SoloCook asks what's actually in your fridge, then shows you what you can cook tonight — scaled for one person, and running entirely on your phone. This is the story of how it got made.
Cooking for yourself means doing arithmetic before you're allowed to start. Halve this, quarter that, then work out what to do with the other three portions of something you've already gone off. Every recipe app quietly assumes there are four of you.
What struck me is that the failure isn't appetite, and it isn't skill. It's friction. Deciding what to cook, checking whether you've got the things, scaling it down, accepting you'll eat it again tomorrow — none of it is hard, and all of it is enough to make you open a delivery app instead.
The delivery apps have no such problem. They know what you want, they remember your address, and they'll take somewhere north of £200 a month for it. A decent week of cooking for yourself costs about £30. The maths isn't close — the friction is just all on the wrong side. So I stopped asking "what should I cook?" and started asking "what can I cook, right now, with what's already in the fridge?"
Built for a household. Serves four, and expects a shop before you begin.
Solve the wrong week. You cook once on Sunday and eat the same thing five times.
A folder of things you'll never make, because none of them match tonight's fridge.
Once it was a friction problem rather than a discovery problem, the shape of the answer came quickly. I worked through it in a few moves.
Don't ask what someone wants to eat. Ask what they already own, and work backwards to the things that are actually cookable tonight.
A grid of ingredients you tap once, kept as state. Typing out a shopping list every time is exactly the friction the app exists to remove.
A hob and a kettle produce a different list to an air fryer and an oven. Appliances turned out to matter as much as ingredients.
Anything you're one ingredient short of gets flagged rather than dropped — so you know what to pick up on the way home.
This was the real lesson. A well-stocked kitchen is the easy case. Most of the design effort went into what the app says when there's barely anything in.
You have everything. Start now, no shop required.
Missing a single thing, and the app names it.
Enough to cook, but a staple or two would unlock far more.
Said plainly, with the two items that would fix it.
I designed SoloCook and then built it conversationally — describing what I wanted, reading every diff that came back, running it on a real device, and correcting course. Two native apps rather than one cross-platform compromise.
That turned out to be a design constraint rather than a shortcut. Because a working screen took minutes, I ended up designing in the running app instead of in a static mockup — which is how the awkward states got caught. What the home screen says before you've cooked anything. What results look like with four ingredients ticked. What happens when an image fails to load.
It's also perfectly happy to hand you something that compiles, looks right and is quietly wrong. So the discipline isn't in the writing, it's in deciding what to check. The ingredient matcher got a test harness before it got an interface, and the harness caught two real bugs on its first run. A navigation change got driven on an actual emulator rather than reasoned about.
Your pantry, bookmarks, streak and saved money live in the operating system's own key-value store. No account, no sign-in, no sync — delete the app and it all goes with it.
A single Cloudflare Worker turns a recipe name into a photograph — Wikipedia first, then a stock search, cached for a month. So each dish costs one lookup, ever.
Tap what you've got. Get the things you can actually cook tonight.


Built and working
Not claimed
I'd rather show a real product with honest gaps than invent a results slide.
Free. No account. Nothing leaves your phone.