Organize threads

Organization starts to matter after the first few successful investigations. A good sidebar should tell you what the team is working on now, where a line of research lives, and how to find an older decision without reconstructing the org chart that existed at the time.

Folders, names, pins, and search help with that navigation. They do not change a thread's content, access, ownership, or research history.

Create and use a folder

Select New folder and name a durable grouping: a research program, product area, model family, or collaboration. Start a thread from the folder or choose it in the new-thread composer.

Avoid a folder for every small task. A folder is useful when several independent threads belong together but should not share one research history.

For example, a “Protein models” folder might contain separate threads for confidence calibration, mutation effects, and steering. Each question remains independent, while the folder makes the broader program easy to browse.

Name threads for the question

Prefer names that will make sense in search results later:

• Compare probe transfer across medical QA datasets
• Why does checkpoint B lose structure confidence?
• Steer generation toward synthesizable compounds

Names such as “test,” “new run,” or “Friday investigation” force a future reader to open the thread before learning why it exists. Rename a thread as the question becomes clearer; the research record does not change.

Keep important work visible

Pin a folder or thread when you are likely to return to it during the current phase of work. Think of pins as a working set, not a permanent honor roll.

Unpin finished work rather than deleting it. Delete only when you are sure the research record is no longer needed; a share link is not a backup.

Find an older thread

Use Search threads to search titles. A folder scopes the visible list, so return to the top level when you want to search more broadly.

Once you find a long-running thread, read Summary before scrolling the conversation. It should tell you where the research stands, what is active, and which reports matter. Use Timeline when you need to reconstruct how a conclusion changed.

Choose the right boundary

• Continue the thread when the next question depends on its reasoning or evidence.
• Fork when you want to branch from a particular response and preserve that exact starting point.
• Start a new thread when the question can stand on its own, even if it belongs in the same folder.

A useful test is to imagine sharing only the new question with a collaborator. If they would need the existing thread to understand why the work matters or how to interpret it, continue or fork. If the question is self-contained, start a new thread and attach only the earlier evidence you need.

See Threads for the conceptual boundary.