Weak letter/word tracking + targeted practice mode

[GUNPARK-GOOKIM]

Right now Monkeytype tells you overall wpm/accuracy but not which letters or words are actually costing you speed. I'd like to propose:

• Track correct/incorrect counts per character across tests (persisted with account stats like everything else already is)
• Show it as an opt-in keyboard heatmap + weakest-letters/words list, tucked away from the main test UI
• Add a practice mode that biases word selection toward your weak letters/words instead of pure random

I built a small prototype validating this (per-keystroke scoring, heatmap, biased practice mode):[ [repo link: (https://github.com/GUNPARK-GOOKIM/keystroke-lab)]]

Open questions: would this extend the existing profile stats or be its own mode? Any concern it works against the minimalist design? Should letter data be stored server-side from day one, or is client-side a fine first cut?

Not attached to an implementation — mainly checking if this is a direction you'd want before I put time into a PR.

[GUNPARK-GOOKIM]

here is my project link https://keystroke-lab-psi.vercel.app/

[fehmer]

this is planned

[GUNPARK-GOOKIM]

you can checkout my repo and project i have already implemented that