Lexicle
How I Built Lexicle – The Word Game Based on Meaning
I’ve always loved daily word-relation games, but found them too hard to play because of how unexpectedly the words relate. Most rely on something called word2vec, which measures how “close” words appear next to each other in text, but not how similar they are in meaning. This creates maddening moments (like in a recent Proximity puzzle) where the target word was “firefly” and the closest words to it are “owl” and “frog” rather than “insect,” “bug,” or even “fly.”
My partner commented that if someone could solve that problem and make the game more fun to play, it could be really cool…
So I got to work. 🙂
A New Way to Relate
After ruling out word2vec for good, I found my way to Sentence-BERT, which is a framework used in sorting and filtering, and is great at understanding the semantic relationships between sentences. I tweaked it to process individual words instead – specifically every 5-letter word in English minus offensive and unusable words – and got a result that was about 90% of the way there. For the remaining 10%, I decided to build a manual review form where my partner and I could examine each word relationship one by one…
This was… not my most efficient idea 🤦🏻♂️
Secret Sauce
After realizing the pain of manual review, I remembered AI exists and built a script to send each word pair, one by one, to an AI endpoint for review. My prompt asked it to “assess how closely the meanings of the two words align, overlap, or connect in real-world usage, definitions, or typical contexts.” (The full prompt is much longer and took dozens of iterations to dial in.)
Though this wasn’t cheap – each API request costs money – it was efficient and worth it for the peace of mind of having words truly relate to each other.
Putting It All Together
With words now related, I set out to make the game more accessible by offering a free “warm” one to start, along with a free one for every guess. This prevents frustrating dead ends, and allows me to add a negative constraint – 11 guesses or bust.
I also quickly learned I needed to reveal at least one letter of the mystery word for the game to be playable, so with subsequent guesses the letters reveal. You get three shown by the final guess. I have seen people get the word in just one guess – 10/10! – but there’s a bit of luck involved, though not as much as Wordle.
Also, all Lexicle puzzles will never repeat, so if you’ve played the archive, you can rule out many words. 😉
Let me know if you find a word that is missing and should be included or removed! I always welcome feedback, and can be reached via the contact page or in the comments below. The game is only as good as its words after all.
Happy Playing,
Matt