I built DiscernAI because I’m bad at making decisions. Not for lack of information. I had way too much of it. I’d open twelve tabs, read a pile of conflicting opinions, and somehow end up more stuck than when I started. So I did the thing engineers do when they’re uncomfortable: I decided to write code until the feeling went away.
That was the first mistake.
The problem wasn’t the one I thought
My theory was that people need better structure for decisions. A framework. Criteria, weights, a final score. Tidy and rational, the way I wish my own brain worked.
So I built exactly that. Then I handed it to a few early users and watched them go through the whole flow, look at the recommendation the app spit out, and then calmly go do the thing they were already going to do.
That was annoying. It was also the most useful thing that happened in the whole project. People don’t want a better answer. They want to trust the answer they already have. The decision is usually sitting there half-formed, and what’s missing is permission to act on it.
So I stopped optimizing and started clarifying. The model’s job isn’t to decide. It’s to lay out what you already care about, make the tradeoffs visible, and then get out of the way.
What I’d do differently
I spent the first two months building the scoring engine before I’d talked to a single person about an actual decision they were stuck on. Classic. When I finally had those conversations, nearly every assumption I’d baked in turned out to be a little off, tilted toward how I decide things and not how most people do.
If I did it again: five conversations before one line of code.
The stack (SwiftUI, Gemini, Supabase, CoreData) was mostly fine. What I got wrong was where the product actually lives. I thought it was the infrastructure. It was the prompt: the framing, the questions, the tone. I spent maybe 60% of my time on architecture and 40% on the part users actually felt. Should have been the other way around.
What surprised me
350+ people signed up with zero ad spend. That was the first signal that the problem might be real. People are worn out by the number of choices and the noise around each one.
But the use cases caught me off guard. I built this imagining big, weighty life decisions. What people actually brought to it: which show to watch tonight, which dog food is least bad, should I take this job or the other one. The trivial decisions and the huge ones often needed the same thing, which is a way to get out of your own head for a second and see what you actually care about.
Why I keep coming back to it
I’m a PM by trade, so I usually work between what’s buildable and what people need. DiscernAI was the first time I owned the whole surface myself: the idea, the design, the code, the words on the landing page. I learned more about product from that than from years of pushing other people’s roadmaps.
I built it because I think tools should sharpen human judgment, not stand in for it. The more software tries to decide for us, the more I want tools that help people stay in charge of their own choices.
