Seven days of meal photos. The app was short 800 calories a day.

Today’s letter

The one thing your dinner can tell you that the app cannot

For years I repeated a line I had read in a dozen places. People underreport what they eat. They forget the oil in the pan, the handful from the bag, the second pour of wine. I said it the way you say something obvious… and I never once asked how anybody knew.

Then I read the study that had measured the thing I kept assuming. Twenty women photographed every meal for seven days into an AI app, the kind that reads a plate from a picture. The same week, researchers tracked what each woman’s body actually burned using water with a tracer in it, and in a week where weight holds steady, that figure stands in for what she ate.

“So I was lying to myself?”

You are not careless about this. The chicken got weighed. Every plate got photographed before anyone was allowed to eat, and the coffee went in with the milk counted. Some of those days you felt awful, and you logged anyway.

Then the scale sat still for three weeks, and the advice was the same everywhere you looked. If the deficit is not working, you are eating more than you think. Be honest with yourself. So you cut a little further, logged a little tighter, and started treating your own memory as the suspect.

Here’s what gets me.

You were not the unreliable part of that arrangement.

Ask your doctor for this heart test♥️

There's a test that could tell you exactly what's happening inside your arteries right now.

Doctors used to order it routinely.

Not anymore.

It quietly disappeared from standard checkups.

Why?

Because what it reveals could change how millions of people think about heart health… and put billions in heart medication sales at risk.

You won't get this test at your next physical.

But you can still find out what it checks for and what to do about it.

=> See the test doctors stopped ordering..and what they don’t want you to know

P.S. If your last checkup came back "normal," that doesn't mean your heart is safe. This test checks something most panels completely miss.

Let me show you how I know. Not with my opinion, but with the study.

Researchers at University Hospital Zurich took twenty women living with obesity and ran them through one ordinary week. Every meal went into SNAQ, an app built on the same idea as the ones on your phone: point the camera at the plate, let the software name the food and estimate the portion.

At the start of that week each woman drank doubly labeled water, the tracer method nutrition science uses as its reference standard. It reports how much energy a body spent across seven days, measured rather than remembered. Then the two sets of numbers were laid side by side.

What they found

Validation study, 2026. Serra et al., npj Digital Medicine.

The app’s total (817 short a day)

Against the measured figure, the app ran about 25 percent low, an average gap of 817 calories a day. The older method, asking a woman to recall yesterday’s food, ran about 50 percent low.

The tracer (measured, not recalled)

Measured energy showed no significant relationship with what either method reported, and the app’s day to day stability across the week came out at 0.00, where 1.00 would be a perfect match.

A group average may be off by hundreds. One woman’s daily number, on this evidence, may carry almost nothing.

In other words, the app was not simply low for everyone by the same amount. For one woman on one day it could fall short by thousands, for another it could run over, and nothing in the number told you which had happened.

Her honesty had nothing to do with that gap. Neither does yours.

So why would a camera and a database struggle this much with a plate of food?

A photograph carries no weight and no density. The software has to name the food, guess the volume, then look the numbers up, and all three steps can go wide: the oil the pan soaked up, the food sitting behind other food, the drink in the cup. In July the NIH ran a harder version of this test, photographing 102 meals cooked in a research kitchen where ingredients are weighed to a tenth of a gram, and four popular apps came back about a third low, with most of the miss sitting in fat. That one was presented at a conference and has not cleared peer review, so hold it loosely. Both sets of results point the same way, and the Zurich team wrote plainly that the answer cannot be to hand the error back to the women using the app.

She never stood a chance of making that number honest. Neither did you, I’d bet.

That’s the part no one says out loud… the software was allowed to be wrong by hundreds, and you were the one asked to explain the gap.

Try this today

Tonight, leave the app closed for one meal. Eat the dinner you were already going to eat, and add one thing to it: a boiled egg, a spoonful of beans or lentils, a handful of almonds, or a cup of berries. Take nothing off the plate. Then note the time you finish, and note the time hunger actually comes back. Do the same tomorrow with the same addition.

Your own clock is measuring something. The app was only ever estimating.

Promise me you’ll try it today? Saying it out loud is half of what makes it stick.

I’m in. I’ll try it today →

This is not magic, and it is not a new rule to follow. A meal carrying protein and fiber tends to hold longer, and how long it holds is something your body reports back to you, for free, every day of the week.

I got there late. For years I told people the gap was memory, and part of the gap was the instrument the whole time.

There’s still time for you. The number was handed to you with instructions to trust it over your own hunger, and you are allowed to stop doing that today.

You were never the unreliable measurement here…

It was never your fault.

— Cole

P.S. How many different food tracking apps have you downloaded over the years? Just the number, hit reply. I’ll tell you tomorrow where that sits against everyone else.

Ask your doctor for this heart test♥️

Keep Reading

Cutting the trans fat earned the label. The saturated fat paid for it.

The exercise barely moved the scale. Six pounds came back without it.

180 people cut sweetness for six months, and not one lost the craving

Four days by a window, and the blood sugar held steadier

8am and 5pm, same doctor. The orders came out sixteen points apart.

They missed days. Twelve weeks later, it made no difference.