Many ideas that would make the world better never become companies. One big reason is money.
A company spends money long before it makes any. Someone has to build the product, someone has to find customers, and someone has to keep things running once those customers arrive. Investors cover that gap in exchange for a chance at a large return, and plenty of useful companies can offer one. When an idea can't, like a service that helps families find benefits they already qualify for, or a project that brings clean water to a town without it, it usually exists only because a few people decide to carry it anyway, often for years and at real cost to themselves. Many never get started at all.
That is the barrier I want to break down, and AI is what makes it possible. At Metropolis we are building pipelines in polis, the platform the studio runs on, that take a company from idea to launch and then keep it running. They use the same building blocks we use to run the studio itself: shared knowledge, planning, and AI agents that do the work. If that works, starting and operating a company costs far less. An idea wouldn't have to promise venture returns. It would only have to pay for itself, with grants and donations where it needs them.
Cheaper to build, cheaper to run
In AI isn't a genie, I argued that AI delivers when it works inside a system built around it. In The difference a year makes, I described what that looks like at Metropolis today. Agents plan, build, review, and release our software, and I approve the decisions that matter. Work that used to need a team of engineers now needs me and the system.
Software is only one part of a company, and the rest of it still runs on people's time. The next step for polis is to take the same approach to that work, so a pipeline can carry an idea through research, launch, and day-to-day operation, with a person approving the decisions that matter along the way.
Every company would run on the building blocks we use to run Metropolis itself, so a new company wouldn't start from nothing. It would still need what is specific to it: a customer, a problem worth solving, knowledge of the field, and someone to oversee it. The platform would carry the rest. When a company needs something polis doesn't have yet, the people running it can request it. If the new piece only makes sense for that company, it stays with that company. If others could use it, it becomes part of polis. Each company leaves more behind for the next, so every company should be cheaper to start than the one before it. A company that costs a fraction of what it does today to start and run doesn't have to grow huge to justify its existence.
Companies like GiveCard
I have followed GiveCard since the beginning, and I know its founders. Lurein Perera started it as a student at Boston College, after someone on the street asked him for money and he had no cash to give. That moment grew into a platform that gets money to people quickly, on a card that works without a bank account. Cities and nonprofits use it for guaranteed income programs and disaster relief, including payments to survivors of the Los Angeles wildfires. It has issued more than 100,000 cards and works with more than 100 government agencies.
Getting there took years from a small team. Moving money is hard for any young company, and GiveCard's founders have put in an enormous number of hours. I think it has been worth it, because they have helped people at some of the hardest moments of their lives.
GiveCard is the kind of company I want to see more of. Most ideas like it never get that far, because few founders can carry that much for that long. So I want a second rule for what gets built: a company people need is worth building as long as it can break even after grants and donations. Law firms have a name for this kind of work. They take on pro bono cases because they have the skills and the cases are worth doing, even though they don't pay like the rest.
What we would build
Some of the most useful companies would do something simple: gather information that is already public and make it easy to understand.
Local government is one example. City councils and school boards make decisions that shape people's streets, taxes, and schools, and they publish them as long agendas and minutes that almost nobody reads. A service that tells you in plain language what your town is deciding this week, and how it affects you, would make it easier to show up when it matters.
Benefits are another. Food assistance, tax credits, and help with utilities and childcare come from different programs, each with its own rules and forms. The IRS says nearly 1 in 5 eligible taxpayers don't claim the Earned Income Tax Credit, often because they don't know they qualify. A service that takes one household's situation and tells them what they can get, and how to apply, would put money in the hands of people who are already owed it.
Hospitals have been required since 2021 to publish their prices, but they do it in files built for regulators that patients can't use. Turning those files into an answer to "what will this cost me, and where is it cheaper?" would help people before the bill arrives.
A company doesn't have to be an app, either. Clean water and reliable food are crises in many places, and solving them means building real things: wells, treatment systems, farms, and the supply chains behind them. People would do that work. A company built on polis could research the options, plan the project, price it, source the materials, and track progress, so the people on the ground can focus on building.
A second studio
The simplest way to follow that second rule is to give it its own studio. polis is being built to run more than one studio. Each one would use the same building blocks and pipelines, set up with its own rules for which ideas to take on and what counts as success. Building and running a company is the same work whether the goal is profit or impact. Only the rule for what gets built changes.
That studio would judge a company by whether people need it and whether it can sustain itself. Its companies would feed polis like any other, so the impact work would make commercial companies cheaper to start, and the commercial work would do the same for impact companies.
Much of the public conversation about AI treats it as a weapon or a profit machine, and some of those fears are fair. It can also be a value creation engine: a way to build things that used to cost too much to exist. If the system we are building works, this is how I intend to use it.