NYX AI
A personal AI that lives in your home, protects your privacy, and never sends your secrets to the cloud. Intelligence without compromise.
Imagine a brilliant assistant
who lives in your house
NyxAI is like having a full-time personal assistant, business manager, and security guard — all running on small computers in your home. They know your schedule, help run your beauty salon, track your health, manage your finances, and protect your data with the same security banks use.
It lives at home
Unlike ChatGPT or Siri, NyxAI runs on physical hardware in your home. Your conversations, health data, and finances never leave your house unless you want them to.
You talk via Telegram
Send a message, get an answer. "What's on my calendar?" "Draft a reply to that client." "Log my blood pressure." It just works.
It's actually smart
It uses the same AI models as ChatGPT and Claude — but routes sensitive data to its own local brain, and only sends safe things to the cloud.
Why not just
use ChatGPT?
When you use ChatGPT, Google, or Siri — everything you type gets sent to their servers. They see it all. With NyxAI, sensitive data stays home.
Regular AI
- Health data goes to OpenAI's servers
- Financial info goes to the cloud
- Private conversations stored by Big Tech
- No control over your data
- If they get hacked, your data leaks
NyxAI
- Health data processed on your hardware
- Financial info never leaves your network
- Private conversations stay in your house
- You own the hardware, you own the data
- Cloud gets hacked — your secrets are safe
Three agents
work together
Think of it like a small company with three employees, each with a clear role. They communicate but never step outside their domain.
Nyx
The one you talk to. Handles questions, morning briefings, calendar, emails, health tracking, and salon operations. Warm, multilingual, proactive.
Kai
Writes all code. When Nyx needs a feature or automation, she hands it to Kai. He only writes — never deploys directly. Safe and controlled.
Guardian
Watches everything. Checks every command. Ensures sensitive data never leaks to the cloud. Monitors all four computers every five minutes.
What happens when
you send a message
Here's the journey of a simple request — say you type
"Log my weight: 82kg"
You send a Telegram message
"Log my weight: 82kg" — arrives at the Gateway, a small Raspberry Pi that serves as your front door.
Guardian scans it
Guardian spots "weight" — health data. Flagged as CRITICAL. This message will not go to the cloud. No exceptions.
Nyx processes it locally
The message goes to a small AI brain on an NVIDIA Jetson — a tiny, powerful computer with no internet connection.
Data saved in your encrypted vault
Your weight is saved to an encrypted file on your hardware. Even if someone stole the computer, they couldn't read it.
You get a reply
"Logged: 82kg. Down 0.5kg from last week 💪" — and nobody outside your house ever saw that data.
Four tiny computers
three security zones
Think of it like a house with different rooms. Some are locked, some have no windows to the outside, and one has a guard at the door.
🌐 Browser Room
Can look at websites, but can't access your private rooms. If a hacker attacks through a web page, they're stuck here. No keys, no passwords, no access.
📡 Gateway Room
The front desk. Receives your messages, coordinates everything, stores files, and runs the Privacy Proxy that scrubs personal info before it reaches the cloud.
🔒 The Vault — Two Jetsons
No internet access at all. These two computers handle everything sensitive — health, finances, security monitoring. They physically cannot connect to the internet. Your most private data lives here, encrypted, on hardware you own.
Nine domains of
intelligent assistance
Just message Nyx. She figures out the rest — routing each task to the right model, the right tier, the right security level.
Personal Assistant
Morning briefings, reminders, calendar, Q&A, daily nudges, decision support
Business Ops
Content for Alice Beauty Studio, client communications, booking management, brand strategy
Tech & Code
Automations, integrations, website work — Kai handles all engineering tasks
Finance
Expense tracking, budgets, invoicing, tax preparation
Health & Wellness
Health logging, habit tracking, medication reminders
Learning
Study help, language practice — Dutch, English, Ukrainian, Russian
Creative
Writing, brainstorming, gift ideas, event planning, content creation
Communication
Message drafting, thread summaries, follow-up tracking
Travel
Trip planning, packing lists, time zones — passport info stays local
It picks the right
brain for each job
Ten different AI models — from a free local one to the most powerful cloud models. Each task is automatically routed to the cheapest one that can handle it well.
🏠 Local Brain
Used for: Health data, finances, medication reminders, anything personal. Runs on the Jetsons. Costs $0. Slower but completely private.
☁️ Cloud Brains
Used for: Blog posts, research, coding, complex analysis. Goes through the Privacy Proxy which scrubs your name, address, and personal info first.
Enterprise-grade security
in plain English
Reviewed against the same standards banks and governments use. Five layers protect your data.
Network isolation
The Browser computer literally cannot talk to the Jetsons. A digital firewall sits between them. If a hacker gets into the browser, they can't reach your data. It's a separate building with no connecting hallway.
Identity scrubbing
Before anything goes to the cloud, a Privacy Proxy replaces your name with "[PERSON_1]", your address with "[LOCATION_1]". The cloud AI helps without ever knowing who you are.
Full encryption
Jetsons use full-disk encryption — if stolen, data is unreadable. Health and finance files use end-to-end encryption. Even the file server can't read them. Your vault key lives on a USB drive you keep yourself.
Anti-manipulation
Hackers try to trick AI by hiding instructions in documents. NyxAI has a Semantic Firewall that detects these attacks and a trust system that never follows instructions from untrusted sources.
Constant monitoring
Every five minutes, Guardian checks all four computers are alive, AI model files haven't been tampered with, and nothing unusual is happening. Something's wrong — you get an immediate Telegram alert.