cognira

What it is

cognira is an independent, self-learning AI assistant that learns from the people who use it and shows exactly what it learned. You teach it in plain sentences and it saves useful information to your private memory immediately, displaying which memory each reply used. A small number of taught concepts can become verified shared knowledge that all users benefit from, but only after several independent people teach the same thing days apart and it passes public checks.

The service runs from a single 16GB Apple Silicon iMac in a living room, not a GPU cluster. cognira Retrain 12B is the current model, retrained on a Gemma 4 base and served through MLX. Every plan gets access to the full model with no reduced functionality.

cognira Entity is the local version that runs entirely on your own hardware with the same models, local memory, and optional weight-level retraining, with nothing leaving your machine. The service is free to use with optional Pro and Max paid tiers that raise weekly token allowances.

Quick answers

What is cognira Entity?

cognira Entity is the local product — a self-learning AI that lives in your machine. It runs the same models as the hosted chat (cognira Retrain 12B and Omega), but inference, memory, and optional weight-level retraining all stay on your disk with nothing leaving the machine. It is included on every plan.

Does cognira really learn from my chats?

Yes. You can teach it a preference, fact, or word and it is saved to your private memory immediately — the next reply shows which memory it used with a one-tap Forget button. A small number of taught words can become shared knowledge only after several independent people teach the same thing days apart and it is reworded so it is never anyone's literal sentence. Every rejection is logged publicly at cognira.dev/learning.

How smart is cognira?

It is a compact, locally served model, and frontier systems will still beat it on breadth and trivia. Its edge is the loop: confirmed teaching works immediately from private memory, then verified updates can improve the model where people actually use it.

Is it actually green?

The service runs on one 16GB iMac, caps needless output, and pauses background learning when inference needs the machine. The company does not publish a carbon number without proper metering because most AI energy claims are difficult to prove.

How much does the Platform API cost?

The API costs £1.50 per million tokens for input and output on either model. The minimum top-up is £5. Chat plans (Free, Pro, Max) are a separate monthly subscription.