About

Hi, I'm Mike — a software engineer, and right now the only user of SelfCtrl.

Why I built this

I wanted to track macronutrients, mostly for homecooked meals, and every app I tried made that harder than it should be. Logging a real meal — not a barcode, not a chain restaurant — usually meant typing out every ingredient and weight by hand. Skip that effort and the numbers stop meaning much. I'd start strong on a Monday and start dreading logging by Thursday, not because I lost motivation but because the app made honest food tracking more tedious than fast food.

The apps that made logging easier came with a different cost: adverts, tracking cookies, barcode databases padded with branded junk, and a premium subscription that felt steep for something this simple. I didn't want to hand over photos of my meals and my weight history to a company whose business model depends on doing more with that data than just showing it back to me.

I'd spent decades building software for a living, so "build it myself" wasn't a leap — it was the obvious next step. A photo of a meal is usually enough information to estimate calories and macronutrients reasonably well; AI models got good enough to do that estimation cheaply, so I stopped seeing typing every ingredient by hand as a necessary evil and started seeing it as a workaround for expensive premium monthly plans.

What I ended up building

Photograph or describe a whole meal or part of it, let AI estimate the nutrition, and get on with the day. No ads, no tracking cookies, no more data collected than the app actually needs, and no data handed to a broker. If the estimate's off, a quick note nudges it — "no rice" or "double the chicken" — rather than starting the entry over.

I also wanted a tool that's honest about its own limits rather than pretending to more precision than it has: AI estimates are approximate, confidence is shown per meal entry, and the app doesn't claim to replace a dietitian or a food scale when you actually need one. It doesn't nag you to log everything, nor to slow down with your eating or drinking. It doesn't claim to know you better than you know yourself, nor what motivates you. It just does one job, does it well, and does it cheaply.

How I'm running it

I'm keeping this deliberately small and cheap to run — no VC money, no outside investors, no growth targets, no pressure to add engagement features that don't serve the actual job. Running costs are pennies per week at this scale, not zero, but genuinely small: each meal item logged or refined costs a fraction of a penny in AI estimation, with storage and data processing being only a few pennies a month. I'd rather keep it lean and price accordingly than build something that only works with a huge user base subsidising it.

It's mine first. I'm the one using it every day, which means the incentives are aligned in a way they usually aren't: if a feature doesn't earn its place in my own logging, it doesn't ship. If it turns out other people have the same problem — wanting honest tracking of real food without ads, dark patterns, or a data-hungry business model behind it — I'd like to find out, and I'm building toward that deliberately rather than rushing it.

SelfCtrl is UK-based and English-first. Other languages are planned for later, once the English version has earned it.

— Mike