D/W SEQUENCE LAB On your device. Not on a server.
01

Set the search

Search settings

The word to avoid. Never sent anywhere.

A timed search returns the best found, unless it proves optimality sooner. Exhaustive searches may take a very long time.

Already have a starting sequence?

Up to six comma-separated words, in order. They count toward the six guesses and must obey all earlier feedback. The engine checks legality.

C++ search engine → WebAssembly.
Runs in a background worker; Stop ends it immediately.

02

The sequence

READY

Six turns. No undos. The answer stays unguessed.

FINAL SCORE—
—valid words left
—unused letters

Ready when you are. Enter the known answer to find a six-guess sequence.

Six-guess sequence with feedback, remaining valid words, unused distinct letters, and score after each guess
TurnGuess / feedbackValid
words		Unused
letters		Score
Search time —Nodes —Score ceiling —

WHAT WE'RE MAXIMIZING

More possibilities.
Fewer letters spent.

The objective is final valid words × unused distinct letters, measured after all six guesses—not the sum of the row scores. Every guess must remain legal under the feedback so far, and cannot be the known answer.

Optimal means the search proved no better sequence exists. A time limit or manual stop gives an unproven best known sequence, if one was found. With a fixed prefix, any proof is conditional on that exact prefix. The score ceiling is an upper bound, not a promised result.