# How Minds works

A practical map of memory, knowledge, learning, and control.

Source: https://docs.minds.sh/docs/concepts



Minds combines storage and computation in a dedicated Akasha instance. The useful question is not “which subsystem sounds most advanced?” It is “what should this application remember, retrieve, or learn?”

## Follow the work [#follow-the-work]

1. **Keep a source.** Store an event, a document, or another useful input with enough context to identify it later.
2. **Connect what it means.** Represent relevant entities and relationships in the knowledge graph.
3. **Retrieve what matters.** Search for the context needed by the current task.
4. **Review new knowledge.** Use learning isolation when incoming information should not be accepted immediately.
5. **Inspect the result.** Keep lineage and operation evidence so a conclusion remains explainable.

These are application design choices, not a promise that every ingestion operation automatically runs every stage.

<Cards>
  <Card title="Dedicated instances" href="/docs/concepts/instances">
    The boundary around a Mind and the Free/Pro runtime requirement.
  </Card>

  <Card title="Six kinds of memory" href="/docs/concepts/memory">
    Choose the right representation for each kind of context.
  </Card>

  <Card title="Knowledge graphs" href="/docs/concepts/knowledge-graphs">
    Facts become more useful when their relationships and history are explicit.
  </Card>

  <Card title="Learning Isolation" href="/docs/concepts/learning-isolation">
    Separate new information from accepted knowledge.
  </Card>
</Cards>