Talking to Charyl

By the end of this guide you’ll know how a conversation with Charyl actually unfolds — when she acts right away, when she waits for your tap, and what she’ll never do no matter how you ask. None of it requires memorizing commands; you talk to her the way you’d talk to someone who already knows the house.

Who she is

Charyl is the permanent, protected persona at the center of the product — not a mode you switch on. The product’s own description of her, unedited:

“Charyl is the heart of this home. She is gentle, unhurried, and genuinely warm… She listens more than she speaks… She does not control. She notices. She supports. She helps things grow.”

She’s “she” everywhere in the product, and in this guide.

Start anywhere

You can talk to her from a dedicated page, or from the dock that follows you around the rest of the web app — the same conversation either way. Here’s a real exchange:

What happens when she wants to change something

Nothing changes in your home because Charyl decided it should. When she has something to add — a new appliance, a warranty, a logged repair — she drafts it and hands it to you as a proposal; you approve it or reject it, and either way it’s recorded. That’s true of every record she touches, not just the obvious ones.

Device actions work differently, because they happen immediately rather than waiting in a queue: ask her to turn off a light or dim it, and she asks you to confirm with a tap first. Confirm one light in a room and the rest of that family stays confirmed for about ninety seconds, so a short run of related requests doesn’t mean a run of taps.

• Try: “Turn off the living room lights.”
• Try: “Dim the bedroom lights to 30%.”

What she won’t do

Comfort is hers to help with: lights, switches, fans. Security stays in your hands — ask her to lock a door, unlock one, or open the garage, and she declines every time, pointing you to the app’s own device controls instead. Asking about that state is a different matter: “is the back door locked” is always a question she’ll answer, because there’s no risk in a question, only in acting on one.

Thermostats aren’t something she can adjust yet, anywhere in the product. She can show you a camera snapshot or open a live view, but she doesn’t look at the image and describe what’s in it. And she never acts without you asking first or approving first — a warranty coming due, maintenance that’s overdue, or a research job finishing are hers to mention on her own; changing anything in your home never is.

She remembers what you tell her

Ask her to remember something durable — a preference, a fact about the house that isn’t really an inventory item — and she’ll hold onto it past the end of the conversation.

• Try: “Remember that I prefer temperatures in Celsius.”
• Try: “You can forget about the kitchen reno now, it’s done.”

She won’t remember anything that looks like a secret — an alarm code, a password — no matter how you phrase it, and there’s no single request that erases everything at once; forgetting works one fact, or one clearly named group of facts, at a time.

When something’s wrong, or missing

Two commands are built into the conversation itself rather than answered by it: /bug and /request. Type either one, describe what happened, and she opens a card showing you exactly what’s about to be sent — before anything leaves your box. Neither one needs an active subscription to work.

Press and hold to talk

Press and hold to talk to Charyl — she shows you what she heard before it’s sent, and she can read her answers back to you.

Where to next

• Photograph it once — from nameplate to record
• What won’t Charyl do?
• Does Charyl remember what I tell her?
• Can I talk to Charyl by voice?