# Inspect learning and computation

Understand what each cognitive and neural view controls.

Source: https://docs.minds.sh/docs/workspace/cognitive



These views expose engine capabilities beyond storing and retrieving a record. Choose one because it fits a task, not because every application needs to enable or tune it.

| Workspace area      | What it represents                                         | Learn more                                              |
| ------------------- | ---------------------------------------------------------- | ------------------------------------------------------- |
| Cognitive Cycle     | A cycle that coordinates context and processing            | [Reasoning and cycles](/docs/concepts/reasoning)        |
| Neurosymbolic       | Retrieval and learned signals combined with explicit rules | [Reasoning API](/docs/api/instance/reasoning)           |
| Continual Learning  | Training tasks and state that change over time             | [Continual learning](/docs/concepts/continual-learning) |
| SSM / Consolidation | Memory processing and consolidation controls               | [Memory API](/docs/api/instance/memory)                 |
| Self-Improvement    | Inspection of proposed changes and learning state          | [Learning isolation](/docs/concepts/learning-isolation) |
| SNN Networks        | Event-based neural network operations                      | [Neural systems](/docs/concepts/neural-systems)         |
| ML Models           | Model lifecycle and inference                              | [Models](/docs/concepts/models)                         |
| MHN Patterns        | Associative storage and pattern recall                     | [Associative memory](/docs/concepts/memory)             |

## Verify the operation, not just the status label [#verify-the-operation-not-just-the-status-label]

Read a job's input, configuration, state, and resulting artifact. A completed label is insufficient when the handler is still a placeholder. The current source has acknowledgment-only continual checkpoint operations and a BPTT training path without an executed training loop. These limitations are explicit in [instance coverage](/docs/api/instance/coverage).

For experimentation, use a test dataset and a measurable outcome. Preserve the version of the model, inputs, and configuration so you can compare results or return to a known state through a verified recovery path.