How I Run 5 Companies with Zero Hired Employees
The architecture, build order, and operating system behind GameTape, Sena, Precis, TrueStandard, and Gavel. No prompts. Just the maps.
A YC partner published a video describing how a company should be built with AI from the ground up. It described my company better than I could. The one line that matters most, she walked right past: "give the model as much context as you would give an employee." She never said how.
That "how" is this document. I run a multi-product portfolio company, GameTape, Sena, Precis, TrueStandard, and Gavel, with zero hired employees. Not zero people: a few cofounders and a stack of agents. Zero W2 staff, zero recurring contractors. This is the actual operating system, the maps and not the prompts.
Nothing here requires a frontier breakthrough. The models are already good enough. What follows is what to build, in what order, and why most of it replaces work that used to need a team.
5
products run solo
0
hired employees
~40
posts/week the stack drafts
52
playbooks built, none by hand
Why Is the Bottleneck Context, Not the Model?
The models are smart enough. The bottleneck is the company-specific context locked in your senior people's heads. Whoever cracks that extraction unlocks the rest.
Pick any frontier model. For almost any business problem you have, it is no longer the constraint. The constraint is that it shows up on day one knowing everything on the public internet and nothing about your company.
Think about what a senior person knows. Not the process doc. The reasons. Why you fired that customer. Why the obvious pricing change blew up last time. The five-second judgment your best operator makes that they could not explain if you paid them to. None of it is written down. It lives in heads. That is the bottleneck.
The first time I pointed an agent at one of my businesses, I gave it the docs and the dashboards and watched it make confident, plausible, wrong calls. It was not dumb. It was new. A new hire would have asked. The agent just guessed. The fix was never a better model. It was getting the context out of my head and into a system the agent could query.
AI
The next teardown, without checking back
About one a week from the five companies I run with co-founders, AI agents, and zero hired employees — what I changed, and the part that broke.
One click to unsubscribe.
What Does It Mean to Run a Company as a Closed Loop?
Open loops lose information. Closed loops compound. Every important action should produce an artifact the system learns from.
An open loop is the old way: decide, act, and let the lesson die in someone's head or a Slack thread nobody reads again. A closed loop captures the action as an artifact, feeds it back, and improves the next run.
Concretely, in my stack: every agent run is logged. Every draft gets checked by a second model against a list the first model never sees, so bad output gets killed before I see it. A nightly job reads its own transcripts and rewrites the skills. My content stack drafts about 40 posts a week and kills the weak ones on its own. I built 52 playbooks without writing a single one by hand.
The surprising part: the highest-leverage component is not any single agent. It is the loop that makes the agents better while I sleep. Almost no mid-market company has a running self-improving loop. It is the cheapest competitive moat available right now.
What Are the 5 Operating Primitives?
These five are the difference between bolting AI onto old software and running the company on agents that wrap deterministic tools.
A single context layer
The important data denormalized into one place an agent can query, not scattered across six SaaS tools. An agent cannot reason across a junk drawer. Neither can a new hire.
An internal tool registry
The reusable tools and skills the agents call. One tool per job, parameterized. When the toolset is clean, the agent picks the right one; when it is a mess, it fumbles.
Skills kept DRY and MECE
One job, one tool, no ten overlapping versions. Mutually exclusive, collectively exhaustive. This is what keeps the system legible as it grows.
A self-improving loop
The nightly process that reads the transcripts and rewrites the skills. This is the part that compounds. The system is measurably better this month than last, and I did not touch most of it.
Transparency by default
Every agent run is logged and visible. A new cofounder ramps by reading how the best operator actually used the system, not by sitting through onboarding. The floor rises.
How Is Each Company Actually Built?
For each product, the question is the same: what is the one function that would normally need an employee, and which agent owns it?
GameTape
Replaced a coach.
How: Ambient coaching layer for founders: ambient observation turned into daily debriefs and longitudinal patterns.
Sena
Replaced an ops coordinator.
How: Portfolio company, separate brand: intake, routing, and follow-up over chat instead of a custom app.
Precis
Replaced a research team.
How: Longevity research engine: a pipeline that scores cross-expert consensus across thousands of sources.
TrueStandard
Replaced a fact-checker.
How: AI fact-verification: a multi-model council that flags fabricated claims before publish.
Gavel
Replaced a strategist.
How: Cited answers for founders: cited, framework-grounded answers in place of generic advice.
The architecture categories repeat: intent capture, a matching or scoring engine, a generation step, a verification step, and a data-enrichment pipeline feeding the context layer. The categories are public. The prompts are not.
What Patterns Hold Across All 5?
Five things are true at every company. They are also the order to build in.
- Data layer first, always. Agents on top of clean workflows, never instead of them.
- Customer interactions over chat, email, and web, not bespoke apps you have to maintain.
- Daily monitoring plus weekly evals replace manual QA loops.
- Token-maxing, not headcount-maxing. Run an uncomfortably high API bill, because it replaces far more expensive headcount. My model bill is the entire ops team for five companies.
- One cofounder in the loop on the big calls. Humans at the edge guiding the system, not routing information through it.
What Would This Find in Your Company?
If one person runs five companies with zero hired employees, your company is carrying weight a stack of agents could lift. The audit finds where.
The hard part for an existing company is not the technology. It is that the most valuable context lives in the heads of people too busy to write it down and too senior to be asked. The company that should win the AI transition is sitting on the asset and cannot get it out of the building.
That gap compounds every quarter you wait, because the cost of running these workflows keeps falling. The work that costs real money in tokens today is cheap in two years. The window to leapfrog is open now. An audit maps where your company overstaffs work agents could do, and in what order to build your own AI Operating System.
Frequently Asked Questions
Do you have zero employees?
Zero hired employees: no W2 staff, no recurring contractors. Founder-led, with a few cofounders and a stack of agents. Not one person doing everything, one person plus agents replacing whole teams.
Will you share the actual prompts and code?
No, by design. You get the architecture, the layers, the build order, the agent categories. The prompts, configs, and code are the moat.
Is this only for startups?
Both startups and existing companies. A startup designs AI-native from day one. An existing company has the more valuable context but it is locked in heads and tangled in legacy process. Extracting it is the work the audit does.
What is the AI Operating System?
A system that captures your data and your team's expertise, then runs your workflows, so the business stops being bottlenecked by any one person. The Operator Method is the how; your AI Operating System is what you get.
How do I get this for my company?
Start with an audit: 10 to 14 days, a fixed deliverable mapping where you overstaff work agents could do, and in what order to build it. Apply at agrahri.com.
Want this built for your company?
I take a few audits a month. The application is a short form. The audit maps where your company is carrying weight a stack of agents could lift, and the order to build your AI Operating System.
Apply for an audit