The Right Protein Number Depends on Whether You Move

On July 31st, researchers at the University of Wisconsin-Madison published a review that went through more than 350 studies, in humans, mice, insects and yeast, and landed on a conclusion I didn’t expect to read this week: for most adults, eating less protein than they currently do would probably serve them better. Bailey Knopf and Dudley Lamming’s paper, “The hallmarks of protein and amino acid restriction in aging and longevity,” found that lower protein intake tracked with better metabolic function, less inflammation, less cellular damage, and activation of the same pathways that show up in essentially every serious longevity intervention we know about. The mechanism, as far as the review can pin it down, runs through specific amino acids, methionine, isoleucine and valine, which in excess seem to push the body toward growth-signaling pathways that are useful when you’re young and building muscle and considerably less useful when you’re older and mostly sitting at a desk. In one mouse study, isoleucine restriction alone added 33% to lifespan.

Here’s the line from Lamming that actually stopped me: most people are relatively sedentary, and a lot of people are eating more protein than they need, probably with real costs. I build a food-tracking app. Mr BITE sets you a protein target on day one, computes it off your bodyweight and your goal, shows you a ring that fills up as you log meals, and treats hitting that number as an unambiguous win. That’s not a fringe design decision, it’s close to industry standard, every tracker I’ve used does some version of it. And the review is a pretty direct challenge to the assumption sitting underneath it: that more protein, within reason, is basically always fine, so a generous target is a safe target.

It isn’t a clean “protein is bad” story, though, and the review doesn’t pretend it is. The benefit of eating less shows up specifically in sedentary people. Flip the activity level and the relationship flips with it: people who train, especially against resistance, use that protein to build and hold onto muscle, and higher intake helps them do it. Older adults trying to stave off sarcopenia are their own case again, protein is one of the few levers that reliably helps there, sedentary or not. So the honest read isn’t “lower everyone’s target.” It’s that a single number computed from bodyweight and a generic activity multiplier is standing in for at least three different situations that call for different advice, and the app can’t tell which one it’s looking at.

That’s the part worth sitting with as a builder, not just a reader. A protein ring is an easy feature to justify: it’s legible, it’s motivating, closing it feels good, and “eat enough protein” has been safe, uncontroversial advice for so long that nobody questions the target underneath the ring, only whether you hit it. But if the number itself is wrong for a given person, sedentary and already eating plenty, then the app is spending its one daily nudge convincing someone to do more of something that isn’t helping them, while calling it a win.

The fix isn’t dramatic. It’s asking the app to actually use the activity data it already collects, steps, logged workouts, whatever signal it has, to set a target that means something different for a runner than for someone who hasn’t left their desk today, instead of running everyone through the same bodyweight formula and calling the output personalized. That’s more work than a static target, and it’s a worse-looking spec on a roadmap doc than “protein ring,” which is probably why it isn’t the default anywhere yet, including in the app I work on. But a target that’s confidently wrong for a lot of the people looking at it isn’t personalization, it’s a guess with a nice UI around it, and I’d rather build the thing that’s actually looking at the person in front of it.

BITE of the Week: 1.5.0 Is Live, and a Day Spent Finding Every Way a Number Could Lie

Since Issue 3, BITE Pro actually cleared app review: 1.5.0 (iOS build 97, Android build 128) is the version in your hands right now if you’ve updated, carrying everything Collins described last issue, the redesigned logging sheet, the recipe-forking rule, the second coach revision. Main has already moved past it: 1.6.0 (iOS build 98, Android build 129) is cut and carries the next merge, recipe-only meal plans and shopping-list parity with what you’ve actually planned, plus a pile of library fixes. It’s queued for the next store build, not in your hands yet.

The bigger story of the week didn’t ship a version number at all. It was a single-day audit on August 4th that went hunting for every place a logged number could be quietly wrong, and found more than I’d like to admit. A barcode scan logged at more than one serving was saving the product’s per-100g macros inflated by exactly the portion count, permanently, so every future re-log of that product carried the error forward. Typing a decimal into the voice or photo card’s macro fields round-tripped through parseFloat on every keystroke, so “12.5” landed as “125” and clearing a field snapped it to zero. And the app’s own gram-conversion table had a teaspoon at 15g and a tablespoon at 25g on a scale where a cup is 240, three times and 1.7 times too big, so any spoon-measured ingredient logged through that path carried a real error into its macros. All three fixes are merged to main now, none of them are in 1.6.0’s build, they’ll ship with whatever comes after it.

I don’t love that any of these existed. I do think finding three separate ways a precise-looking number could be silently wrong, on the same day, from the same audit, is a decent argument for running that audit more often than we have been.

Three Quick Bites

1. A 350-paper review says most sedentary adults are probably eating more protein than helps them. University of Wisconsin-Madison researchers Bailey Knopf and Dudley Lamming synthesized decades of human and animal research and found that restricting protein, and specifically the amino acids methionine, isoleucine and valine, improved metabolic markers and extended lifespan across multiple species; isoleucine restriction alone added 33% to lifespan in one mouse study. The catch, and it’s the whole essay above: the benefit was strongest in sedentary subjects, not people who train. Published July 31st. Cell Press Blue

2. A 3-year trial found a plain daily multivitamin may help older adults stay independent. A secondary analysis of the COSMOS trial, presented at NUTRITION 2026, followed more than 16,000 adults aged 60 and older and found the group taking a daily multivitamin-mineral supplement reported meaningfully better physical functioning and lower symptom burden after three years than the placebo group. It’s preliminary and not yet peer-reviewed, and researchers are clear it’s not a substitute for diet or exercise, but it’s a notably boring, cheap intervention showing up in a large trial. ScienceDaily

3. GLP-1 drugs are now linked to a real, if small, hair-loss risk. A University of Pennsylvania-led study published in The BMJ compared more than 12,000 GLP-1 starters against similar patients starting other diabetes drugs and found a 37% higher rate of non-scarring, reversible alopecia diagnoses versus an SGLT-2 inhibitor, and a 68% higher rate versus a DPP-4 inhibitor. Absolute risk stays low, but it’s one more side effect worth knowing about if you or someone you’re tracking for is on one. ScienceDaily

Tool of the Week: Open Food Facts

This week’s pick is the open database sitting quietly underneath a lot of the barcode scanning this issue’s bug hunt was about. Open Food Facts is a free, nonprofit, crowdsourced database of millions of food products worldwide, ingredients, allergens, nutrition facts and a Nutri-Score, all contributed by volunteers and available with no login wall through the website or a free API. For readers: it’s a fast way to check what a barcode scan actually told an app, ingredient list included, which most proprietary databases won’t show you. For builders: it’s the same open data OpenNutriTracker (Issue 2’s pick) draws from, and it’s a good reference for what a barcode-to-macros pipeline should look like, and for what not to hardcode: gram-per-unit tables like the one that bit us this week are exactly the kind of thing worth double-checking against a real, actively maintained source instead of quietly wrong constants. openfoodfacts.org. Free, open data, nonprofit.

Thanks for reading. If you’d rather your protein target actually knew whether you moved today, Mr BITE is free on iOS and Android.

Michael, building Mr BITE