Multiplist
Multiplist
Practical context handoff • reviewed October 22, 2026

Use the current plan.
Keep the evidence.

When AI returns to an old decision, identify the source it used. Preserve the earlier record and make the reference for the current task explicit.

Real public artifacts • authored demonstration records

Our example compares two existing Multiplist tutorials: the earlier narrated Claude-context episode and the later silent selected-notes episode. Both remain valid published artifacts. This task uses the later edition as its production reference.

The two notes below are newly authored handoff examples, not historical approval transcripts. They have not been saved to a vault or retrieved through a fresh assistant session in this production. The product-screen image in the video is an earlier real capture; connection and retrieval steps are labeled illustrations.

Carry the correction forward

One task. One explicit current record.

  1. Ask which source supports the old decision. Request its title, date and passage.
  2. Keep the earlier record when its history matters. It remains a separate published example. [1]
  3. Write the current task's reference, scope, supporting evidence and what it replaces. [2]
  4. In your intended Multiplist account and realm, import the reviewed note and confirm the saved text. These files are ready to import; no new import is claimed here.
  5. Connect, authorize and enable Multiplist in the supported assistant. Check account and workspace controls. [3]
  6. Request both records explicitly. Compare the returned wording with the original before drafting.
  7. Review native memories, instructions and separate reference files where needed. A saved CURRENT label does not change them automatically. [4]

What changed, and what did not?

RecordUse in this exampleBoundary
October 9 Claude-context episodeEarlier narrated exampleKeep its published version unchanged.
October 19 selected-notes episodeCurrent production reference for this taskUse its silent caption-led format; review before a new release.

The existing local release copies were inspected: the first has an audio stream; the second has none. This checks those files. It does not establish that another assistant has retrieved a note, resolved a conflict or received permission to publish.

A request to try after saving

Retrieve the original content of "Current task reference - silent selected-notes cut" and "Earlier reference - narrated Claude context tutorial" from Multiplist. Show each title, date and relevant passage. Identify which record applies to this task and what remains unchanged. Then draft caption cards for the current edition. Flag conflicts or missing access before continuing; do not infer that a CURRENT label changes other memories automatically.

This is an example request, not a transcript of a successful new run. If retrieval fails, supply the reviewed note directly or resolve access before relying on the result.

Check the next piece of work

For this example, caption cards should follow the silent selected-notes reference, while leaving the earlier narrated episode intact. Check the actual output; an acknowledgment that AI read the record is not proof.

October 19 is the referenced artifact's publication date. October 22 is this record's review date. Neither label is a claim about when someone approved every choice or about a permanent production policy.

Included in this kit

The guide, two complete authored Markdown records, an example retrieval prompt, a next-task checklist, a structured record file, local checksums and the actual Multiplist logo. Public video and documentation links require a network; the records, citations and return links work offline without JavaScript.

No private chats, account details, source IDs, local filesystem paths, embeddings, voice recordings, model caches, credentials or full upstream documentation are included. Review your own text for personal context before sharing. Multiplist is hosted; review its privacy policy before saving sensitive material.

Inspectable evidence

Evidence 1 • complete authored record

Earlier reference

# Earlier reference - narrated Claude context tutorial

Artifact published: October 9, 2026
Record reviewed: October 22, 2026
Status in this demonstration: Earlier example; retain as history.

Public tutorial: https://www.youtube.com/watch?v=wa_NQvhxaTA
Companion: https://multiplist.ai/learn/how-does-claude-cowork-memory-differ-from-claude-chat

The existing local release copy includes a narration track. It demonstrates carrying research between Claude surfaces. It remains a separate published episode.

Boundary: Its narrated format is not an instruction to add narration to the later selected-notes tutorial. Check the edition requested for the current task.

This is an authored description of public production artifacts, not a quoted conversation, extracted seed, approval transcript or fresh assistant response.

SHA-256 of exact UTF-8 Markdown bytes: ccd6ac9169945de6aa452950285b8ff1dd93a4f7a2bf5f0eaef82e548a3d8c29

Checksums detect snapshot changes; they do not establish authorship, approval or currency.

Return to the checklist ↑

Evidence 2 • complete authored record

Current task reference

# Current task reference - silent selected-notes cut

Record reviewed: October 22, 2026
Referenced artifact published: October 19, 2026
Scope: This demonstration's next production handoff.
Status in this demonstration: CURRENT reference to request explicitly.

Use the silent, captioned selected-notes tutorial as the reference for this task. Keep its reading pauses and evidence guide attached. Do not add narration merely because an earlier tutorial used it.

Public tutorial: https://www.youtube.com/watch?v=33puBgBGPF4
Public short: https://www.youtube.com/shorts/cY_TjurajNA
Public guide: https://multiplist.ai/demos/selected-notes-guide/index.html

Replaces for this task: Using the earlier narrated Claude context episode as the production-format reference.
Preserves: The earlier published episode and its original evidence. Do not overwrite or remove it.

Observed basis: The existing local selected-notes release copy has no audio stream; captions follow deliberate reading holds.
Next step: Prepare caption cards from this reference and inspect the resulting words and pace before release.
Open check: Confirm which edition the next task actually requests. This note does not establish a permanent policy for all future videos.

This is a newly authored demonstration handoff based on public artifacts. It is not a historical approval record, a saved vault source, an extracted seed or a live AI result. Review date is not approval date.

SHA-256 of exact UTF-8 Markdown bytes: e9075fa4b2f3c67949715119396b77df4429b0fa1c013b5c21bca7182d38bf4c

Checksums detect snapshot changes; they do not establish authorship, approval or currency.

Return to the checklist ↑

Evidence 3 • official documentation, checked October 22, 2026

Claude custom connectors

Remote connectors require configuration, authorization and enablement. Free users are limited to one custom connector; organization controls apply. This is separate from changing Claude's native memory.

Current Claude connector guidance (online).

Return to the connection step ↑
Evidence 4 • official documentation, checked October 22, 2026

Native memory is a separate layer

ChatGPT allows corrections to remembered information, with available controls varying by account and workspace. Deleting a chat alone need not remove a separate saved memory. Claude also provides controls for reviewing and changing memory; search and memory have their own availability and boundaries.

OpenAI memory guidance · Claude memory guidance (online).

Return to the native-memory step ↑