Signal / Human Systems

MATTHEW
WALKER

I built an AI operating system for my body.
19 data sources. 87 intelligence tools.
One hypothesis: compounding data changes everything.

Read the experiment See the platform
// live vitals — loading...
Weight
— lbs
—
HRV
— ms
—
Recovery
— %
—
RHR
— bpm
—
Journey progress—
Zone 2 this week—
Weekly loss rate—
19 DATA SOURCES 87 AI TOOLS WEIGHT 287.7 LBS HRV 54MS ↑ IMPROVING GOAL DATE 2026-07-31 RECOVERY 75% LAMBDAS 42 PHASE IGNITION 19 DATA SOURCES 87 AI TOOLS WEIGHT 287.7 LBS HRV 54MS ↑ IMPROVING GOAL DATE 2026-07-31 RECOVERY 75% LAMBDAS 42 PHASE IGNITION
The experiment

I AM THE
EXPERI
MENT

Started Feb 9, 2026

Most people track their health with one or two apps. A step counter. A calorie tracker. Maybe a sleep watch. The data lives in silos, never talks to each other, and generates the same generic advice regardless of what's actually happening.

I wired everything together. 19 sources feeding a single database — wearables, continuous glucose, lab results, genome variants, DEXA scans, daily journal. Then I built an AI with 87 tools to query it in natural language.

"When you instrument your body the way you'd instrument a production system, you stop guessing. The data doesn't lie — but it does consistently surprise you."

This site is the lab notebook. The health transformation is real. But it's the proof of concept, not the product. The product is a new model for how humans and AI can work together on problems that are personal, longitudinal, and uniquely yours.

Lost
14.3 lbs
Remaining
102.7 lbs
Current
287.7 lbs
Goal date
Jul 31
START 302 LBS NOW 287.7 ← 12.2% GOAL 185 LBS
Now
287.7 lbs — Active
14.3 lbs lost. HRV improving. Zone 2 target exceeded this week.
May 5
250 lbs — Ignition
Sleep apnea risk drops substantially. Walking pace improves naturally. 50 days out.
Jun 14
220 lbs — Push
Zone 2 becomes accessible at real workout intensity.
Jul 11
200 lbs — Grind
Visceral fat normalization target in range.
Jul 31
185 lbs — Goal
FFMI enters athletic range. 117 lbs total. Chapter 1 ends.
// data sources — 19 active streams
Whoop 4.0
Withings Body+
Strava
Apple Health
MacroFactor
Eight Sleep
Garmin
Hevy
Habitify
Notion Journal
Todoist
Dexcom CGM
Lab Draws ×7
DEXA Scans
Genome 110 SNPs
Blood Pressure
State of Mind
Weather
Supplements
The platform

Not a product.

A custom AWS system built on Claude. Proof of concept that lives in production and gets smarter every week.

87
MCP tools
Natural language queries across every data point. Ask Claude anything about 18 months of longitudinal health data.
19
Data sources
Wearables, labs, CGM, genome, DEXA — fully automated ingestion every morning before 11 AM.
7
AI emails / week
Daily Brief, Digest, Chronicle, Compass, Nutrition Review, Weekly Plate, Monthly Letter — all written by Claude from live data.
42
AWS Lambdas
8 CDK stacks. EventBridge scheduling. Fully automated. Runs every morning without intervention.
23
Correlations / week
Pearson matrix every Sunday with FDR correction. Feeds the hypothesis engine that tests ideas about my biology.
A
Architecture grade
Reviewed 16 times by a 12-member AI board of directors. Current grade: A. All systems nominal.
Deep dive into the architecture →
Latest writing

The lab notebook.

// First post coming at 250 lbs — Ignition milestone
// Second post: the architecture deep dive
All writing →