> ## Documentation Index
> Fetch the complete documentation index at: https://project-brain.rocheaiops.net/llms.txt
> Use this file to discover all available pages before exploring further.

# How B.R.A.I.N. learns

> Implemented learning is explicit, attributable, and bounded; broader adaptation is future work.

## Implemented today

Explicit owner teaching can become active after structural and authority validation. Corrections create successors. Document knowledge retains source artifacts, versions, and citations. One bounded consolidation process can create attributed **candidates**; it does not autonomously modify policy, schema, or code.

## Future

Preference inference, pattern learning, hypotheses, routines, skills, autonomous self-improvement, and model adaptation are not implemented. They need evaluation, promotion rules, monitoring, and rollback before they can affect canonical state.

| Maturity                                      | Status            |
| --------------------------------------------- | ----------------- |
| Explicit teaching and correction              | Accepted          |
| Document knowledge and cited owner correction | Accepted          |
| Candidate generation with attribution         | Accepted, bounded |
| Autonomous preference/routine learning        | Future            |
| Model or code adaptation                      | Future            |

See [self-improvement](/roadmap/self-improvement).