Returning to a shared project after a holiday, a busy week, or simple neglect usually means scrolling and guilt. If the project lives in Thoughts and is synced, there is a faster way back in: ask ChatGPT or Claude what changed, who added what, and what is waiting on you — then read only the thoughts that matter. This guide gives you a five-minute catch-up ritual, the exact questions to ask, and an honest account of what AI catch-up can and cannot see.

What you need

The project must be synced and shared in Thoughts — catch-up happens at the cloud boundary, so local-only projects have nothing to catch up from. Connect your assistant once (ChatGPT · Claude). Everything below then works in either, from the same library, as described in one second brain for ChatGPT and Claude.

Minute 1: Orient — what is this project, at a glance?

What is my Launch project about? Give me the short version, then list its most recent thoughts with who added each.

The project summary plus a recent-thoughts list rebuilds your mental map faster than opening the app cold. You are looking for shape, not detail: what areas moved, what looks new since you last looked.

Minute 2–3: Diff — what changed while you were gone?

What has someone else changed in Launch in the last two weeks? Group it by person.

Show me thoughts Sam added, and separately anything Sam last edited.

Keep those two questions distinct, because Thoughts does: authorship (who created a thought) and last-editorship (who most recently changed it) answer different needs. “What is new from Sam?” and “what did Sam touch?” are both legitimate; confusing them misattributes work. Also ask after your own exposure:

Do I have any unread invitations or new shared-thought notifications?

Reading notifications through the assistant never marks them read or accepts an invitation — it is a peek, and the deciding still happens in Thoughts, as the MCP guide explains.

Minute 4: Read the two or three that matter, in full

Read the pricing thought in Launch in full and tell me: what was decided, what is still open, and anything that mentions me or waits on me.

Insist on full reads for decision-shaped thoughts; previews truncate, and a summary of a truncation is how misunderstandings start. Ask the assistant to quote the line behind any claim you will repeat to someone else. These retrieval habits come from asking AI about your notes — they matter most exactly here, where other people's words are involved.

Minute 5: Leave a trace of your return

Draft a short thought for Launch: what I have caught up on, the two things I am picking up, and the one question I still have. Save it to the project.

This closing move is what turns catch-up into collaboration. Your trace tells the group you are back, shows what you absorbed, and files your commitments where the work lives — visible to members, since it is a shared project, which is the point. Review the draft before it saves; the assistant creates only when you ask, and cannot edit it afterwards, so corrections are yours in the app.

What AI catch-up cannot see — plan around it

  • It is not an audit log. Activity shows each thought's latest visible state in your window, not every edit or deletion. “What does the note say now?” is answerable; “what did it say last Tuesday?” is not.
  • Timing follows sync, not capture. A thought a teammate captured offline appears when it syncs. Very recent offline work may simply not have arrived yet.
  • Locked and local content stays invisible. A collaborator's private local notes about the project are theirs; catch-up covers the shared record, which is the correct boundary.
  • Text only. Attachments and recordings are not read through the connector. If a thought says “see attached quote”, open it in Thoughts.

None of these undermine the ritual; they define it. Five minutes gets you oriented, current on people, deep on the essentials, and visibly back — which is four things a scroll never delivered. Fold the same ritual into your week and it merges naturally with the weekly review.

Frequently asked questions

How can AI help me catch up on a shared Thoughts project?

Ask ChatGPT or Claude for the project summary and recent thoughts, what others changed in a time window grouped by person, and full reads of the few decision thoughts that matter. Finish by saving a short caught-up thought so collaborators see you are back.

Can the assistant show me a full edit history?

No. Activity shows each thought's latest visible state within a window, ordered by when the server received it. It is a catch-up feed, not an audit log, and it does not show deletions or previous revisions.

What is the difference between who added and who last edited?

Authorship is who created a thought; last-editorship is who most recently changed it. Thoughts tracks both separately, so ask each question explicitly to avoid misattributing work.

Does reading notifications through AI mark them read or accept invitations?

No. The assistant can show invitation and shared-thought notification summaries, but reading them changes nothing. You review and respond to invitations in Thoughts.