About Hot & Cold
Hi, I'm Alex
I built Hot & Cold because I wanted a word game that tested understanding rather than spelling. Wordle is brilliant at the letter-level puzzle, but I kept wondering: what would it feel like to navigate meaning instead?
The first prototype was a weekend hack in late 2025. I plugged OpenAI's embedding model into a Next.js page, picked a random word, and typed guesses into a text box. The ranking feedback was surprisingly addictive, so I kept building.
How the ranking works
Every word in English gets converted into a high-dimensional vector (a list of numbers that captures its meaning). When you guess, the server measures the distance between your word's vector and the secret word's vector, then tells you how many other words are closer. Rank #1 is the answer. Rank #50,000+ is ice cold.
The model is OpenAI's text-embedding-3-small. It was trained on a massive text corpus, which means it picks up relationships that feel unintuitive sometimes—and that's what makes each round interesting.
The stack
Next.js for the front end, FastAPI for the API layer, Supabase Postgres with pgvector for storing and searching embeddings, and Redis for caching today's rankings. The daily word refreshes at midnight UTC and is the same for every player worldwide.
Get in touch
Spot a bug, have a feature idea, or want to say hi? Drop me a line on the contact page. I read every message.
