The next frontier of AI agents isn't a smarter model. It's a model that knows you.

LLMs have gotten remarkably good at understanding language in the abstract. Yet they still fail at understanding intent — not because the model is dumb, but because intent isn't only semantic. It's contextual, personal, and accumulated over time.

The vertical AI wave is a partial answer — constrain the domain, reduce the ambiguity. But you end up with a dozen specialized tools instead of one that actually gets you.

What if the real unlock is longitudinal memory? An AI that watches, remembers, and builds a model of you specifically. The more context it holds, the less you have to explain — and the more it feels like thinking with you rather than prompting at it.

That's the thesis on which we are building AirJelly. Would love to know any thoughts from this community: is personalized context the missing variable in AI intent accuracy — or is it still fundamentally a model capability problem?