Partner Story · in conversation with Oskar Glauser, founder of Mira and author of The Code Was Never the Hard Part. This is a written interview submitted through Submit Your Story. The claims below are his own, presented as they were given to us. No payment was involved.
Oskar Glauser, Stockholm. Photo: courtesy of Oskar Glauser.
In the past twelve months, Oskar Glauser has built more than twenty apps. He did it mostly in the evenings and on Fridays, alongside a design studio, and he spent less than $100 a month on tools to do it.
Two of those apps shipped. One of the two, he has since shut down.
Most people would tell that story as a confession. Glauser tells it as a method.
"Building felt like progress, so I kept building," he says. The ratio, twenty to two, is the most useful number he has, because of what it reveals about the eighteen that never left his laptop.
A designer who could finally build
Glauser is a designer who builds, based in Stockholm. For twenty years he has turned ideas into products, his own and those of companies such as H&M, Klarna and Greenely. For most of that time, he could only draw what should exist. Someone else had to write it.
That changed in early 2025. He started with HTML prototypes for a client, then built parts of an app in Swift, then a whole design system in Flutter, with AI writing the code.
But he is precise about how he uses it, and it is not the way most people do.
"I never start by asking for code," he says. "I write the plan together with the AI first: every screen, the edge cases, the tech stack. Then I ask it to tear the plan apart before a line gets written."
That second step, the one where the machine is asked to argue with you, is where he says most of the expensive mistakes get caught: the wrong stack, the overbuilt first version, the login choice he would have regretted a month later.
Mira: interviews at the scale of a survey
The clearest result of that year is Mira.
It comes from a frustration Glauser saw inside almost every company he worked with. Teams make product decisions on the strength of "we did some interviews last quarter", because talking to real users takes too long to arrange. Surveys are quick, but shallow. Real interviews are deep, but rare.
Mira tries to have both. A team shares one link, in an onboarding email, inside the app, in a message. An AI interviewer then holds a natural conversation with each user, by voice or by text, in that user's own language, and asks follow-up questions the way a good researcher would. When the conversations are done, it pulls them into a report: the findings, the recommendations, and the reasons, in the users' own words.
The question on the front page is the one that costs startups the most money: why aren't your new users converting?
Mira runs on OpenAI's latest models, at different levels for different parts of the work. The first working version took one weekend.
The lesson that cost him the most
To understand why Glauser counts shipped products rather than built ones, you have to go back to 2008.
That year he co-founded Videofy. By most measures it did everything right. It started small, proved the idea worked, raised money on real numbers and reached number one in the Swedish App Store. It grew to half a million users and 40 million video views a month, and at its peak it brought in around €800,000 a year, mostly from advertising.
It still failed.
"We had picked a race against Instagram and YouTube, with a fraction of a fraction of their resources," Glauser says. When the giants turned to video, they took Videofy's users and its advertisers within a short time. The advertising model, which had carried the revenue, was part of why the fall was so fast.
"It took me four years to see that the code was never the problem."
What came next was quieter and has lasted longer. Glauser bootstrapped Minutemailer, an email marketing tool for small businesses. It has now run for more than a decade without outside money. Hundreds of small companies send about two million emails a month through it.
One business that raised money on impressive numbers and disappeared. Another that never raised a cent and is still standing. That contrast is the spine of everything he now teaches.
What the two had that the eighteen did not
So what separated the two shipped apps from the rest?
Not the quality of the code. Not the features. Glauser says both shipped products shared two things the others lacked: each started from a problem he had seen over and over, and for each he knew exactly who to show it to on the first day.
Mira is the example. It did not go looking for an audience. It went straight to Minutemailer's customers, people Glauser already knew and who already had the problem.
The other eighteen were, in a sense, solutions waiting for someone to need them. With AI, they were cheap enough to build that nothing stopped him from building them. That is the trap he now writes about. When building costs almost nothing, building stops being evidence of anything.
Why he shut down Tempo
The second shipped app was Tempo, an AI-supported work tool for small startup teams, a little like Trello. Glauser has closed it, and he wanted that decision in this story.
"I shut it down because the way I work had already moved past the workflow it was built around," he says. He no longer needed issues and boards in that form, and it was hard to make Tempo different enough from the tools already on the market.
He is not embarrassed by it. "Shutting it down on purpose felt better than letting it drift."
For anyone building with AI, this may be the most practical lesson in his whole account. A product that costs a weekend to build can quietly cost months of attention to keep alive. Deciding to stop is a skill, and Glauser treats it as part of the method rather than a failure of it.
The book: everything around the code
That method is now a book. The Code Was Never the Hard Part came out as an ebook on Amazon on 15 October, with a paperback to follow in November.
Its argument is simple. Building got cheap. With AI, someone who has never written code can have a working app by Sunday night. What did not get cheaper is everything around the code: picking a problem worth solving, finding out whether anyone wants it, explaining it to a stranger, and getting it in front of the first hundred people. So launch platforms fill up with good-looking products nobody uses.
Most startup advice, Glauser points out, was written when building was the expensive part, for teams with funding. His book is written for the opposite reader: one person, often building for the first time, usually next to a day job. It is short enough to read in a weekend, and every tool in it comes with an exercise, so the reader's own idea goes through the method while they read.
What comes next
Glauser would like to hear from people building with AI, especially first-timers who have launched something and heard nothing back. And for teams that want to talk to more of their users, Mira is open to try.
The first chapter of the book and templates for every tool in it are free at neverthehardpart.com. Mira is at letmira.com.
---
This is a Partner Story: a written interview with a founder building with AI, submitted through Submit Your Story and published free of charge. Statements about the founder's products, revenue and results are his own. Building something with AI? Tell us about it.
Would you try this product?
