Ask the sky in the Journal
Ask the sky is the Journal's built-in AI assistant. You type or say a question about your journal, and it answers — and it can act on your journal with the Journal's own actions, so what it does leaves the same history your own tap would. How it works across the family is on Ask the Sky across the suite.
Open it
- Tap "AI" — the tall zenith star — top right of the home views, the escritoire and a journal's page, or in the top line of compose, design and bring to life.
- Or, on an entry's ⋯ menu, tap "ask the sky about it". The box opens holding the entry's title — or the start of its words when it has none. Nothing is sent until you tap "send".
The "AI" button shows only while the suite's AI switch and settings › "ask the sky" › "zenith ai in this app" are both on. "ask the sky about it" isn't offered on a hidden entry.
Who answers
Cielo, the tulpa — the suite's word for the AI who answers — who comes with the suite itself, unless Zenith AI has made another tulpa active. The empty sheet says "ask cielo anything about your journal."
The sheet
The header, after the name:
| button | what it does |
|---|---|
| the eye | "the screen stays here" / "the screen goes too". On, the sky also reads what is on your screen with your next ask: the view, how many stars you can see, and the entry open in the reader. A hidden entry is never sent; for an entry in a locked journal it says only "an entry in a locked journal is open. its words are not sent.". It starts off every time the sheet opens |
| the lens button | only when Zenith AI is installed and offers lenses: "zenith ai's lenses · none on". A lens is an answer style — "how the sky answers, not who answers". A tap on a chip switches it on or off for the whole suite |
| memory | ✦ "these talks are remembered" or ✧ "these talks are not remembered" (Gold). On, the sky remembers your last ten talks; off, each ask stands alone. Without journal gold, a tap opens the Gold sheet |
| read aloud 🔊 | "answers are read aloud" / "answers are not read aloud". On, each answer is spoken as it arrives, in the tulpa's voice where the phone has it. Not the same as "spoken labels" in ease |
The talk — the last twelve exchanges, "you: …" then the answer, with a line for each thing it did: "done", "refused — you said no", or what happened. "asking…" while it works.
"⚑ report" under every reply — flags it as wrong or harmful; the reply then says "reported". The flag is kept on this phone, and nothing is sent. See AI safety — Zenith Guard & consent.
The ask — "e.g. what did i write about the sea?". The mic, "speak the ask", uses the phone's own speech recogniser. Then "send", or the keyboard's send.
Under the talk: "✦ ai-generated — verify anything that matters.", with "acceptable use ↗" and "privacy ↗".
The first time it sends your words
The first time your words would go to an AI company, the suite's own sheet asks first — the same in every Zenith app. Nothing is sent until you accept.
- It is titled "Send to <the service>?" — for example "Send to Claude (Anthropic)?" — and says what goes, to whom, under your own key.
- Its last line: "You can change your mind anytime in settings › ask the sky › “sent with your yes”. On-device (Gemini Nano) never leaves your phone and needs no permission."
- "Send to <the service>" is your yes, and the ask goes now. "Not now" sends nothing and records nothing; a tap beside the sheet, or back, is the same.
A yes is for one company. Switch to another service and its first send asks again. See The ask the sky settings.
If it isn't set up
The sheet says "the sky isn't set up yet —" and why — "the suite's ai is switched off", "zenith ai is switched off in this app", or "no key is saved for …" — with "set up the sky in settings ›". See The ask the sky settings.
With on-device chosen on a phone that doesn't have Google's on-device model, every ask is answered "on-device ai isn't available on this phone — choose another service in settings · ask the sky", and nothing else happens.
How an ask runs
- It may take several steps; if one ask runs out of steps, it says "that took more steps than one ask allows — here is what was done."
- When what it does opens another screen, the sheet closes, but the ask still finishes. Its reply is waiting the next time you open ask the sky. Anything in it that would need your yes is taken as a no.
- If something unfinished is open on top — compose, design, bring to life, a passphrase, something arrived, a file to open, journal gold — it opens nothing, and says that screen "is open on top, holding something not finished; it stays until you close it".
- A show it builds is only a preview until you ask it to keep it.
- A Gold feature asked for without journal gold is refused in words — "that is part of journal gold (…); it opens on the gold page in settings" — rather than opening the Gold sheet.
- Anything that can't be undone asks you first, or is refused while that is switched off. See Things it asks first.
What it can't see
It never sees a hidden entry, or an entry in a locked journal. Locking, opening a locked journal and the app lock are never the sky's.
Other Zenith apps
It can also use the other Zenith apps' own actions on this phone — "add a task in Zenith Tasks" reaches Tasks — but only when you name that app. If the other app doesn't answer, the line under the reply says "no response from <app>". See How the Journal fits the suite.
See also.
What the sky can do — entries & journals · What the sky can do — design, screens, settings & more · The ask the sky settings · Ask the Sky across the suite