I Built the Perfect Chess Training Tool... And Immediately Ghosted It

I did what everyone says. Puzzles. YouTube. Opening prep. My rating moved maybe 50 points in either direction and always came back to the same place. I wasn't improving. I was just busy.

The thing that bothered me most was that none of the training was about me. The puzzles had nothing to do with the mistakes I actually make. The openings I studied weren't the positions I was losing from. It was generic content for a generic player — and I am not a generic player. I'm a specific one, with specific bad habits, losing in specific ways.

So I built something.

rt2k — you upload your PGN, it runs Stockfish locally in your browser, detects your recurring mistake patterns, and generates puzzles from your actual games. No account. No server. Nothing leaves your machine.

The idea was simple: stop training on random positions. Train on the positions that are actually killing your rating.
I finished building it. Deployed it. Told myself this would finally move the needle.

Then I never opened it again 🤣

I'm not sure what that says about me. Maybe the building was the point and the using was always going to feel like work. Maybe I don't actually want to see what my games reveal about my chess. Maybe I already know the answer and don't want it confirmed.

I'm posting this because I genuinely don't know if the tool is useful or if my whole premise is wrong. Does training on your own mistakes actually work better than generic puzzles? Or is there a reason nobody has done this and just stuck with Lichess practice?
Try it if you want. Tell me what you actually think. I can handle it.

link: rt2k.pages.dev/analyze