Changing Threadbear itself
You can change Threadbear by asking a thread to. Start one in your Threadbear folder (your fork, the copy setup ran from) and say what you'd like different. If a setting or a patch would do it, the agent uses that rather than changing Threadbear's own code, so Mike's updates come in cleanly later. A thread there gets its own copy of the code to work in, and when the change is ready and its checks pass, the agent lands it on your main branch.
Updates then arrive by themselves. Within a few minutes of a change landing on main, the runner builds it and switches over without cutting off any thread that's partway through a turn, and the app builds its new version in the background. The app swaps in the new version when you next quit it, when you choose Restart to Update in the Threadbear menu, or by itself when you've left it alone for a while with nothing typed. The window stays where it is while it does: the new version opens the same thread underneath and takes over, so you shouldn't see it go.
The Updates icon on the left-hand rail shows what's on its way for the app, the runner and the Claude Code and Codex tools. A grey dot means something is in progress and will sort itself out, and a coloured dot means it needs you. Right-click it for a quick Restart to Update.
Mike's updates
Your copy is a fork of Mike's, and it doesn't change unless you say so. About once a day Mike publishes a release. Your runner checks for new ones every few hours, and only trusts releases signed with the key setup saved. When there's one your copy doesn't have, the Updates icon gets a coloured dot, and its From Mike section lists what's new. Bring Them In starts a thread that merges Mike's changes with your own, sorts out any clashes, runs the checks and asks you before it lands. From a thread, tb updates says the same, and tb updates check looks again now.
The activity log
Turn on Keep an activity log in Settings (⌘,) > General and the app notes down what you do in it, so the agents working on Threadbear can see how you really use it: what you click, which threads and pages you open, the menu items and shortcuts you use, the tooltips you stop to read, and when you send a message or answer a prompt. It never notes what you type or the words of a message, only which button, row or item it was, by its tooltip or its name in the menu (so clicking a thread's row puts that thread's title in it). An agent reads it with tb activity, or tb activity --summary for what's used most.
It's off until you turn it on. It's kept in the activity folder of Threadbear's data folder, a file a day, and each file is deleted after 30 days. Threadbear never sends it anywhere, though once an agent has read some of it, that's part of its thread like anything else it reads.