Pieter Levels: From 12 Startups in 12 Months to a $10M Year
Pieter Levels turned repeated small bets into a portfolio of internet businesses—and in September 2026 reported passing $10M a year when business revenue and investment gains were combined, while his operating businesses remained radically lean.
Pieter Levels
@levelsio · XNomad List, Remote OK, Photo AI & more
Pieter Levels built his reputation by doing something that sounds almost irresponsible when described as a business strategy: he kept shipping.
Instead of spending years trying to find one perfect startup idea, Levels treated the internet as a laboratory. Build something small. Put it online. See if anyone cares. Keep the winners. Kill the rest.
That philosophy started with his famous “12 Startups in 12 Months” experiment and eventually produced businesses such as Nomad List, Remote OK, Photo AI and Interior AI.
By September 2026, the experiment had become something much larger. Levels reported passing $10 million a year in combined revenue and investment gains, with operating business revenue still around $200,000–$250,000 per month and a reported profit margin of about 94.5%.
The important detail is that the $10M number was not simply SaaS revenue. Levels explicitly described it as a combination of business income and investment gains. The operating companies were still much smaller than the headline number.
That distinction makes the story more interesting, not less.
Levels' early approach was intentionally anti-startup.
Instead of asking what company he should spend the next five years building, he asked how many small products he could get into the world.
The result was a year of rapid experiments.
Most failed.
His current project archive makes that unusually clear. Levels describes it almost like a resume of experiments: a small number became durable successes, some made money but did not last, and a much larger number failed or remained unfinished.
That failure ratio is part of the strategy.
If every idea has to succeed, experimentation becomes emotionally expensive. If failure is expected, shipping becomes much easier.
The lesson is not “launch twelve startups every year.”
The lesson is to make individual experiments cheap enough that failure does not threaten the entire business.
One of the experiments that escaped the lab was Nomad List.
The original idea was simple: create a useful database for people working and traveling remotely.
But the value was not only the information.
It was the network around it.
People compared cities, discussed destinations, looked at costs, and eventually interacted with other people living similar lifestyles.
This was an early example of Levels discovering that a useful dataset can become a community when people have a reason to return.
The product did not need a giant organization behind it.
It needed enough utility that a small group of users cared.
Remote OK followed a similar principle.
Remote workers needed jobs.
Companies needed remote talent.
The product connected the two sides without trying to become a massive general-purpose employment platform.
That narrowness mattered.
A small founder can often win by removing everything that does not directly support the core transaction.
Levels repeatedly demonstrated that he would rather operate a narrow product with clear economics than a huge product with a huge organization.
Over time, Levels stopped looking like a founder with one startup and started looking like a small internet company operating a portfolio.
Some projects produced meaningful revenue.
Some became experiments.
Some were abandoned.
Some became technical playgrounds.
The portfolio structure created an important advantage: Levels did not need every new project to become a company.
A project could simply teach him something.
A product could produce traffic.
A failed idea could provide technology for the next one.
A successful product could finance experimentation elsewhere.
This is one of the biggest differences between the portfolio approach and conventional venture-backed startup building.
The goal is not necessarily to create one enormous outcome.
The goal is to continuously increase the number of things that can work.
Photo AI became another major example of the model.
AI dramatically reduced the amount of engineering required to turn an idea into a usable product.
Levels could combine existing models, APIs, infrastructure and his own product layer instead of building every capability from scratch.
The result was not “AI did the startup for him.”
The hard part moved.
When software becomes easier to build, choosing what to build and getting people to use it become more important.
Levels' public work increasingly reflected that reality.
By 2026, Levels was using AI coding tools aggressively.
But the interesting part was not that he used AI.
It was what he used AI for.
He began replacing parts of his infrastructure stack with software he could build himself.
Scraping services.
APIs.
Internal tooling.
Small operational systems.
Instead of automatically accepting a monthly SaaS bill as the cost of doing business, he increasingly asked whether the dependency itself was worth keeping.
In September 2026, he described replacing roughly $25,000 per month of SaaS and infrastructure dependencies with systems he built himself.
That is a very different definition of “efficiency” from simply negotiating a cheaper subscription.
He was reducing the number of businesses sitting between his code and his customers.
On September 23, 2026, Levels published the most striking financial milestone of this phase: he said he had passed $10M per year in revenue plus investment gains.
He also reported approximately 94.5% profit.
But the mechanics matter.
His operating businesses were still reported around $200K–$250K per month.
The rest of the headline figure came from investment gains.
Levels described a loop:
lower expenses → higher profit → more capital → investments → investment returns → more capital
This is where the portfolio founder model becomes a personal capital-allocation model.
The company is no longer just producing income.
It is producing capital that can compound elsewhere.
A $10M company with a huge payroll is one kind of achievement.
A tiny team generating substantial cash is another.
Levels has spent years pushing toward the second model.
His businesses are designed to be operated with very few people.
That creates an unusual relationship between revenue and complexity.
More revenue does not automatically require more employees.
More customers do not automatically require a new department.
More features do not automatically require a large engineering team.
The constraint is deliberately kept low.
The internet often turns Levels into a “genius founder” story.
His own project archive tells a less flattering and more useful story.
Most of his projects did not become meaningful businesses.
That is not a footnote.
It is the system.
If only a minority of projects succeed, then the founder's advantage comes from being able to launch enough experiments and survive enough failures to reach the next winner.
That is why his story should not be interpreted as “copy Pieter's products.”
The transferable asset is the operating system.
Across the portfolio, a few principles keep repeating.
Levels frequently ships before he has perfect confidence.
The market provides information that planning cannot.
A failed project should cost weeks, not years.
Products with direct traffic, search demand, communities or audiences are less dependent on paid acquisition.
Revenue is useful.
Complexity is not automatically useful.
Once a product produces reliable cash flow, the founder can fund more experiments without outside capital.
Every failed project leaves behind code, knowledge, distribution, SEO data, customer conversations or technical infrastructure.
The most interesting part of Pieter Levels' story in 2026 is that the original thesis has survived several technology cycles.
The tools changed.
The businesses changed.
AI changed the cost of software.
But the underlying behavior remained:
Build something.
Ship it.
Measure reality.
Keep what works.
Delete what doesn't.
Then use the winner to finance the next experiment.
That is how a year of twelve startup experiments eventually turned into a portfolio of internet businesses—and, by Levels' own September 2026 accounting, a year above $10M when operating income and investment gains are considered.
The biggest lesson is not to become Pieter Levels.
It is to lower the cost of being wrong.
If your first product needs six months before anyone can use it, every idea becomes a huge bet.
If you can build a useful first version in a weekend, you can run ten bets instead of one.
And when AI makes software dramatically cheaper to build, the competitive advantage moves toward taste, distribution, speed, and persistence.
Levels spent a decade proving the first three.
The latest chapter suggests the fourth—capital allocation—may be just as important.