptions, inspect the pieces in context, and make the decisions that require a coherent whole.

4. Check and release.

   Use the checks appropriate to the work: functioning software, a readable interface, visual continuity, or a complete creative release.

5. Learn and preserve.

   Keep the useful context: what worked, what failed, and what the next attempt should know. Let actual use inform the next brief.

Parallelism is valuable when it moves the project forward. Adding agents can also add handoffs, duplicated work, and review demands. I want to measure the result of the whole loop, including the time spent coordinating it.

05 / THE EXPERIMENTS

The work is where
the thesis gets tested.

The projects on this site explore different parts of the same question. They give the idea something concrete to push against.

The projects have different purposes and different standards for success. Their shared value is what they teach me about carrying a direction across many kinds of work.

06 / THE COMPANY

Capacity is only
part of the company.

Producing more work does not create demand on its own. A company still needs a reason for people to care, a way to reach them, and a product or experience that keeps being useful.

I want the swarm to help me test those questions as seriously as it helps me build. An impressive prototype can be the beginning of learning what people want. It can also reveal that a direction needs to change.

There is a practical cost to this model: tools, inference, infrastructure, integration, review, and the founder’s attention. A useful system should make those costs visible. I care about how much good work comes out of the complete process, rather than how busy the agents appear.

The billion-dollar ambition makes the question bigger. Can a company built this way create lasting value at global scale? That outcome depends on the work meeting the world, and it remains something to pursue and test.

07 / THE TEST

What would make
this thesis true?

I want to judge the experiment by evidence that survives beyond a convincing demo. The questions I want to keep asking are:

Useful outcomes

Does the process consistently produce finished work that someone can use, experience, or build on?

Quality

Can errors be found and corrected? Can the work meet the standards of its own medium?

Cost and attention

Does the output justify the compute, tools, coordination, and human review it takes?

Continuity

Can a project develop over time without repeatedly losing its context or creative identity?

Demand

Do people return, participate, and find value in what is being made?

Those questions make room for failure. Some ideas will ask too much of the current tools. Some will need more human collaboration. Some will turn out to be the wrong projects. I want the record to include what changes my mind.

08 / THE BET

A bigger ambition
for one person.

I’m building toward a company whose reach can grow far beyond the size of its founding team. Art, software, music, characters, and new ways of participating are all part of that ambition.

The billion-dollar company is a possibility I want to investigate through real work. Every release is another chance to learn what the model can carry, where it breaks, and what becomes possible next.

One founder gives the work a point of view. A coordinated swarm gives that founder more ways to make the point of view real. The company will be built in the relationship between that capacity and the people it serves.

How much can one person build?

R.R. / Emrecan Doğramacı
Founder, Internal Agency