Where this started
I was studying Software Engineering in Turkey when the idea for Ulter first took shape. I kept running into the same question, over and over: why should anyone have to juggle multiple different websites and apps just to figure out which AI model actually fits what they’re trying to do? Every model had its own tab, its own login, its own quirks — and the average person was left guessing. That question never really left me, and it’s the reason Ulter exists today.
What Ulter is building
Okran.ai — the multimodel platform
All the frontier models in one place, with an autorouting layer that sends each request to the model that fits it. It is the answer to the tab-juggling problem, and it is the for-profit side of Ulter — the revenue that lets the company conduct and carry its own research and grow.
ElaNa — the offline model
Ulter’s first offline model: the Okran LLM shrunk by 95–99% so it can run without a live connection. It is built for the people and organizations for whom privacy isn’t optional — individuals, corporations, and governments — and for everyone whose internet isn’t fast or constant.
The goal underneath it
The platform is what exists today. The goal is our own model — the Okran LLM — quantified and reduced through Project Quantum Reduction until it is small enough to become ElaNa and run offline. Okran.ai funds that work; the research is what the funding is for.
The research log
Every research direction behind Ulter is written up in the same shape — what it is, its status, and why it matters — so the log stays easy to scan as it grows. The posts follow the dependency chain: platform, then model, then reduction, then the offline model.
-
Entry 01 · Active development
Okran.ai — The Multimodel Platform
All the frontier models in one place — and the business that funds everything else in this log.
-
Entry 02 · In development
The Okran LLM
Ulter’s own model — the goal the platform is building toward, and the model ElaNa will be made from.
-
Entry 03 · Research memo drafted
Project Quantum Reduction
Quantifying the Okran LLM with quantum-inspired tensor network compression. Without it, there is no offline model.
-
Entry 04 · In development
Project ElaNa — The Offline Model
A 95–99% smaller Okran LLM that runs with no connection, and syncs compressed weights down from the server whenever one is available.