iOS · in development
Snap a plate.Keep the day honest.
Your camera already knows what dinner was. Point it at the plate and Plateknow reads back the calories and the macros — no database to search, nothing to weigh, no account before you have seen whether it works.
What it is not
No eleven-tap search-and-confirm for one meal
No food scale
No account before the first result
No streak that punishes you for missing a day
The whole product
Shutter to macros, before anything asks you for anything.
Point and shoot.
One photograph of the plate as it is in front of you. Not styled, not from above, not in good light. The one you would actually take.
Watch it read the plate.
The segmentation is shown, not hidden — you see the salmon cut out from the quinoa from the avocado. This is the part most apps skip, and it is the part that makes the number believable.
Correct it in one tap.
Because it will be wrong sometimes, and an app that pretends otherwise is lying to you. Wrong item, wrong portion, half of it left on the plate — all one tap, on the result card, always visible.
The log
A collection, not a spreadsheet.
Meals come back as cut-out photographs of the food you actually ate. Scrolling last Tuesday should feel like looking at photos, because that is the only version of this anyone opens twice.
Four days, as the log shows them.
Why the last app did not stick
It was the typing.
Self-monitoring works — that is not in question. Sticking with it is. In the largest systematic review of the evidence, only 25% of participants were still self-monitoring by the end of the intervention; the eighteen-month trial inside it watched weekly adherence fall from 84% to 56%.
People do not quit because they stop caring. They quit because logging one meal costs eleven taps — search, pick a brand, guess the portion, change the unit, confirm — and eleven taps three times a day is a part-time job. Removing the typing is the entire thesis of this app.
Burke et al., 2011, Journal of the American Dietetic Association — systematic review of self-monitoring in weight loss. Adherence figures from the Wing 18-month randomised trial (n=314) within it.
Rules we build against
Four things this app will not do.
It will not tell you that you went over.
A number is information. "You went over" is a verdict, and nobody asked for one.
It will not punish a broken streak.
You can have a streak if you want one. Breaking it will not be framed as a failure, because it is not.
It will not require a target.
There is a mode that sets no goal at all and simply shows you the day. For a lot of people that is the entire useful product.
It will not be marketed to teenagers.
A systematic review of 38 studies found regular calorie-app use correlates with higher disordered-eating symptoms, and adolescents are the most vulnerable group in it. Whatever the acquisition cost says, that is not a user we go after.
Before the App Store
Get it before it is public — and tell us what to build.
Early access on TestFlight, and no charge during the beta.
The app is being built now, for a December submission. If you leave a note below it is read by the person writing the code — the whole reason this box exists is that pre-launch is the only time a feature request is cheap.
FAQ
How accurate is a photograph, really?
Accurate enough to be useful, and not accurate enough to be a lab. Photographs carry no depth, so portion size is the hard part — which is exactly why correction is one tap and always on screen rather than buried. There is a longer, more honest answer on the accuracy page.
Is it free?
The free tier is not settled yet, and we would rather leave this blank than promise a number the App Store listing might not match. The waitlist gets the answer before anyone else does. The full answer, including what the subscription is for, goes here before launch — and it will be the same answer the App Store listing gives. Two different free-tier promises in two places is the single most damaging review theme in this category.
Does it work with the food I actually eat?
That depends on where you eat, and it is the thing most likely to be wrong at launch. A model trained mostly on Western restaurant plates will mis-read a night-market meal or a convenience-store lunch. If that is your food, say so on the waitlist form — it is how the training priorities get set.
When does it ship?
Submission is planned for December, which is deliberate: search demand for this category roughly doubles between August and January in every market we measured. The waitlist gets TestFlight before that.
Android?
Not at launch. iOS first, and honestly rather than as a placeholder — one platform built properly beats two built halfway.