Information theoryWordle Cyborg
A Wordle Clone With an Entropy-Powered Cyborg Sidekick
A full-stack Wordle clone with a Cyborg Mode that suggests the best next guess using Shannon entropy.

Rochester, N.Y. First I built a full Wordle clone, with tile flips, shakes and a 10,656-word list of valid guesses. Then I gave it a cheat code. Turn on Cyborg Mode, the board goes red and an Ask Cyborg button appears. It sends your guesses and tile colors to a Flask endpoint that suggests the next word.
The solver filters the 2,314 possible answers down to the ones consistent with everything you've seen, then picks the guess whose color patterns spread those answers out the most, measured by Shannon entropy. Duplicate letters are the easy thing to get wrong, so feedback is scored in two passes: greens first, then yellows from what's left. The first suggestion is always SALET. I mocked up the UI in Figma before writing any code.
How it works
Wordle Cyborg, start to finish, in 4 steps.
- Step 1:
History
Guesses and tile colors sent to Flask
- Step 2:
Filter
Keep answers consistent with every clue
- Step 3:
Score
Entropy of each guess's color-pattern buckets
- Step 4:
Suggest
Best guess, or the last few candidates
Graphic: The Sahil Bachu