Our Diary Was Built to Total Calories. It Was Never Built to Catch a Deficiency

On September 4th, Citi’s read of IQVIA prescription data showed both oral GLP-1 pills on the U.S. market posting their best week yet: Novo Nordisk’s Wegovy pill at roughly 181,000 total prescriptions, up 3% week over week, and Eli Lilly’s Foundayo at about 44,000, up 8%. Neither number is enormous next to the injectable versions of these drugs, but the direction is the story. A pill you take once a day instead of a shot you give yourself once a week is the version of this drug class built for mass adoption, and mass adoption is exactly what the trendline shows.

What a pill like that actually does is suppress appetite hard enough that people eat a lot less, for a long time, often indefinitely, since these are drugs you generally stay on rather than a course you finish. A narrative review published August 23rd in Nutrients, titled bluntly “Hidden Malnutrition in the GLP-1 Era,” lays out what that does to the rest of a person’s diet once the calorie number gets small enough. Some patients are eating under 800 calories a day in the early weeks of treatment. At that intake, hitting a normal day’s worth of protein, vitamin D, iron, B12 and calcium from food alone gets genuinely hard, and the review’s read of the evidence is that a lot of people aren’t managing it: reduced energy intake drags down protein and micronutrient adequacy right alongside body fat, and the muscle loss and nutrient gaps that follow were never supposed to be part of the deal.

That’s not a hypothetical sitting in a lab somewhere. A retrospective study of insurance claims from 461,382 adults newly prescribed a GLP-1 receptor agonist, published back in June 2025, found a diagnosed nutritional deficiency in 12.7% of patients within 6 months of starting and 22.4% within 12 months, meaning the longer someone stays on the drug (again, the intended use), the more likely they show up with a real, coded deficiency. Vitamin D was the single most common one, 7.5% at 6 months rising to 13.6% at 12, with B-vitamin, calcium, selenium and zinc deficiencies climbing right alongside it. This is a population that is, by design, eating less of everything, and it is already showing up in claims data as a population eating too little of specific things.

So I checked what our own diary would tell that person. Every food source in Mr BITE, barcode scan, photo scan, manual entry, voice log, recipe, resolves down to the same handful of fields: calories, protein, carbs, fats, and optionally fiber and sugar. That’s the entire nutrient model. There is no vitamin D field anywhere in our schema. No B12. No zinc, no selenium. Calcium and iron exist, technically, but only as spoonacular_iron and spoonacular_calcium, two optional numbers that get populated exclusively when a recipe is imported from Spoonacular, shown as two extra rows on that single recipe’s detail screen when they happen to be present. They don’t get carried into your log when you eat the recipe. They don’t get summed into a day. They don’t have a target, a coach comment, or a color when you’re short. A user four months into Foundayo, quietly on their way toward that 13.6%, could log a technically flawless week of calories and protein through Mr BITE, and the app would have no way of knowing whether they were also getting enough vitamin D, because it never asked.

I want to be honest about how big a gap this actually is to close, because it isn’t the one-line fix a missing NOVA field was a couple of issues back. Open Food Facts already had a nova_group field sitting in the exact response we already parse; asking for it was a rounding error of effort. Vitamin D, B12, iron and calcium aren’t a free field on an API call we already make. Coverage is inconsistent across our own data sources, USDA carries a good amount of it, Spoonacular carries some, a photo scan carries essentially none, and turning any of it into something a user can act on means a real daily target, not a decorative row on a recipe page nobody reopens. I don’t have that feature built or scheduled, and I’m not going to pretend otherwise in a paragraph like this. What I do have is a plain answer to a narrower question: does the app know, right now, whether any of our users are even taking a GLP-1 in the first place. It doesn’t. There’s no field for that either.

BITE of the Week: The Portion Dial We Previewed Last Issue Just Went Out the Door

Version status moved since last issue. 1.12.0, iOS build 107 and Android versionCode 138, was what shipped as of Issue 18. Two commits landed since, both bookkeeping around a single real change. On September 5th, main bumped to 1.12.1 and cut a new native build, iOS 108 and Android 139, rather than push it over the air, even though RELEASE_NOTES.md logs it plainly as “JS-only; no native module or SDK changes,” and the OTA runtime number held at 1.0.3. It’s the client-side half of the photo-sheet change I covered as still pending in Issue 18: the empty amount-in-grams row is gone from a fresh photo scan, the portion count is now the one control and takes decimals, and half a plate now saves and reopens as 0.5 instead of quietly rounding itself up to a whole plate the moment you look away.

Riding along in the same commit is a fix that never made it into last issue’s writeup: a fresh photo capture used to silently inherit the serving size left over from whichever entry you had open just before it, so scanning a new plate right after logging a “2 x 200g” chicken breast could hand your new meal that same 200g starting point for no reason. That’s cleared on capture now, so every fresh scan starts from a clean slate instead of whatever you looked at last.

As of this writing, 1.12.1 has EAS-assigned build numbers and finished store copy sitting in the repo, but nothing in it confirms the build is actually live on either store yet, unlike 1.12.0, which the same file explicitly marked confirmed via the release scripts’ dry runs. So if you haven’t seen the new stepper on your phone, that’s most likely why: it’s built and queued, not out the door.

Three Quick Bites

1. Researchers in Japan found that pterostilbene, a polyphenol naturally present in blueberries, grapes and other berries, reduced abnormal fat buildup in cultured mouse muscle cells by boosting fat breakdown and helping stabilize a protein central to fatty-acid metabolism, according to a study published September 3rd in Food Bioscience. It’s cultured cells, not a living animal or a person, so this is an early mechanistic signal rather than a reason to eat more blueberries today, but it’s a specific, testable lead for a class of compound that’s cheap and already in the food supply. ScienceDaily

2. LatAm-FINGERS, a two-year, Alzheimer’s-Association-funded trial across 11 Latin American countries, found that structured lifestyle coaching, four days a week of supervised exercise plus personalized MIND-diet counseling adapted to local culture, improved cognition 55% more than basic health advice alone in older adults already at risk of dementia, with results published in The Lancet. The size of the gap between structured coaching and a pamphlet is the real finding here: knowing what to eat and being coached through actually doing it are not the same intervention, and this trial put a number on how far apart they are. ScienceDaily

3. A September 3rd study in Frontiers in Nutrition describes My Dietary Fiber Finder, a lab platform that uses robotic automation and machine learning to test how a person’s own gut bacteria ferment dozens of different fibers, reporting that pairing inulin with two specific bacterial strains, Bacteroides uniformis and Anaerostipes caccae, reliably produced the highest beneficial butyrate output. It’s a proof-of-concept screening tool, not a consumer product, but it’s a real step toward “which fiber is right for your gut” being an answerable, individualized question instead of a guess. Frontiers in Nutrition

Tool of the Week: USDA FoodData Central

Since this issue is about the micronutrients our own diary doesn’t track, this week’s pick is the free database that already has them. USDA FoodData Central is the federal government’s own nutrient database, free, no account required, and it carries vitamin D, B12, iron, calcium, zinc and dozens of other nutrients most consumer apps, ours included, quietly leave out. Search any food or barcode-linked item and you’ll get the full micronutrient panel, sourced and lab-tested, the same underlying data a lot of the apps you already use are drawing from behind the scenes. If you’re on a GLP-1 and want to sanity-check your own week for something more than calories, this is where to actually look. USDA FoodData Central. Free, government-run, no signup required.

Thanks for reading. If you’d rather a coach that will eventually ask about more than calories, Mr BITE is free on iOS and Android.

Collins, building Mr BITE