Memory Catch-Up

​

Why Catch-Up is the best first step

Everyday Memory learns gradually—one candidate fact per completed Ask Anything turn. That works well once Memory already has context, but it is slow when you are starting from zero. Memory Catch-Up solves that bootstrap problem. In one run it:

• Scans the last 30 calendar days of personal chats and meeting transcripts
• Includes the current workspace when one is active
• Compares candidates against saved Memory before writing anything
• Saves high-confidence facts automatically as a trusted Alter action

Think of Catch-Up as seeding the notebook. Everyday Memory keeps it updated afterward.

​

When to use it

Run Memory Catch-Up when:

• You just enabled Use Memory and want a head start
• Settings → Memory is empty but you have weeks of Alter activity
• You return after a break and want recent goals, decisions, and preferences captured
• You switched to a new workspace and want project context from recent work
• You would otherwise paste the same background into every new chat

Catch-Up is a one-time-style backfill for the last 30 calendar days, not a continuous background indexer. You can run it again later if you want another 30-day refresh.

​

How to start Catch-Up

Catch-Up appears only when the chat is writable, idle, and not already running another action or attached context. If Memory was disabled, Catch-Up turns Use Memory on for the run so the review can complete.

​

What Catch-Up reviews

​

Time window

• 30 inclusive calendar dates ending today in your local timezone
• Example: if today is July 16, the window starts June 17

​

Sources

Catch-Up exhaustively paginates through chats and meetings in that window. It does not stop after the first page of results.

​

What it saves

Catch-Up looks for durable, high-confidence items such as:

• Stable preferences
• Active goals
• Clear decisions
• Useful facts
• Meaningful recent changes

It skips secrets, credentials, sensitive personal data, reminders, short-lived logistics, raw excerpts, and low-confidence guesses. Each saved item is one concise standalone fact with provenance from the matching scope.

​

What to expect during the run

At the start, Alter tells you—in plain language—that it will review the last 30 days, compare against saved Memory, and save only genuinely new or updated details. During the run Alter:

1. Reads chats in the date window
2. Reads meetings in the date window
3. Compares candidates against existing saved Memory
4. Remembers new facts or replaces outdated ones when evidence is clear

Catch-Up does not:

• Ask you to approve each individual save
• Delete or forget existing memories
• Review workspaces other than the active one
• Broaden the 30-day window

When it finishes, you get a structured summary with counts and bullets for added or updated facts, plus a short plain-language explanation of how Memory works and a few example questions to try next.

​

After Catch-Up

Recommended follow-ups:

1. Forget anything that is wrong or no longer relevant in Settings → Memory
2. Turn on Ask before saving if you want tighter control over future automatic saves
3. Link Memory to custom actions that should reuse saved facts or search history

​

Catch-Up vs everyday Memory

Use both: Catch-Up to seed Memory from history, everyday Memory to maintain it as you work.

​

Privacy notes

Catch-Up reads local chats and transcripts already stored on your Mac. Saved results remain local Memory entries. The review itself uses your configured generative model to compare and phrase concise facts. Treat that like any other Alter action that calls a model. See also Local data and privacy.

​

Related docs

• Memory overview
• Manage Memory
• Meetings
• Alter Actions