Nomon is a free browser extension that notices when you’ve stopped evaluating what the AI gives you & quietly says so, in one line under your prompt. A mirror, not a nanny. It never interrupts, never blocks a reply, and never reads what you write.
Works with ChatGPT, Claude, Gemini, Grok, Copilot, Perplexity, Mistral, Meta AI, DeepSeek, Qwen, Kimi, MiniMax, HuggingChat and Doubao · Everything stays on your device
This is a call only you can weigh. Want 30 seconds to note your own read before the answer loads? Still your call.
Three quiet mechanisms. No score chasing, no streaks, no guilt.
A lightweight extension watches how you and the AI are talking — questions versus commands, pace, pushback. Never the content itself, and nothing leaves your browser.
When you stop evaluating, a one-line strip appears under your message. One of five signals, each with its own name and meaning. Never a banner, never a block.
Patterns roll into one card — your shape and your trends, in plain words. No raw scores, no leaderboard, nothing to grind.
Your habits with AI aren’t per-tool, so the mirror isn’t either. Nomon works inside the chat interfaces you already use, and your weekly card reflects all of them together. Switching tools never resets the picture.
One warning light can’t mean five different things — so Nomon doesn’t use one. Each signal has its own name, colour and voice, and reads as quiet data under your message. There is no red anywhere.
You’ve delegated a whole task in your first message. A gentle prompt to put down your own starting point first.An invitation, never a gate.
You’ve accepted several answers in a row without editing or pushing back. In-session, right now.Noticing, not judging.
Across sessions, you’re asking less and accepting more. A longer-term pattern.Lives in your weekly digest, never as an alarm.
This conflicts with a goal you set. Nomon just quotes your past self back — it never decides what you should protect.Only ever fires on goals you wrote yourself.
A high-stakes question where the thinking is the point. An invitation to a beat before the answer loads.The answer still arrives — never delayed.
One dial, four settings — one for each dot in the mark. Switch any time; the panel on the right shows exactly what you’d see.
The default. Subtle inline strips for Loop and Drift beside your messages — never a pop-up, never a card.
The first three modes never hold you up. Guard is the only one that can pause before send — you switch it on yourself, it only acts on goals you wrote, and it is always bypassable.
Every signal rolls up into a single, shareable card — your shape, your trends, and one question worth sitting with. Shapes, not rankings: there are no raw scores and no leaderboard, because the point is comparison with your own last week, not with strangers.
Explorer · steady
Mostly research and learning mode. You pushed back a little more than last week.
One to sit with: what did you figure out yourself this week, without AI?
Fair questions, direct answers. Nomon exists to counter cognitive offloading — the well-studied habit of leaning on a tool until you stop thinking for yourself. And if a tool watches how you use AI, you should be suspicious of it by default. Here’s exactly where the lines are.
No. Nomon reads the shape of a conversation — message length, question-versus-command patterns, pace — never the words. Message text is not stored, analysed as content, or transmitted.
Not unless you switch it on. All scoring runs locally in your browser. Contributing anonymised counts to research is off by default, opt-in, visible in a transparency view, and revocable — with full deletion on request.
It’s built not to. Signals are one-line strips under messages you’ve already sent. No pop-ups, no red, no streaks, no guilt mechanics. If it ever feels like a nag, that’s a bug in the product, not a feature you have to manage.
No. Nomon measures how you use AI, not how much. Use AI as often as you like — heavy use with your judgement engaged is exactly the point.
Only if you ask it to. Guard mode — off by default — briefly holds a message that clearly conflicts with a goal you wrote, and even then one click sends it anyway. The other four modes never touch your messages.
Ghost mode. One click and nothing appears in-session at all — you only get the weekly digest. It’s a first-class mode, not a buried setting.
A researcher, not a growth team. Nomon is built in Edinburgh by Dr Adam Chalmers, a University of Edinburgh academic working on computational social science and AI governance. All feedback can be directed to hello@nomon-app.com.