01
Raw data is insert-only
Every payload is stored exactly as it arrived and never edited. A parser bug is fixed by re-running the parser over the original bytes — not by patching history until the chart looks right.
Cohort 001 · Day 26/90 · Intake closed
One subject. Ninety days.
Nothing averaged.
N of 1 is not a fitness app. It is the instrument built to run one single-subject experiment properly: every reading keeps the source it came from, every derived number can be recomputed from scratch, and no measurement is quietly smoothed into another one to make a nicer chart.
- Subjects
- 1
- Seats open
- 0
- Days per cohort
- 90
- Sources wired
- 7
Why the list is this short
There is one person on it, and that is the design.
A cohort is one subject at a time. Each one gets a protocol written for them, targets chosen for them, and a human reading the rows that land — which is exactly as scalable as it sounds.
Cohort 001 started on 2026-08-01 and runs 90 days. Its single seat is taken. Cohort 002 opens when 001 has been written up, and not before: an experiment that admits new subjects halfway through is an experiment that measured nothing.
So the honest state of this product is: full, working, and not for sale. The waitlist is how you hear when that changes.
What it reads
Seven inputs, one record, no guessing about where a number came from.
- WHOOPRecoveries, sleeps, cycles, workoutsRead-only OAuth against WHOOP's official API. Nothing is ever written back.
- Apple HealthSteps, heart rate, active energyPushed from the phone. A transport, never treated as the origin of a number.
- MacroFactorCalories, macros, weight trendExport file. Days the app restates later are corrected, not duplicated.
- DEXAFat mass, lean mass, visceral fat, per regionTyped in from the scan report. The reference measurement everything else is judged against.
- Blood panelsLipids, glucose, hormones, markersTyped in from the lab PDF, with the reference range that came with it.
- BIA scaleWeight, body-fat estimateKept as its own series at lower confidence. Never averaged into the DEXA line.
- By handCheck-ins, sets and RIR, meals, interventionsThirty seconds a time. If logging is slow, the log stops, and the experiment ends.
Connected data is read for one purpose: to show the subject their own experiment. It is never sold, never shared with advertisers, and never used to train anything. The privacy policy says so in detail.
Three rules that do not bend
02
Blank is not zero
An unmeasured waist is not a waist of zero centimetres, and an unlogged day is not a day of fasting. Empty stays empty all the way down, because one fabricated zero contaminates every average drawn through it.
03
Sources are never merged
A DEXA scan and a bathroom scale both report body fat. They stay two lines, each carrying its own confidence — the average of a good measurement and a bad one is a bad measurement that looks trustworthy.
Waitlist
Leave an address. Hear about it when a seat opens.
This is not a queue and there is no position to climb. When Cohort 001 ends, the next subject is chosen for the experiment that makes sense to run next — and everyone on the list hears about it either way.
What gets stored: your address, the one line you write, and the date. No tracking pixels, no analytics, no third-party mailer. Ask and it is deleted, same day.