01/15/2026
This is totally unrelated to hammers, but here's a nerdy rabbit hole I've gone down...
I love to ride my bike, and I also love playing with data. My Garmin bike computer displays gradient, but it's frustratingly inaccurate. The problem? It relies on GPS (slow, only accurate to maybe 10 meters) and a barometric altimeter (imprecise and unstable).
So I thought: why not measure the tilt of the bicycle frame directly?
I spent a significant part of my career doing instrumentation and measurement, so I started experimenting with tilt meters. A simple bubble level won't work—any acceleration when you speed up or brake throws it off completely.
First, I tried a tiny Bluetooth tilt meter taped to my top tube (WIT Motion WT9011DCL—under $20 from China). Super accurate and stable, but I couldn't keep my phone awake on long rides to watch the data in real time.
So I built my own standalone display and logger with a touchscreen, using a Waveshare ESP32-S3 development board (about $30). Here's the thing: I don't write code. The only programming class I ever took was 50 years ago using punch cards. But with AI help, I wrote and compiled 1,000 lines of code to make this work.
The device measures slope using gyroscopes with corrections from accelerometers. The gyros measure rotation rate, which you integrate over time—but integration amplifies errors. That's where Kalman filtering comes in, a probability-based method for correcting telemetry that's completely new to me and absolutely fascinating.
I've got it working and logging data to a micro SD card. There's still snow and black ice everywhere, so no road test yet, but bench testing looks great.
The hardware cost: $30-50. The software, based on my invested time: roughly a million dollars. Software always costs more than hardware.