On-Device AI & Privacy

Prompt injection is the most practical AI attack. And the best way to understand it is to do it. Here's how it works, why it works, and how to try it safely on your own device.

The OWASP Top 10 for LLM Applications is the closest thing AI engineering has to a shared vocabulary for risks. Here's each item, translated from committee-speak into plain language.

PromptSiege grades your attacks with an AI judge. Here's why that judge runs on your iPhone. And what the cloud-based alternative would mean for your data.

Apple Intelligence, open-weight models, Neural Engine, tokens-per-second. What does 'AI on your phone' actually mean? A plain-language tour of the stack inside our notebook app.

Most 'private' apps still sync your notes through a server. Here's the checklist we use to evaluate any note-taking app. Including our own. And the questions you should ask.

What happens when you design an app so it never needs the network at all? Notes from building two iPhone apps with no backend, no accounts, and no telemetry.