Science
Selfra is built on real research.
Selfra is a consumer product, not a clinical trial. The thesis is built on decades of nutritional, longevity, and behavior change research. Here is some of it, and how it shapes what we ship.
01
The compounding effect of micronutrient adequacy on healthspan.
More than a third of adults in wealthy countries fall below the estimated average requirement for at least one essential micronutrient over any given month. These shortfalls are subclinical for years. They surface as fatigue, cognitive drift, and elevated long-term disease risk well before they appear on a standard panel. Adequacy held across decades, not heroics held across weeks, is what the longest-lived populations share.
Selfra is built to surface adequacy at the resolution it actually matters: daily.
02
Continuous nutrition signals versus episodic bloodwork.
Bloodwork is a snapshot. It reflects what your body has done recently, filtered through a few dozen markers, taken once a quarter at best. The intake that produced that snapshot is invisible to your clinician and usually to you. Continuous, passive nutrition data closes the gap. It is to bloodwork what a CGM is to a fasting glucose.
Selfra is the input layer that makes everything downstream of food legible in real time.
03
Mediterranean diets, Blue Zones, and what they actually share.
Across the populations that age best, the common thread is not a specific cuisine. It is structural: most calories from plants, protein consumed in moderation and from varied sources, ultra-processed food a rare event, meals shared and unhurried. The exact ingredients change continent by continent. The pattern does not.
Selfra is designed to recognize the pattern, not police the menu.
04
What people think they eat versus what they actually eat.
The underreporting literature is brutal. People underestimate caloric intake by 20 to 50 percent, with the gap widest among those most motivated to track. They forget snacks, undercount portions, and edit out the choices they regret. This is not dishonesty. It is the human memory doing its job. Passive observation collapses the gap.
Selfra removes the need to remember at all.
05
Why portion estimation from imagery has crossed the line into useful.
Vision models in 2025 and 2026 estimate portion volume from a single camera angle within a margin that is good enough to drive nutritional inference. Combined with longitudinal context, the cooking patterns of a particular household, the staples in a particular pantry, the error rate falls below the recall error rate of every food-tracking app on the market. The tool is finally good enough to be passive.
That is what makes Selfra possible now and not five years ago.
We will keep adding to this page as we publish our own research and as the field advances. Last updated January 1, 1970.