Today’s AI models are fluent and capable, yet fundamentally impersonal. By default, they optimize a single response for the entirety of humanity. We believe personalization is what unlocks the next order of magnitude in AI value: the shift from passive tools that people use, to proactive, model-centric systems that self-evolve to suit their needs. In our future, AI won’t just know the world; it will actually know you.
For a system to truly grow through interaction, we have to answer a few questions again: how does it learn, what counts as a learning signal, and what should the system around the model look like?
We break that work into four directions.
The system has to learn from immediate interaction, not only from a centralized offline training cycle. Every conversation should be treated as a signal the model can absorb: your correction, follow-up, change of direction, and even silence all matter.
What the system learns has to remain. Otherwise, every conversation starts with a new user. We study two complementary paths: persona vectors that persist learned state in a form that can condition the model, and memory files that keep a textual representation of what the system has learned.
In most systems, the model is a component at the end of a pipeline. Memory, personalization, content selection, and interface logic sit outside it, with code orchestrating the model. We want the opposite relationship: code moves to the side, the model gets the freedom to learn, and the user keeps final control over the experience.
Offline benchmarks tell us whether a model is good for a population. They do not tell the model whether the sentence it just wrote actually caught you. To help a system grow through interaction, we need evaluation that turns live user behavior into learning signals in real time, correcting and encouraging the model as the experience unfolds.
We are researchers and engineers who ship. Every research direction connects to a live product serving millions of users. We validate ideas through rapid experimentation, not just paper benchmarks — and we publish what we learn.
AI is becoming the primary interface between people and information, entertainment, education, and work. If these systems cannot adapt to individuals, they will flattenhuman experience rather than enrich it. Getting personalization right isn’t a product optimization — it’s a prerequisite for AI that genuinely serves people.
Read the longer mission essay, Let the world grow because of you, for the argument behind this agenda.
We’re Kaon. We’re here to make AI personalized.