We built the tool we needed,
in the order we needed it

House AI

House AI turns a photo of a room into a redesign you can act on, then lets you buy what you see. We shipped it, and then we did the unglamorous part: we scaled it. Paid acquisition on Meta and Apple Search Ads. Creator partnerships across five continents. Store listings localised into 40+ languages. App Store optimisation, paywall tests, price changes, onboarding rebuilds.

It worked. House AI has passed 150,000 downloads, runs on iOS, Android and the web in 175+ countries, and reached the top 10 iOS apps in the Design category in the United States.

What scaling it actually taught us

Growth is where the stack breaks, and you only find out once the numbers get big enough to matter.

We were running five products to operate one app. An attribution tool that knew where a user came from but not what they paid. A subscription layer that knew what they paid but not where they came from. Product analytics holding the same human under a third user id. Error tracking that could tell us what crashed but not who it crashed for. Paywall tooling that needed an App Store release to change a price.

Five dashboards, five bills, five definitions of a user, and one founder every month doing the reconciliation by hand. At our worst month, the attribution tool was reporting revenue at roughly four times what the bank had actually deposited. Neither tool was lying. They just had no shared ground truth, because there wasn't one — each of them was measuring from its own island.

Then we did the other arithmetic. Apple was taking roughly 30% of every subscription before we saw a cent. Add the stack of monthly SaaS bills on top of it, each taking its own cut of the same revenue, and a meaningful share of what the app earned was gone before it reached us.

So we built the thing that would have saved us: one SDK, one user object, one dashboard, and a way to take a payment on the web for roughly 3% instead of 30%. We shipped it into House AI first. We ran it against our own money before we showed it to anyone.

Why we're being careful about what we claim

We spent four years being told by other people's landing pages that something was supported, integrating it, and finding out it wasn't. That is why this site has a Coverage section that lists what is not built yet, and why the ad platforms we cannot connect say “coming soon” instead of appearing in a logo row.

Fictura is invite-only right now because we would rather onboard ten founders properly than a thousand badly.