Humans: conversations on a belief state
A conversation that keeps what it concluded, and shows why.
A chatbot starts over with every exchange. This experience runs on a persistent belief state instead. It holds an evolving understanding of the person, the situation around them and the tensions still open inside that understanding. In development.
Four properties
Salient
It finds what matters, not only what was typed into the prompt.
Proactive
It raises emerging conclusions, dissonance and open questions before you know what to ask.
Continuous
Beliefs carry across conversations and evolve over time. Nothing resets at the end of an exchange.
Transparent
Every reply sits next to its reasoning state: current beliefs, supporting and conflicting evidence, confidence, competing explanations, unresolved dissonance, and what would change the conclusion.
Not a more conversational interface. A direct interface to an evolving, inspectable model of thought.
What you see
Ask a question. The answer arrives with a thinking trace. Each card shows a dissonance: two beliefs that do not reconcile, how hard they collide, and how confident the belief state is in each side. The beliefs formed while answering stream in as they form.
Same platform underneath
The conversation, the API and agents share one belief state. A belief formed in a conversation is available to every other component reading that session.