What to remember, and what to forget

A memory should earn its shelf space.

That is true for the drawer full of mysterious cables. It is also a useful rule for the context you keep in an AI assistant. More details do not automatically make better context. The useful ones help with the next conversation.

Keep details you would otherwise repeat

Start with information that stays useful across several conversations:

• A preference that affects your plans.
• A constraint you often have to explain.
• Background for a project you are still working on.

Be concrete. “I prefer quiet places for dinner” gives an assistant more to work with than “I like nice restaurants.” Both may be true. One does a job.

Birch has memories you can review and manage. Use them for context you want available again, rather than assuming every passing remark should become a fixture.

Let temporary context be temporary

Some details have a short working life. A plan for this week may stop being useful next week. A preference can change. Old context can steer a new answer in the wrong direction.

Review what you keep when circumstances change. Edit what is still useful; remove what no longer fits. Birch also gives memories expiration dates, so context does not have to live indefinitely.

Separate remembering from doing

A memory describes context. A reminder asks for a next step at a particular time.

If you need to revisit a decision on Friday, set a reminder with a clear action. If you want a recurring preference available in later conversations, keep it as a memory.

You can use both. Just give each a clear job.

Try a small review

Look at the details you have saved and ask: would I explain this again today? If the answer is no, it may be time to edit or remove it.

Then keep the few things that make the next conversation easier. The drawer of cables is a separate afternoon.