AWAITING SIGNAL...
OPERATIVE
No recent games.
No open lobbies.
The code is open source, please help us improve it!
View on GitHub Join our DiscordThis is a community project. Get involved:
Words exist as vectors in high-dimensional space. Similar meanings cluster together. We use OpenAI text-embedding-3-large to measure semantic distance.
We measure how "close" two words are using cosine similarity - the cosine of the angle between their vectors. Raw values range from -1 (opposite) to 1 (identical).
cosine(A, B) = (A · B) / (||A|| × ||B||)
Raw cosine values cluster between 20-60%, making differences hard to read. We apply a sigmoid transform to spread them across 0-100%:
s = xn / (xn + (c(1-x))n)
Nerd mode (in options) shows the raw cosine similarity alongside the transformed value.
OpenAI text-embedding-3-large - 3072 dimensions, state-of-the-art semantic understanding.
Win ranked matches to climb. New operatives calibrate faster due to higher volatility.
Free to play. Donations cover server and API costs.
Supporters unlock custom theme colours and an exclusive badge. One-time donation, permanent access.
Any amount unlocks all supporter perks.
Join a match to chat.
// decode. deduce. eliminate.
Ranked requires Google sign-in + 5 casual games
Establishing connection...
Operatives: 0/6
Minimum 2 required
Operatives: 1/6
Add at least 1 AI target
DATABASE: -
Locked in: 0/0