Account settings: how the assistant behaves

Overview

Settings holds the handful of choices that change how the assistant behaves for you. They are personal — they follow your account, not a device — and every one of them has a sensible default, so you can ignore this page entirely and nothing will be worse for it.

When a conversation outgrows the model

Every model has a context window. When a conversation no longer fits, something has to give, and you decide what:

• Ask me — the turn pauses and asks before it does anything that costs more. Recommended, and the default.
• Summarise automatically — older turns are condensed so the conversation keeps its thread without carrying every word.
• Always send the full context — nothing is dropped or condensed. The most faithful option, and the most expensive one.

Queue instead of stopping

If your allowance runs out mid-task, you can have the request queued rather than refused. Queued work waits for off-peak capacity and runs then; you can see and cancel everything waiting under Queued requests, with its estimated cost and expected run time.

How answers are written

• Uncertainty markers — when the assistant is not sure, it says so in the answer instead of sounding equally confident about everything.
• Wellbeing check-ins — if a conversation suggests you are having a hard time, the assistant may acknowledge it. Switch it off if you would rather it stayed strictly on task.

Videos you share

Shared photo and song videos end on a short Made with ToRun card. On a paid plan you can turn it off for your own shares.

Starting over

Reset the welcome wizard brings back the first-run tour. It changes nothing about your data — it only forgets that you have already seen it.