Nobody uses Parallax, including me
I pulled the production numbers for Parallax this afternoon. Eight users. Seven couples. Forty two answers, ever. The last person signed up on 14 July. The last time anyone answered a daily drop was 16 July.
Nobody uses it. I don't use it.
Parallax is a couples app I put on the App Store at the end of June. First commit on 21 June, live eight days later. Daily prompt for both of you, neither sees the other's answer until both have submitted, with the reveal gate enforced in row level security rather than the client. I'm still happy with that part. The engineering is fine.
Building an iOS app had been on my list for years. That was the actual goal, and that's the whole problem. I wanted to have built an iOS app. I didn't start from a couple with a problem.
It isn't that I skipped the research. I did a lot of it. Therapists warn that scoring couples is toxic and a low number can start a fight, so I designed the wavelength score to read as curiosity rather than a verdict. The APA issued an advisory about AI as therapist, so I ruled out mediation during a fight and positioned it between moments. I read the Habermas Machine result on AI drafted statements out-endorsing human mediators and wrote eight safety requirements that had to be met before Refocus could ship to anyone real.
Refocus is the feature I thought hardest about. It has been used twice in production.
I researched the domain carefully and never once checked whether anyone wanted the product.
The clearest tell showed up when friends finally tried it and said the onboarding was super confusing. I went looking for the cause, expecting missing features. What I found was that the app shipped its demo data as if it were real. A brand new couple opened Parallax and saw a partner called Dani, a 23 day streak, and a 76% wavelength score, from the first second. Fabricated intimacy, handed to them as theirs.
I'd built something that demonstrated a relationship rather than supported one. The seed data was just saying it out loud.
The part I find harder to sit with is that I'd already written this lesson down. In why Studio didn't work I said the thing plainly: work out who the user actually is before you decide what to hand them. I published that on 1 June. I made the first Parallax commit on 21 June.
Then, when the numbers didn't move, I rewrote the app. I tore out the daily drop product, made it open on the fight instead of the pitch, and let a solo user in without a partner, which fixed a genuine two sided cold start problem I'd flagged in my own decision log back in June. I shipped v2.0.0 on 25 August.
The last real user activity in the database was 16 July. I rebuilt the product for an audience that had already left, and I did it without speaking to a single one of them. That's the same mistake wearing different clothes.
Ragnarok is the counterexample and it's uncomfortable how simple the difference is. There was a gym, 441 members, an owner who was fed up with the incumbent software, and a migration date. The problem existed before I did. I just had to not get it wrong.
So I'm starting Parallax again, and this time the build comes last. I don't yet know whether v2 was too small a change or the wrong direction entirely, and I'm suspicious of my instinct to answer that by opening the editor. Still early, and the honest position is that eight users over eleven weeks is not a signal to interpret. It's an answer.