It's like Wordle, but instead guessing a word and being given the positions of the correct letters, you are given the "similarity" between your guess word and the target word.

The similarity is calculated using the cosine similarity between the GloVe embeddings of the guess and target words. The similarity value is between -1 and +1, where higher similarity is better, and similarity is 1 when the guess word equals the target word.

Valid words are the most common 50,000 English words with a GloVe embedding. The target word is randomly chosen from Wordle's list of target words that are in the most common 50,000 English words with a Glove embedding.

Inspired by Semantle, which uses word2vec embeddings instead of GloVe embeddings.