An AI-native venture studio
One founder. Fleets of AI agents. A portfolio of startups.
The Unnamed Roads is the new way to test and run startups. Instead of raising a big round and hiring a team, one founder runs a portfolio of companies with fleets of AI agents doing the research, code, design, writing and operations. The skill is managing the agents and running lean: small tests, low cost, a human yes before anything goes out. Every project is listed here with an honest status.
- Projects
- 14
- Live
- 3
- Testing
- 8
- Parked
- 3
How a company gets started here
The same four steps for every idea. AI makes each step fast and cheap, so many ideas can be tested instead of betting everything on one.
- 1
Pick one real problem
AI scans markets, forums, search data and competitors. We pick a problem that a specific group of people clearly has.
- 2
Build a test in days
AI writes the code, the copy and the design for the smallest thing that can be tested: a landing page, a prototype or a simple tool.
- 3
Measure real demand
The test goes live. We decide in advance what success looks like and let the numbers, not opinions, decide.
- 4
Keep, change or park
Projects that work keep growing. The rest are parked in public, so the lessons are not lost and the next idea starts faster.
Current projects
The companies the studio is running or testing right now. Live means real users can use it today, Testing means we are checking if people want it, and Parked means it is paused.
Live (3)
Running and open to real users or customers.
- Live SportsThe Hockey Analytics Short, proactive hockey insights instead of dashboards. Visit The Hockey Analytics →
- Live FamilyFöräldraledighetsplanen Plan Swedish parental leave days and pay in minutes, also from inside AI assistants. Open the planner →
- Live SportsThe Hockey Brain One club-owned database for all of a hockey club's data. Visit The Hockey Brain →
Testing (8)
Being tested for real demand before more is built.
- Testing B2B softwareThe Outcome Layer Pay for finished work, not consultant hours. Visit the site →
- Testing MarketingThe Atomic Network Neighbourhood postcards and referral codes, triggered by real purchases. See the pilot →
- Testing PrintThe Print Route A print engine that turns digital orders into physical mail and print. Visit the site →
- Testing FamilyCPA – Car Planner Assistant Know what a car really costs before you buy or lease it. Visit live site →
- Testing B2B softwareTBS – The BRF Styrelse An AI assistant for Swedish housing cooperative boards.
- Testing B2B serviceThe Data Labelers Verified AI training data, labelled in Sweden or on your own platform. Visit The Data Labelers →
- Testing B2B softwareThe Mention Engine Find out whether AI assistants mention your company, and how to fix it.
- Testing RetailThe Shelf Route Price tags, shelf labels and car signs, designed, printed and delivered.
Parked (3)
Paused, but could come back later.
- Parked B2B softwareThe Agent Fabric Infrastructure for running AI agents privately, close to company data. Visit the site →
- Parked EducationSimulation Trainer AI role-play for practising hard conversations.
- Parked FamilyBaby Checklist Simple preparation checklists for expecting parents. Visit the site →
Earlier experiments that have been stopped are listed in the closed archive.
How a venture studio differs from an accelerator
Traditional accelerator
- Invests in outside founders
- Months of building before the first real test
- Needs a team for code, design, marketing and analysis
- Few bets, large budgets per bet
AI venture studio
- Starts and owns its own companies
- A first test is live within days
- AI does the code, design, marketing and analysis
- Many small bets, cheap to test, honest status for each
More detail: agency vs venture studio vs incubator.
Field Notes
All Field Notes →- Side Project To Profitable Product The allure of building a successful online business is undeniable. With the rise of digital platforms and tools, it's easier than ever to turn an idea into a th
- MCP Explained: Model Context Protocol for AI Agents What MCP is, how to build servers and clients—and why our venture studio ships every new agent on MCP (200+ pipelines in production).
- How I Run 7 Websites With One Person and AI Agents A detailed breakdown of the automation stack behind The Unnamed Roads studio: self-hosted on Hetzner, n8n workflows, AI-generated content pipelines, and what actually went wrong building it.
Common questions
- What is The Unnamed Roads?
- The Unnamed Roads is an AI-native venture studio based in Stockholm, Sweden. It builds its own companies (it does not invest in other startups, and it is not an accelerator programme or agency): AI does almost all of the work of researching, building, launching and testing them, and one founder, Emil Ingemar Karlsson, decides which ideas to pursue and approves everything that goes public.
- Can one founder run several startups with AI agents?
- Yes, if the agents are managed well. At The Unnamed Roads, fleets of AI agents handle research, code, content, analysis and routine operations across the portfolio, with cost limits and a human approval gate. The founder chooses problems, allocates attention and decides what goes out. That keeps costs low enough that a startup test does not need outside capital to start.
- How do you start a company with AI?
- Start with one clear problem for one group of people. Use AI to research the market and competitors, draft the offer, and build a simple website or prototype in days instead of months. Put it in front of real people, measure whether they sign up, pay or come back, and use AI to analyse the results. Keep what works, change what does not, and park the rest. The How it works guide describes each step in detail.
- How can I test a startup idea with AI?
- Write down what has to be true for the idea to work, then let AI build the smallest possible test: a landing page, a prototype or a manual service. Decide in advance what result counts as success (for example sign-ups, replies or payments), run the test for a fixed period, and let AI help you read the numbers. A failed test that took a week is a cheap lesson.
- What is an AI venture studio?
- A venture studio starts its own companies instead of investing in other people’s. An AI venture studio does this with AI doing most of the work that would normally need a team: research, coding, design, writing, data analysis and operations. That makes it possible for one person to start and test many companies in parallel.
- How is this different from a startup accelerator?
- A traditional accelerator invests in outside founders and supports them with mentoring and a programme. The Unnamed Roads builds its own companies, uses AI instead of a large team, and publishes every project with an honest status: Live, Testing or Parked.
- Does AI really do almost all of the work?
- Yes, for most of the hands-on work: market research, code, websites, copy, analysis and routine operations are produced by AI tools and agents. A human still chooses which problems are worth solving, talks to customers, checks the quality, and must approve anything that is published, deployed or sent to someone outside the studio.
Newsletter
Field Notes from the studio
What we learn from starting and testing companies with AI: what worked, what did not, and what it cost. Only sent when there is something worth sharing.
No spam. Unsubscribe anytime.
Have an idea, a problem worth solving, or want to collaborate? Get in touch.