With deterministic story execution and coherent persistence in place, the project identified its next uncertainty: can the player communicate naturally without giving a probabilistic model control of the story or credit for language the learner did not produce?
The planned player action is unrestricted typed language. An interpretation layer would map that transient input to one of the communicative intentions already allowed by the authored situation. The deterministic engine would still decide the canonical transition.
The first approved batch is intentionally a contract boundary, not a production model call. It must distinguish raw transient input, interpreted meaning, canonical intent, linguistic observation, uncertainty, and failure.
This is a planning milestone, not a completed capability. The value of the next phase will be in testing whether freedom of expression can grow while narrative truth, privacy, and evidence remain constrained.
← Back to HelloStranger