The planning document was three pages long. Justin had written it over a weekend in late January, and it described the second year of the software factory. It was the shortest strategic plan in StrongDM's history, and it was the most ambitious.
More twins. The Digital Twin Universe would expand beyond Okta, Jira, Slack, Google Docs, Drive, and Sheets. Year two would add behavioral clones for GitHub, AWS IAM, Azure AD, PagerDuty, and Datadog. Each new twin extended the surface area of the simulation. Each new twin meant more scenarios could run against more failure modes without touching production. The twin universe would grow until it modeled every service the company depended on.
More scenarios. The scenario library, which had grown from one to over three hundred in year one, would target a thousand by July. Not through volume for its own sake, but through diversity. Adversarial scenarios. Edge-case scenarios. Scenarios that modeled cascading failures across twin boundaries. Scenarios that tested not just whether the software worked, but whether it worked when everything around it was breaking.
More agents. Not more types of agents—the Attractor pipeline already supported Claude, Codex, Gemini, and others through its pluggable codergen backend. More concurrent agents. More parallel work. The factory would scale horizontally, running multiple convergence loops simultaneously on independent work streams.
The factory scales. The team doesn't.
That was the line in the planning document that Jay read three times. Three people. Still three people. The factory had produced, in its first six months, more validated code than a team of twenty could have produced in the same period, and the plan for year two was to produce five times more with the same three people.
"You're not hiring," Jay said. It was half a question.
"We're hiring one person," Justin said. "We posted the listing."
"One person. For a factory that you want to scale by five-x."
"The factory scales through agents, scenarios, and twins. Not through people. People write specifications and scenarios. Three people can write specifications and scenarios for a system of arbitrary complexity. Four people can do it more comfortably, with better coverage and more creative diversity. Five people would introduce coordination overhead. The optimal team size is small."
Navan looked at the planning document. "You've modeled the team size as a variable."
"I've modeled everything as a variable. Team size has diminishing returns past four for our current architecture. If the architecture changes, the model changes. But I don't expect the architecture to change."
Jay sat with that. Three people and a thousand scenarios and a dozen digital twins and a fleet of agents running convergence loops twenty-four hours a day. The factory didn't need more people. The factory needed more intent, more clarity, more diverse descriptions of what the software should do. And that was a function of insight, not headcount.
"Year two," Navan said, and he said it the way someone says the name of a place they haven't been yet but can already see.
Jay closed the document. Justin opened his whiteboard marker. The equation was still there, stable, waiting for the second year to validate what the first year had proven.
"The factory scales through agents, scenarios, and twins. Not through people." Every VP of Engineering needs to read this sentence and sit with the discomfort.