Cheeks

Creator, designer and product lead

A weight-neutral food tracker that never counts calories, for people who want to eat enough.

THE CHALLENGE

Most food trackers are built around calories and limits. For people with a history of restriction or binge eating, or anyone relearning how to feed themselves on or after GLP-1 medication, that framing does real harm. But those same people often still want to know whether they're getting enough protein, fiber and variety to support their training and their gut.

Cheeks is in beta on TestFlight and launches on the App Store soon. Follow along at cheeks.world.

HOW MIGHT WE

How might we help people eat enough without counting anything they've learned to fear?

THE PRINCIPLES

Rules written before any screens

I wrote the product principles first and treated them as non-negotiable. Every feature request gets checked against them, and when one conflicts, the principle wins.

Additive framing. Cheeks asks what you got today, never how much is left.
Calories are a floor, not a ceiling. Energy is hidden by default and only answers "am I fueling enough?"
Movement never earns food. There are no calories burned anywhere. A training day just means fuel up.
No red, no limits, no streaks. Nothing turns a warning color and nothing counts down.
Flags are trends, not days. Anything worth noticing shows up across weeks, in the weekly review, never as a push notification.
THE HAMSTER

A mascot that can't be disappointed in you

A pixel-art hamster fills its cheeks as you log. It fills with what you add, like a new plant this week or your first meal of the day, and never with the amount you ate. Once the cheeks are full, they stay full for the rest of the day.

Early versions had the hamster carry its food back to a burrow at the end of the day. I cut it. Cheeks emptying after eating could read as purging, and that was enough of a risk to lose a feature I liked.

There is no overstuffed state in the art, the animation or the code. A big day and a light day get the same calm, happy hamster.

The hamster with full cheeks on Today.Today's meals, each showing what it added in protein and fiber.
LANGUAGE

Copy that's tested like code

Every word a person reads lives in one string catalogue, and an automated test fails the build if it finds a banned term. The list includes limit, over, too much, burn, earn, cheat, guilt, junk, deficient, remaining, left, "to go" and back on track.

Counts are collections. "12 plants collected," never "18 to go."
Say movement, not workout, and log a meal rather than add one.
No hoarding words. I renamed the plant "stash" after it started to read too close to binge eating.
No diagnosis. "No oily fish this week, want ideas?" is fine. "You might be deficient" never is.
ONBOARDING

Personal without asking for weight

Onboarding asks what's motivating you, how you move and your age range, which is enough to pick a cited protein range. It never asks for weight or sex, and it lets you turn off measurements entirely, so Cheeks names foods instead of showing grams.

LOGGING

Rough counts fully

Every entry can be rough or exact, and a rough amount counts just as much. Food data comes from USDA FoodData Central and the Canadian Nutrient File, bundled on the phone.

Search USDA foods and brand-name products
Snap it to have the phone suggest the foods in a photo
Describe it by typing or saying a meal
Scan a barcode or read a nutrition label
Ask Siri to log a meal

Photo and description parsing run on Apple's on-device models, so a meal photo never leaves the phone. Nutrients always come from the food data, never from the model.

Log food, with Snap it, Describe it, Scan barcode and Add your own food.Saved, with hearted meals, recipes and saved foods.
INSIGHTS

The week leads with wins

The weekly review opens with new plants, protein consistency and hearted meals. Insights shows the food groups you ate as a plate you can push around with a finger, for a week, a month or all time.

WHAT I LEARNED
Harm lives in small details. The riskiest moments weren't features, they were a word like "left" or an animation of cheeks emptying.
Write the principles down where the work happens. Rules that sit in a doc get negotiated away. Rules enforced by a test don't.
Building it myself changed my design decisions. I designed and built Cheeks with an AI-assisted workflow, which let me test an idea in the real app in an afternoon instead of arguing about it in a mockup.
WHERE IT IS NOW

In beta, launching soon

Cheeks is in beta on TestFlight and has been submitted to the App Store. Follow along at cheeks.world.

Next case study
Storybook
Contact

Let's make something that does good.

347-852-8466 / Los Angeles, CA