Memory keeps the record
The workspace stores the workshop note, task definition, first brief, finding, corrected brief, and checkpoint. It preserves the unsupported Saturday statement as correction history instead of erasing it.
That record answers questions such as:
- What material existed before the task began?
- Which version came first?
- What did the review change?
- Which final file belongs to this job?
Memory stops at preservation and organization. It does not decide what the next task needs.
Retrieval chooses the useful slice
The task asks for a concise workshop content brief. Retrieval selects the audience, two-hour offer, and confirmed-detail constraint. It returns schedule and price as unknown because the note contains neither.
Input: a question, scope, and source collection. Transformation: locate and compare relevant passages. Output: selected excerpts, source pointers, exclusions, and unresolved claims.
The selection ends when the packet can support the requested brief without importing unrelated material.
Why the separation matters
A small packet reduces distraction, but its deeper value is accountability. The reviewer can compare the first brief with the note chosen for the task. When “Saturday sessions are available” appears, the source trail exposes the missing support.
The correction returns to memory. Future retrieval can find both the corrected brief and the reason the Saturday sentence must not return.
This split fits repeatable work with several records and a need to reuse corrections. It may be unnecessary when one short document contains the entire answer and no later task depends on the result.