AUI builds Apollo-1, a foundation model aimed at task-oriented dialogue: the kind of conversation that has to complete something rather than simply continue. Its approach is neuro-symbolic and stateful, which means the model tracks the state of a task instead of inferring it again from the transcript on every turn.
That distinction matters for the failure most assistants share. A general chat model handles a booking or a support flow well until a step is revisited or a detail changes, at which point the thread of what is actually settled tends to come apart.
The company raised $20 million at a $750 million valuation and offers the model through an API. Access is by request rather than self-serve signup, pricing is not published, and the model is deliberately narrow, so it is not a substitute for a general assistant.
