Health, Pills, and Rehab Tracking with AI
Health tracking is one of those problems that sounds simple until you actually need it. Take your pills on time. Do your physical therapy exercises. Track your symptoms. There are a thousand apps for each of these things, and every single one of them is either too complicated, too expensive, or too focused on gamification when all I want is a straightforward system that tells me what to do and records that I did it. So I built my own.
PillPopper3000: Medication Management
The name is ridiculous. The problem isn’t. If you take multiple daily medications, keeping track of what you’ve taken, what you haven’t, and what’s running low is surprisingly hard. The iPhone Health app handles basic reminders, but it doesn’t track refill schedules, doesn’t warn you about interactions with new prescriptions, and doesn’t give you the kind of history your doctor actually wants to see.
PillPopper3000 is a local web app (runs on localhost, data stays on my machine) that manages my medication schedule. It knows every medication I take, the dosage, the frequency, the prescribing doctor, the pharmacy, and the refill schedule. It sends me notifications when it’s time to take something, marks it as taken when I confirm, and flags missed doses. At the end of each month, it generates a compliance report I can share with my doctor.
The AI component is small but useful: it reads the text on new prescription labels (using OCR and a language model to parse the notoriously inconsistent formatting of pharmacy labels) and automatically adds the medication to my schedule. It also cross-references new prescriptions against my existing medications for potential interactions, using a medical interaction database. It’s not a replacement for a pharmacist, but it’s a useful early warning system.
Health Tracking: The Data Side
I track a handful of health metrics daily: weight, blood pressure, sleep quality, mood, and exercise. The tracking app is simple by design — a form with five fields that takes 30 seconds to fill out each morning. The value isn’t in any single data point; it’s in the trends over time.
The AI helps with the trend analysis. Once a week, it reviews the data and generates a plain-English summary: “Your blood pressure has been trending up over the past two weeks. Your sleep quality dipped last week, which correlates with increased screen time in the evenings. Your weight is stable.” It’s the kind of insight a doctor would give you if they had time to review your daily logs, which they don’t.
I built this with a simple SQLite database and a Python script that runs the weekly analysis through Claude. The whole thing is about 200 lines of code. It’s not fancy, but it’s consistent — I’ve been using it for over a year, and having that longitudinal data has been genuinely useful in medical appointments.
Knee Rehab: The Exercise Tracker
After knee surgery, my physical therapist gave me a sheet of paper with twelve exercises, each with a specific number of reps, sets, and hold times. The instructions were clear. The problem was compliance. PT exercises are boring, they hurt, and it’s extremely easy to convince yourself that you did three sets when you actually did two.
The knee rehab app is a timer-based exercise guide that walks me through each exercise in sequence. It tells me what to do, counts the reps, times the holds, and logs everything. No way to cheat, no way to forget which exercise comes next, and a complete record of every session that I could show my PT.
The AI’s role here was in the initial setup. I described each exercise to Claude (or showed it the PT’s instructions), and it generated the exercise configuration with appropriate timing, rep counts, and rest periods. When my PT adjusted the program (fewer reps of this, longer hold on that), I told the AI what changed and it updated the configuration. It saved me from manually editing timer configs, which is not how I want to spend my limited post-surgery energy.
Why Personal Health Tools Matter
The common thread across all three tools is ownership. My health data lives on my machine, not in someone else’s cloud. I control what’s tracked, how it’s analyzed, and who sees it. I’m not generating training data for an ad-tech company or feeding my medication list into a system that might sell it to insurance brokers.
This matters more than most people realize. Health data is some of the most sensitive information you have, and most health tracking apps have privacy policies that would make a surveillance state blush. Building my own tools means I get the functionality without the data extraction.
It also means I get exactly the features I need and none of the features I don’t. I don’t need social features, achievement badges, or AI-generated meal plans. I need a thing that tells me to take my blood pressure medication at 8am and records that I took it. Commercial health apps are built to maximize engagement. My tools are built to minimize the time I spend thinking about health tracking so I can spend more time actually being healthy.
The Bigger Picture
These three tools together represent something I believe strongly: AI is most valuable when it’s applied to personal, specific problems. A generic health tracking app has to serve millions of users with wildly different needs. My tools serve exactly one user — me — and they do it perfectly because they were designed for my specific medications, my specific health metrics, and my specific rehab program.
This is the future of personal software, I think. Not apps downloaded from a store, but tools built by (or with) the person who uses them, tailored to their exact needs, running on their own hardware, with their data under their own control. AI makes this possible for people who aren’t professional developers. You don’t need to know Python to describe what you need to an AI and iterate until it works.
PillPopper3000 isn’t going to win any design awards. But it’s kept me on track with my medications for over a year, which is more than any commercial app managed to do. Sometimes the best software is the software you build yourself.
