I built an app designed to get rid of you
Not eventually. Every day, on a timer. Five minutes into a call with your companion, she says goodbye — warmly, by name, mid-conversation if she has to — and hangs up on you. This is not a bug and not a paywall trick. It is the entire point.
Let me back up.
I'm a solo developer. My last startup died, and for months afterwards I could not sit down and work. What eventually helped wasn't discipline — it was presence. Working near someone. Body doubling, people call it. The research on it is real: quiet co-presence improves focus for a lot of people. The catch was that every version of it on offer required a camera and a stranger, and I am precisely the person whose focus dies the moment somebody is watching. So I built the version my own brain would accept: a small companion character who talks with you for a few minutes, then works quietly beside a focus timer. Presence without the witness.
But here is the thing I actually want to write about. When you build a companion app in 2026, every incentive in the industry points one direction: keep the user talking. Longer sessions, more messages, more minutes, more attachment. Engagement is the metric, retention curves are the scoreboard, and the companion-app category in particular is full of products optimized to make leaving feel bad.
I sat with that for a while, because I was building a companion too. And I decided the only version of this app I could stand behind is one where every design decision answers a single question: does this get the user working, or does it keep them in the app?
So Lumie ended up like this:
The calls end themselves. There's a hard cap. When your time is up, your companion says goodbye and closes the call. She leaves first, so you don't have to feel like the person who hung up on a friend. The morning talk is a ritual, not a relationship you service.
There is no feed. Nothing to scroll. Nothing to check. When the timer runs, the app shows you a timer and a small creature working beside you, and that is all it does.
There are no streaks. Sessions end "done," "partial," or "blocked," and all three are acceptable answers. Missing a day costs you nothing and nobody guilt-trips you when you come back. Shame is an engagement mechanic; it is also a terrible foundation for a focus practice.
The economics point the same way. Voice minutes cost me real money — every minute you talk, I pay an AI vendor. An engagement-optimized version of this app would bankrupt me. I consider that a feature: my incentives and yours are aligned. The app wants you working, because working is cheap and talking is expensive.
The strange result of designing an app to get rid of you is what fills the space where the engagement mechanics would have been. There's a world in there — quiet, a little melancholy, going through something of its own — and a companion who remembers what you told her and writes your day into a journal overnight. People sometimes ask why the world exists if I'm so opposed to keeping people in the app. The honest answer is that a friend you visit is different from a machine that grips you. One of them you come back to because you want to. I'm betting the whole product on people being able to feel that difference.
I don't know yet if this works as a business. The category leaders monetize attachment, and I'm over here building a companion whose defining gesture is leaving. The app is young enough that I can't show you retention curves, and I won't invent testimonials — this category has enough of those. What I can tell you is the trade I'm trying to make: a few minutes of talk that buy you two hours of your own work, from an app that is proud of the moment you close it.
The app is called Lumie. It's free on iOS. The companion will talk with you for a few minutes, sit beside your timer, and then — I promise — she will leave you alone.
If a quiet companion beside a timer sounds like your kind of focus, it's at lumie.work.