AI Usage Monitor and Skill Protection

AI is great at giving answers. But if you hand it the thinking too — the judgment, the reasoning, the framing — the underlying skill quietly fades. This extension helps you keep it. It watches how you use AI (currently on Claude), and at the moments that matter most — a real decision, a judgment call, a piece of persuasive writing — it gives you a brief, optional nudge to bring your own thinking FIRST: your hypothesis, your experience, your key points. Then it asks the AI to challenge your thinking rather than just agree with it. It also keeps a simple private dashboard of how you're using AI, and a light "testing your learning" check that resurfaces things you looked up earlier so they actually stick. WHY IT'S DIFFERENT Grounded in learning science: retrieval practice and the testing effect (actively recalling or forming your own answer builds durable skill), desirable difficulty, and debiasing (asking for the opposite view to counter AI's tendency to agree). It only nudges on high-stakes prompts — never on quick lookups or formatting — so it stays out of your way. Private by default. Your history is stored locally in your browser. You choose how classification happens. YOUR PRIVACY, YOUR CHOICE (three modes) Local: a built-in classifier runs entirely in your browser. Nothing leaves your device. Your own key: uses your own Anthropic API key, sent directly to Anthropic from your browser. Shared server (default): uses our hosted classifier with daily limits. Your prompt is classified in transit and never stored. We never store your prompt text on our servers, never sell your data, and never use it for ads. You can turn nudges on/off, turn prompt-text storage on/off, and switch modes at any time. This is an independent personal project, not affiliated with Anthropic or Claude.