AI

The switchA switch in Settings turns it off

In Settings › AI, a switch labelled AI Master Switch shows whether it is on, with a sentence under it saying what that means right now. Below it, under Each job, every job this instance runs has a switch of its own. Flipping one takes effect on the next background sweep, within a few minutes, with no restart and no redeploy.

Turning it back on fills in the recent gap: the sweep classifies entries from the last two days that never got a verdict. Anything older keeps the shape it was saved with.

What stopsFiling stops, and nothing you wrote changes

With AI off, no job the AI has happens.

Area naming

An entry you write stays with the life area blank instead of being filed under one. An area you set yourself is never touched either way.

Task filing

A task you add stays with the life area blank instead of being filed under one, so it groups under Other on the Tasks page. A task under a goal still takes its goal’s area: that half is a rule rather than a model call, and it does not stop. An area you set yourself is never touched either way.

Note filing

A note you start stays with the life area blank instead of being filed under one. An area you set yourself, from under the note’s title, is never touched either way.

Area descriptions

A life area you name stays with the description blank instead of getting a draft of what it means. A description you wrote yourself is never touched either way.

Transcribe a recording

A video or a voice memo you post is stored and playable, with no transcript, no title and no summary written for it.

What changed in a version

A sealed version of a note or a goal keeps its text and its timestamp, with no line saying what moved since the version before it.

Note summaries, task status lines and travel time are planned, tracked as issues #261 to #263, and each arrives here with its own switch when it is built.

What you type is stored exactly as typed, search still searches it, the timeline still lays it out, and no control anywhere in the app breaks or greys out.

The owner's guide on AI covers what the AI reads when it is on, and which text never leaves the instance.

Model and promptEvery job names the model it runs on and shows the prompt it runs with

Each job row holds two controls beside its switch. Model marks the one in use and offers the other, and Opus is the model every job ships on. Prompt opens the text that job will run with, so an edit starts from what is running rather than from a copy of it. Both controls are on the phone as well as the web, on the same rows.

The three filing jobs hold markers in curly brackets where the app drops their inputs in, and an edit that drops one of those markers is refused instead of saved: the app names the markers it needs back, your text stays in the box, and the job goes on running the wording it already had. The other two jobs take their material as a separate message, so their wording has no marker it must keep. Models & prompts covers both controls in full.

The key underneathThe key is your account’s, and there is no second switch

Every job on this page runs on the Anthropic key stored under Settings › Keys. That key belongs to your account and bills your account. Clear the slot and every job stops, whatever the switches above say; store a key and the switches are the only thing left deciding what runs.

The backend keeps no key of its own. It reads its deployment’s ANTHROPIC_API_KEY once, at the first start after an upgrade, to move that key into the owner’s empty slot, and never reads it again. Whoever runs your LifeOps cannot take AI away from an account that stored its own key.