What a memory does
A memory is a saved statement for future tasks: a fact you assert, a position you confirm, or a style example. Use Brain → Memories to manage active statements. A memory candidate is only a proposed statement; it is not active memory yet.
Add a memory deliberately
Choose Add memory, select Fact, Position, or Style, enter a statement, and choose its audience before saving. Private is the default. Saving a statement is your explicit choice, but labeling something a fact does not independently verify it.
For public drafting, a memory must be active and public-approved. A public-approved confirmed position may express your stance; an AI inference or outside source must not be relabeled as your opinion.
Review candidates separately
Edits to a draft can suggest a style example in Brain → Candidates. Read its statement and audience, and check its origin where shown. Choose Confirm only when you want it remembered, or Reject to leave it out. Candidates start private. Confirming does not automatically authorize public use.
The current candidate list offers Confirm and Reject. If a statement needs correction before you agree, reject it and add the correct memory explicitly. You can edit an active memory later, and choose Make public only if you permit public-draft use.
Edit, change audience, or delete
Open Memories to edit a statement, save a new version, change its audience, or delete it after reading the confirmation. A conflicting edit requires comparing with the latest version. Deleting a memory can block dependent drafts; it is not just hiding a row.
Current limitations
Memory does not promise perfect recall of every active statement on every task. Only eligible statements can be considered. Candidates from edits are examples of text you changed, not personality conclusions. Long edits may create no candidate; use Add memory to save a concise example yourself. Confirmation preserves the statement's origin rather than rewriting outside or AI material as owner-authored.