Local

cognira Entity

The same models on your disk. cognira Retrain 12B and Omega 188M, local memory, optional weight retraining. Nothing leaves. Included on every plan.

Meet cognira EntityUseful chats teach it. See what stuck. cognira is independently served. Teach it something and it works from the next reply — every reply that used a memory shows you which one, with a one-tap Forget.

You

Any name ideas for my hiking newsletter?

cognira Retrain

“Trail Notes” — short, and easy to remember, Tom.

In short

How it learns

A guarded learning loop that improves from useful signals without letting one bad chat rewrite the system.

Correct a reply, rate it, or explain a new word. Confirmed meanings enter your private memory immediately so cognira can use them in the same conversation.

Sensitive, unsafe, badly formatted, and down-rated replies stay out. Only short, useful examples can enter the reviewed learning queue.

Background updates wait for enough clean examples and an idle server. A candidate must pass checks before it can replace anything live.

cognira Entity is the local product — a self-learning AI that lives in your machine. Same models as the hosted chat (cognira Retrain 12B and Omega), but inference, memory, and optional weight-level retraining all stay on your disk. Nothing leaves. Included on every plan. Download it at cognira.dev/entity.

Yes, and you can watch it happen. Teach it a preference, a fact, or a word and it's saved to your private memory immediately — the next reply that uses it tells you exactly which memory it drew on, with a one-tap Forget button. A small number of taught words can go on to become shared knowledge every user's assistant can draw on, but only once several independent people teach the same thing days apart, and only after it's reworded so it's never anyone's literal sentence. Every rejection is logged too — see the whole thing, successes and failures, at cognira.dev/learning.

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.

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 We do not publish a carbon number without proper metering.

No. cognira does not forward your question to an external answer API. cognira Retrain 12B is our current model, retrained on top of a Gemma 4 base and served ourselves through MLX on our own machine. cognira Entity runs those same models on your computer. The earlier 188M Omega checkpoint is deprecated in-product, but lives on, open-source, on Hugging Face.

The compact local serving setup keeps costs low. The Free plan covers casual weekly use; Pro and Max raise your weekly allowance and help pay for serving and the guarded learning loop. Every tier gets the same answers. No ads, no selling your data.

Conversations are stored so eligible parts can enter the guarded learning process. We don’t run ads and we don’t sell data. If you want your account and chats removed, email us and we’ll delete them.

cognira is built by Thomas Conway — a solo, independent build, not a lab or a startup with a marketing team.

Get started

cognira is free to use. Confirmed teaching works immediately, and safe, useful replies can become reviewed learning candidates.