Adapt-1 exposes Core as a self-contained adaptive reasoning substrate. It can preserve and revise state, form task-local structure, compare and revise hypotheses, learn from observations and outcomes, return route-defined predictions or decisions, and expose supporting evidence without requiring generated language. Perception adapters and controllers can connect it to external modalities and actuation; they do not perform its reasoning.
The API namespaces are gateways into that substrate. They are not separate products and do not describe Core’s internal architecture.
Usage patterns
These are examples, not a compatibility checklist. Your product does not need to match one card, and its category does not need to appear on this page. A single application can combine several patterns wherever they fit the task.
Structured results
Adapt-1 does not need to generate a paragraph to produce a useful result. A structured prediction, ranked hypothesis, route-defined decision, constraint set, or evidence packet can be consumed directly by application code.
The machine-readable form is the delivery contract, not the source of the capability. The result can reflect evidence and learned state accumulated and revised across use.
The application owns the final side effect. Under a decision-producing contract, an Adapt-1 result can drive the next application decision. The application executes the operation.
Guides
Applications can combine these workflows. For example, an application can preserve operating constraints and revise hypotheses inside a Domain. The application can bind observed transitions and consequences to state. It can also return a route-defined prediction or decision.
When the workflow acts in an external environment, preserve the exact action-to-transition record beside learned scores or selections. This connects Adapt-1’s reasoning to the committed action, observed consequence, reusable action-effect structure, and later goal-directed behavior. 