Amr Mareh · The Journal

Research log · Entry 01

Okran.ai — The Multimodel Platform

All the frontier models in one place — and the business that funds everything else Ulter is researching.

What it is: Okran.ai is Ulter’s multimodel platform — all the frontier models available in one place, with an autorouting layer that sends each request to the model that actually fits it.
Why it exists: Two reasons at once — it’s the product that solves the tab-juggling problem for users, and it’s the revenue that funds Ulter’s own research.
Status: Active development.

The main goal: one place for all the frontier models

Okran.ai is the direct answer to the question that started Ulter. Instead of a person having to choose between many different AI websites — each with its own tab, its own login, its own quirks — Okran.ai is the single place where all the frontier models live, and the autorouting layer takes the choosing off the user’s shoulders entirely. That is what the platform is for: removing the friction between a person and the model that can actually help them.

The frontier models sit side by side on the platform — Grok is one of them — and the routing layer decides which one takes a given request. The user doesn’t need to know which model answered, or keep an account with each of them. They ask; the right model answers.

The other reason: it pays for the research

I’ll be straightforward about the second reason Okran.ai exists. It is the for-profit side of Ulter — the product that earns money — and that money is what lets the company conduct and carry its own research and grow. Everything else in this log costs money to do: training a model of our own, the reduction research, building an offline model. Research has to be funded by something real, and Okran.ai is that something.

So the platform isn’t a detour from the research — it is what makes the research possible. One half serves users today; the other half buys the time to build what comes next.

Where it leads

The platform is what exists today, and what pays the bills. The goal underneath it is our own model — the Okran LLM. A platform that routes between other people’s models solves the friction; owning the model is what lets us decide how it behaves, how small it can get, and where it can run. That’s the next entry in this log.

From there the chain is straightforward: the Okran LLM has to be quantified and reduced through Project Quantum Reduction before it can become ElaNa, the offline model. Each step exists because of the one after it.


Next: The Okran LLM — the model we’re building toward →
← Back to the research log