The studio itself becomes software.
Metropolis is building a venture studio as an operating system: a solo founder leads specialized agents, business capabilities, knowledge, and governed workflows across a future portfolio of independently valuable companies.
When I don’t know something, I build a system to figure it out. When I want something to scale, I build a system for that too.
Metropolis began while I was exploring ideas for AI-native startups. The ideas differed, but each required much of the same machinery: specialized agents and infrastructure for researching markets, making plans, building products, and operating companies. I didn’t know which individual startup was worth building, so I built the system that could find out—and built it to run more than one company, because doing this once was never the point.
Before Metropolis, I spent more than five years building Amazon FSx at AWS. For three of those years, I led a team I founded to automate the service’s expansion across AWS Regions, Availability Zones, and cells—work that reinforced my conviction that even extraordinarily difficult human workflows can be understood, encoded, and automated.
AI has moved the boundary of what can be automated. I believe much of a modern company’s durable value lives in its data and schemas, its workflow definitions, and the knowledge its team has accumulated by operating it. Metropolis is being built to make those assets explicit, executable, and able to improve through use.
Company-building capabilities should not be recreated from zero.
Venture studios began by sharing human expertise, services, capital, and playbooks. AI-native studios give smaller teams more leverage. Metropolis takes the next step: implement the functions of the studio itself as software, so a founder can lead more of an institution rather than merely complete more tasks.
A central team supplies expertise, services, capital, and operating support to many new companies.
Agents and automation change how quickly a human studio team can build and support AI-native products.
Agent-executed business capabilities become reusable parts, initially directed by one founder.
Company A
distinct market and expertise
Company B
distinct market and expertise
Company C
distinct market and expertise
The Metropolis operating system
Agents, governed workflows, knowledge, and evaluation—plus the business systems a company needs but should not rebuild: customer records and lifecycle, product events and analytics, payments and billing, identity, and model training.
Company A, B, and C are illustrative. Everything below the line is built once and inherited; each company's market, product, and private context stays its own.
An evolution of the platform-VC idea, for the AI generation. a16z argued in 2011 that software would eat the world, then built an in-house platform of experts to give its portfolio an advantage its capital alone could not. Metropolis takes both ideas a step further: software that composes the company itself, and a platform of experts implemented as agents operating on knowledge that compounds with use. Read the full argument →
A studio that builds its own companies—and a consulting practice that builds for others.
The same machinery serves both. The studio is where it is proven; consulting is where it meets companies we did not build, which is the fastest way to find out where it genuinely holds up.
Identify where AI changes what is worth building, then plan, build, and operate companies using shared agents, infrastructure, knowledge, and governed workflows. The first portfolio company is in build now.
Engagements with companies we did not found: re-express a function that will not scale as governed agent workflows on the platform, or have an AI-native system built directly. Open now.
Metropolis is built by Metropolis.
Engineering is the first working team: agents plan, build, review, and deliver Metropolis’s own software under founder governance. The shared business systems are being proven against the first portfolio company now, and consulting takes the same machinery to companies Metropolis did not build.
Go deeper—or bring a perspective we need.
Follow the hierarchy, workflow, knowledge, and founder-decision systems from strategic direction to reviewed delivery.
AI transformation and scalability on the platform, or a custom AI-native build. Engagements are open now.
Read the private thesis and, as they are completed, deeper materials on the capability, knowledge, operating, and evidence systems.
Challenge the thesis, introduce a domain, or explore future leadership when a real company earns it.
What capability or market should Metropolis understand?
Share a perspective, problem, or domain worth understanding. It becomes an input to our research.