The fun slot: termdle

Not every project in a skill-building sprint has to be infrastructure. termdle is the “for fun” slot — Wordle, in the terminal, 5-letter word, 6 tries, colored feedback. A portfolio that’s ten shades of infra reads a little joyless; one genuinely playful build earns its place.

The one place “fun” still needed to be correct

Wordle’s classic edge case: guess a letter twice when the answer only contains it once. Naive scoring double-credits the guess — marking both occurrences as “present” when only one letter actually exists to match against. Getting this right (crediting exactly as many occurrences as the answer actually contains) is a small algorithm with a surprisingly easy way to get subtly wrong, and it’s covered by an explicit test rather than left to “looks right when I tried it.”

Testable by construction

play() takes injectable input/output streams instead of hard-coding stdin/stdout, specifically so the entire game loop — not just the scoring function — can run end to end against fake streams in a test, no real terminal required. That’s a small design choice with an outsized payoff: a winning game and a full 6-guess loss (revealing the answer) are both covered by tests that actually drive the CLI, not just unit tests around scoring in isolation.

What it doesn’t do

Full test suite green, scoring algorithm and CLI loop both covered end to end. No real dictionary check on guesses (any 5 letters go through) and no stats persisted between runs — this was meant to stay small and fun, not turn into a real word game with a backlog of its own.

Code: termdle — and yes, you can npm run play it yourself.