Your files stay home. The AI moves in.
Learn local AI by building ten real apps with QVAC. Every lesson is a video, a written walkthrough and a tested commit, and nothing you open ever leaves your machine.
Be honest. You've pasted something private into a chat window.
A screenshot with a client's phone number. A contract with an address. The AI found what you needed in a second, but it read everything on someone else's server first.
The usual way
Your file travels to the model.
Bring AI Home
The model travels to your file.
of people using public AI assistants at work admit entering sensitive information into them. Names, addresses and phone numbers top the list. TELUS Digital, 2025
Every lesson breaks something first.
You start each project with an app that already works without AI. Then you add one QVAC capability at a time, and you watch it fail before you fix it.
Break it
Start from the last checkpoint and reproduce the limitation on a real machine. Rotate the page, stop the service, pull the network.
Find the layer
Is it the model, the transport, the geometry, the state or the network? Most AI bugs aren't in the AI.
Change only that
One tested commit per lesson, with a patch you can apply and a written walkthrough of every file it touches.
Prove it
Every lesson ends with a receipt: the smallest test or measurement that shows it worked, so you never have to take my word for it.
offline-receipt.json
offline
Redact is finished. All 11 lessons are up.
Ten apps, one house.
Each project lives in a room where your private data already sits. Redact guards the front door, because it decides what leaves.
The model proposes.
The app checks.
You decide.
OCR becomes evidence, not a redaction.
A parsed receipt becomes a draft, not a balance.
A camera frame raises an alert only on a clear yes.