When I started Metropolis, I had a handful of startup ideas and no good way to tell which one was worth building. When I don't know something, I build a system to figure it out. When I do know something, I build a system so I don't repeat myself and can scale it. polis came out of both habits, and Metropolis builds it to plan, build, and launch companies.
In The studio is the first customer, I wrote that Metropolis itself was the first company built on polis: whatever the studio needed got built once, and every company after it would inherit the work. Since then we've realized the studio needs a consultancy. We are pausing the venture studio to focus on Metropolis Consulting, the first company running on polis for other people, until we have explored a market deeply enough to know it is worth entering and can see our position in it. We expect most of the building blocks consulting needs to carry over to the studio and its portfolio companies, so this is a small pivot on the same platform we have been building all along.
A venture studio builds companies around problems, and we can't choose those problems well from the outside. Consulting puts us inside companies that are working through them, which builds the knowledge of markets and problems the studio depends on. It also brings in cash, so the studio can be bootstrapped on revenue instead of on money raised before we know what to build.
The company I already know
I think of a company as four things: its people, what it knows, the work it does, and the resources it creates and manages for customers. polis is modeled around those four. As a solo founder, I wanted a first company where all four were already mine, and for me that is technical leadership.
I spent more than five years at AWS, where I worked on Amazon FSx and founded a team that turned up to 18 months of repeated human work per region into an automated system. I have also built products at startups and mentored founders on building systems that scale. If I weren't building Metropolis, I would probably be a fractional CTO.
So the first company running on polis is Metropolis Consulting, which gives startups technical leadership with a platform behind it. airCFO does this for a startup's finances. We do it for its technology.
Who we work with, and what we learn
We work mostly through networks: venture capital and private equity firms, and the companies they back. A firm sees the same problems across its whole portfolio, which makes it the fastest way to learn which problems are common and which are one-offs.
Each engagement shows us what founders struggle with and which of those struggles keep coming up. The ones that repeat are where Metropolis's future companies will come from. What we carry forward from an engagement is method, meaning how a kind of problem gets solved. A client's code, data, and strategy stay with them, and their problem never becomes our product without them.
Built for us first
Every service Metropolis Consulting offers started as something we needed ourselves.
Planning, building, and operating software is how Metropolis builds polis every day. Agents plan, build, review, and release the software, and a person approves the decisions that matter. I described what that looks like in The difference a year makes. For a client, we turn a problem or an opportunity into a plan, build it inside their codebase, and keep running it if they want us to.
Architecture and security reviews came from the same need. To keep improving polis, we have to understand it, so we map its systems, model the threats, and check whether it is ready for what comes next. We do the same for a client's codebase.
AI spend visibility started with our own bills. Every model call polis makes records the tokens it uses, so we know what our AI work costs and where. Clients should be able to see the same about their own spending.
This matters more as more code gets written by AI. Veracode's 2026 report on AI-generated code found that AI now writes roughly half of the code committed at teams that use it, and that about 44% of AI coding tasks introduced a security flaw. Code got cheap to write, but knowing whether it is safe, what it costs to run, and whether it will hold up is as hard as it ever was. As I argued in AI isn't a genie, AI delivers when there is a system and a person around it.
Start with a tool, then come to us
Tools like Lovable, Base44, and Perplexity are a good way to try an idea quickly and cheaply, and I would tell most founders to start there. You drive them yourself, and they are built for starting new apps on their own stack.
Metropolis Consulting is for later, when a company wants to own and grow what it has built. We work inside the system you already have, in your own repository, and a person is accountable for the result. When we finish, we hand everything over or keep running it for you.
How it feeds back
Metropolis builds polis. Metropolis Consulting runs on it, and so will our venture studio and the companies it builds. Everything they do feeds back in: when an engagement needs something polis can't do yet, we build it once, and every engagement and company after it can use it.
Land, expand, and extract works in two dimensions. Inside a market, we land with one kind of customer, expand to similar customers using the same parts, and extract what we build into polis so it can be reused. Across Metropolis, consulting and the venture studio follow the same plan, and consulting goes first. It sees more kinds of problems, faster, than any one company would, so by the time the studio starts a company, much of what that company needs is already in polis.
Running a consultancy also means selling: finding clients, keeping track of relationships, and following up. That pushes us to build sales and customer relationship tools into polis, and nearly every portfolio company will need those too.
As the consultancy grows, we will hire people who can run it on polis without me doing every engagement. The venture studio picks back up once we have found a market worth entering and a position in it. When it does, its first company will start with everything consulting has already built, and the firms we have worked alongside will be the first people we talk to about it.
If you are a seed or Series A team looking to supplement your technical leadership, tell us what you're working on.