Building perpHect — The Honest Version

I am not a developer. I do not have a co-founder. I did not raise money. I built perpHect around a full-time job, mostly in evenings and at weekends, over the better part of a year — and I want to write down what that actually looked like, because the version that usually gets told is smoother than the reality.

The idea was not original. Lots of people have built water reminder apps. That was part of why I thought there was something to do.

Why hydration

I noticed something about myself a couple of years ago. I was having bad afternoons — low energy, difficulty concentrating, the kind of thinking that felt like running through mud. I had tried the usual explanations: sleep, diet, too much screen time. None of them quite fit the pattern.

When I actually tracked what I was drinking on a bad day, the answer was embarrassingly simple. I was barely drinking anything. Not because I had forgotten or was being negligent — but because nothing in my day was prompting me to, and I had a generic "drink 8 glasses" target that felt arbitrary enough that I never really committed to it.

The 8 glasses number, I later learned, has essentially no scientific grounding. It is a population average that lost its context when it entered public use. My actual requirement on a given day depended on how I slept, how stressed I was, whether I was going to train, and what the weather was like. None of the apps I tried knew any of that. They all gave me the same fixed target, every day, regardless of what the day actually looked like.

The no-code bet

I had no idea how to build an app. What I did know was that AI-powered no-code tools had moved far enough that a non-technical person could now build genuinely functional software — not a prototype, actual software — if they were willing to learn the tools and iterate relentlessly.

I spent the first few months learning. A lot of the early builds were wrong in fundamental ways. The personalisation logic was too simple. The interface asked too much of the user. The AI coaching output was generic in the ways that AI output becomes generic when you do not put enough work into the prompt architecture.

I rebuilt the calculation engine three times before I was satisfied that the target it produced was actually more accurate than a fixed number. The five-input model — body profile, sleep quality, stress and energy, planned activity, live weather — came out of that process. Each input changes the output meaningfully. Removing any one of them makes the target worse. That structure took a long time to get right.

The part nobody tells you about

Building the product was the part I understood least going in. But it was not the hardest part. The hardest part was everything else: understanding App Store submission requirements, figuring out how static HTML blog architecture works when your main site is a JavaScript SPA that AI crawlers cannot read, learning enough about SEO to not make the obvious mistakes, writing copy that does not sound like it was written by someone who has read too many SaaS landing pages.

None of this was in the plan because I did not have a plan. I had a product and a belief that it was better than what existed, and I figured out the rest as I needed to.

Where it is now

perpHect is live on the App Store at £3.99 a month after a 7-day free trial. The core mechanic — recalculating your daily water target every morning from five live inputs — works the way I wanted it to. The AI coaching plan it produces each morning is specific to your actual day, not a template. The adaptive reminders fire when you are genuinely behind pace, not on a fixed schedule that becomes invisible within days.

The review count is low because the app is new. That is the honest bottleneck on everything right now. The product is good. Awareness is the work.

Why I am writing this

Partly because Fable — the AI I use for strategic research — told me that "non-technical solo founder builds AI health app" is exactly the kind of story that indie app journalists and Product Hunt communities respond to, and I should write it down. That is true, and I am comfortable admitting it is part of the reason.

But also because when I was trying to figure out how to do this, I could not find many honest accounts of what building a no-code app actually involved — the failed builds, the arbitrary-feeling decisions about which no-code platform to bet on, the months of work that felt like treading water before anything clicked. I would have found that useful. This is my version of it.

If you are building something similar, or thinking about it — feel free to reach out. support@perphect.ai