Martin Mojžíš

I m a solo dev, and I recently ran into a massive data architecture problem while building a personal tracking tool.

I originally tried using standard habit and symptom trackers, but I realized they almost all fail when dealing with overlapping data: situations where one event or routine completely masks the effect of another.

If Input A causes a negative reaction, but I immediately apply Treatment B to cover it up, the raw data simply logs that "Input A led to zero issues." The app basically lies to the user because it treats every log in isolation and can't separate the overlapping variables.

To solve my own problem, I had to stop using generic trackers and build a statistical model (using Elastic Net regression) in the background to calculate true correlations and filter out the noise.