I tracked my body in Excel at 12. Now I’m building the platform I wished existed.

I was a chubby kid. Not dramatically overweight, but enough that I noticed it, and enough that other kids noticed it too. I was about 12 or 13 at the time.

I also had a weird hobby for that age: I liked spreadsheets. Math class was the one subject I actually paid attention in. Numbers made sense to me in a way that most things didn’t.

The Excel summer

One summer I sat down at the family computer, opened Excel, and decided to fix the weight thing myself. I made a tracking sheet. Columns for everything: what I ate, how much I moved, my weight every morning. I built formulas that predicted where my weight would be in two months if I stuck to the plan. I set targets, calculated daily calorie goals, and made a chart that showed the predicted trajectory next to the actual numbers.

Then I tracked it. Every day, for the whole summer.

It worked. I lost the weight. But looking back, the spreadsheet itself wasn’t the interesting part. What actually mattered was that measuring forced me to pay attention. You eat differently when you know you have to write it down afterward. You keep going when the line on the chart is moving the right way. The data made the invisible visible, and that changed my behavior.

This was over 20 years ago. There were no Fitbit watches syncing to your phone over Bluetooth. No smart scales spitting out body fat percentages. You weighed yourself on a bathroom scale, squinted at the number, and wrote it down with a pen.

The 1% method, before I knew it had a name

I didn’t know it at the time, but what I was doing had a name. James Clear would later call it the aggregation of marginal gains, or the 1% method. The idea is simple: you don’t need dramatic changes. You need small, measurable improvements, tracked consistently, compounding over time.

A 12 year old with an Excel sheet was basically doing atomic habits before the book existed. The only tool I had was a spreadsheet and a bathroom scale, but the principle was the same: measure, adjust, repeat. Don’t try to change everything at once. Just make the numbers go in the right direction, day after day.

That mindset stuck with me long after that summer ended.

20 years later, the tools got better. The problem didn’t go away.

I never stopped tracking. The difference is that today, the devices are absurdly good. My Fitbit gives me sleep stages, heart rate variability, VO2 max, SpO2, breathing rate, skin temperature, active zone minutes. My Renpho scale measures a dozen body composition metrics every time I step on it. Weight, body fat, muscle mass, visceral fat, bone mass, BMR. Things I couldn’t have measured as a kid even if I wanted to.

But here’s what hasn’t changed: the data still lives in five different apps.

Fitbit has its app. Renpho has its app. If I ever add a Withings or Oura device, that’s another app. If I want to see how my sleep quality connects to my weight trend, or whether my HRV drops after heavy training weeks, I have to jump between screens and connect the dots in my head.

The devices got better. The central brain that ties it all together never got built.

So I’m building it

The project is called TONND. It’s an open source health dashboard you can self host.

You connect your Fitbit, your Renpho scale, and eventually other devices. Everything shows up in one view: weight, sleep, steps, heart rate, HRV, body composition, recovery score. It runs on Docker, you own the data, and the source code is on GitHub.

I use it every day. The Fitbit sync works, the Renpho integration is live, the dashboard shows real numbers. It’s already useful, but I’m thinking about this in three phases.

Phase 1: Collect (where we are now)

Get data from all your devices into one place. Display it clearly. Show trends over time. Calculate a recovery score from your HRV, sleep efficiency, and resting heart rate. This is working today.

Phase 2: Analyze (coming next)

Feed your health data to AI. Take your last 30 days of sleep, HRV, activity, and weight data, send it to Claude or GPT, and get back something actually useful. Not “drink more water” advice. More like: “Your HRV dropped 15% this week and your training load went up. You probably need a rest day before your long run on Saturday.”

The 12 year old version of me had to interpret his own charts. The 2026 version shouldn’t have to.

Phase 3: Coach (the long-term goal)

Have the AI built directly into the platform. No copy pasting into ChatGPT, no switching to another tool. You open your dashboard and the analysis is already there, personalized to your data, your goals, your history.

That’s the 1% method with modern tools. Measure everything automatically, analyze it with AI, and get specific recommendations based on your actual numbers.

Why open source, why self-hosted

Health data is personal. I don’t want my sleep patterns and body weight sitting on someone else’s server, being used to train someone else’s model. TONND is AGPL-3.0 licensed. You can read every line of code, run it on your own machine, and keep your data where you want it.

If the project dies tomorrow, your data is still in a standard PostgreSQL database on your own server. No lock-in. No export begging. No “we’re shutting down, you have 30 days.”

Come build this with me

The project is early. There’s rough edges everywhere and a lot left to build. More device integrations, better charts, the AI layer, mobile support. If you’re someone who obsesses over health data, likes building tools that solve your own problems, or just wants to contribute to an open source project that does something you’d actually use, come hang out.

Source code: https://github.com/hemati/tonnd

Discord community: https://discord.gg/3qmrFpwzpE

Live app: https://tonnd.com

The kid who made Excel spreadsheets about his weight would be pretty excited about where this is going. I still am.