A browser game where you draw, and AI vision models (the bots) race to guess your doodle. The first bot to guess correctly wins the round.
You do not need to install anything. It runs entirely in the browser. Pick Fake guesser to play without an API key.
Watch a game
Screenshots





Posts
I write about building and playing Doodle Derby: what I learn about vision models and how they see drawings.
How it works
- A game has 10 rounds. You get 30 seconds to draw each word.
- The words get harder during the game: easy (ball, hat, sun), then medium (bicycle, umbrella, guitar), then hard (lighthouse, octopus, helicopter).
- About every 2.5 seconds, every bot gets a snapshot of your drawing. A bot that guesses wrong tries again, so the bots keep guessing until the time ends.
- Each bot shows its confidence on a 10-rung ladder. It rings the bell when it is right.
- The first correct guess scores a point for that bot, and a point for you as the artist.
You can also switch roles. A Doodler bot draws the word, and you guess with the other bots.
Features
- Many providers - Anthropic, OpenAI, Baseten, Fireworks, Nebius Token Factory, Ollama (cloud or local), Together, or any OpenAI-compatible endpoint
- Vision filter - the model list shows which models accept images
- Test button - sends one small image request to check the key and image support
- Winning tally - right guesses, guesses made, API calls, tokens in / out, and cost for each bot
- Share your doodles - images of all 10 drawings and every bot’s guess trail, including a “Wildest guess” award
Privacy
API keys stay in memory for the current browser tab only. The game never saves them, and sends them only to the provider you choose.