Aether — AI Voice Assistant & Daily Briefing
Aether is a voice-first assistant I'm designing to replace my own scattered mornings — opening Gmail, WhatsApp and the calendar separately — with one spoken interface: a daily briefing, and a way to get things done across all three, built on AI models that can be swapped out rather than locked in.
Why I'm building it
Most of my mornings start the same way: open Gmail, open WhatsApp, open the calendar, and spend the first ten minutes just working out what actually needs me today versus what can wait. Running client work as an Odoo consultant means that list changes daily, and none of those apps talk to each other or will just tell me, in one sentence, what matters right now.
Aether is my attempt to fix that for myself first. Instead of four apps to check, one voice to talk to — it tells you what's going on, and asks before it changes anything. That second part matters more to me than the first: an assistant that can act is only worth trusting if it never quietly does something you didn't approve.
What it's being designed to do
A spoken daily briefing
Each morning, a short spoken summary of what's on the calendar today and the next working day, plus which unread emails actually need a reply — ranked, not just listed.
One voice across email, calendar and chat
Gmail, calendar and messaging sit behind a single conversational interface, so you ask once instead of opening three apps to find one answer.
Swappable AI models, not one vendor
Voice, reasoning, vision and on-screen action are each handled by a separate engine, so the assistant isn't locked to a single AI provider — a stronger model for one job can be swapped in without touching the rest.
Confirms before it acts
Anything that leaves the conversation — sending a message, editing a file — shows a plain-language preview first and waits for an explicit yes. Reading and summarizing don't need that; changing something does.
Shows its work
When it has to drive an app directly — clicking, typing, filling a field — it narrates each step as it happens instead of acting silently in the background.
A running activity log
Every action gets logged with a timestamp, in plain language, and anything reversible can be undone from that same log.
Built on swappable AI, on purpose
Models are improving fast enough that betting the whole product on one AI provider felt like the wrong call. The early design routes voice and everyday reasoning to Google's Gemini models, and hands anything that has to actually touch your screen — clicking, typing, driving a real app — to Anthropic's Claude, because that's currently where computer-use feels most reliable. Neither pairing is meant to be permanent. The point of building Aether around separate, swappable engines for voice, reasoning, vision and on-screen action is that a stronger model for any one of those jobs can be dropped in later without rebuilding the assistant around it.
The kind of thing you'll be able to say
- "Brief me on today"A spoken rundown of your calendar and the emails that actually need attention.
- "Message a contact: running ten minutes late"Dictated and previewed, sent only once you say yes.
- "Summarize my unread email"Read-only, so it just answers — no confirmation needed.
- "Open my spreadsheet and total this column"For the tasks that need Aether to actually drive an app, not just talk about it.
How I already use AI in my Odoo work
Aether isn't live yet, but connecting AI safely to a real system already is. If you want to see the same confirm-before-it-acts approach running in production today, not just designed for one day:
Where it stands today
To be plain about it: Aether exists right now as a working concept and an interface I've designed and prototyped — the voice loop, the confirmation flow, the daily briefing, the activity log — not yet as a shipped product. There's no build to download, no user numbers to report, and no launch date I'm willing to commit to. Everything above is what I'm building toward, not a description of something already running.
Want to know when Aether is ready to try?
I'm Muhammad Salman Ali Khan, an Odoo Techno-Functional Consultant in Dubai, building Aether alongside my client work. Get in touch and I'll let you know when there's an actual build worth testing — nothing before that.
Get in touch