Star on GitHub

ai;dr

AI; didn't read.

When someone else's AI pads one sentence into six paragraphs, ai;dr summarizes it for you, so you don't have to read the whole thing — using a tiny model inside the app, on your Mac.

What you copied

ai;dr
A local model summarizes anything you copy.

Runs in-process with llama.cpp, so nothing leaves the Mac.

Shows for ten seconds, or until you click it. No focus stolen.

ai;dr lives in the menu bar and watches the clipboard. When you copy a hundred characters or more — an email thread, a changelog, a terms-of-service page, a document like this one — it hands the text to a small language model (Qwen3-0.6B) running in-process through llama.cpp. There is no server, no Ollama, no API key, and no network call.

The card then drops from the menu-bar icon with a bold one-line headline and a short summary, stays for ten seconds or until you click it, and never steals focus from what you were doing. If you miss one, the menu has "Show clipboard summary" to bring the last one back, and "Copy last summary" to paste it somewhere else.

Nothing is uploaded, because there is nowhere for it to go. The model file ships inside the app bundle, inference runs on your machine's own silicon, and the app makes no network requests at all — so it works on a plane, on a locked-down laptop, or with the Wi-Fi switched off entirely. Your clipboard is never modified, and no summary is written to disk.

The hundred-character floor exists so the app stays quiet during ordinary work: copying a file path, a variable name, a URL, or a one-line command produces nothing at all. It wakes up for the things you actually skim — release notes, a wall of Slack backscroll, a legal paragraph, a stack trace, the paragraph a colleague pasted with "thoughts?" attached.

The download is around 680 MB, and almost all of that is the model itself rather than the app. It is signed with a Developer ID and notarized by Apple, so it opens on first launch like anything else you install. Apple Silicon only, because the inference is built against the Metal backend.

select · ⌘C · that's the whole product

In-process

llama.cpp runs Qwen3-0.6B inside the app binary. No server to start, no daemon to babysit, no API key to paste.

Private by construction

Your clipboard never leaves the machine — there is nothing it could be sent to. Works offline.

Zero ritual

No window, no typing, no buttons. Copy text like you already do; read one line instead of four paragraphs.

Download for macOS

Apple Silicon · ~620 MB, model included · open the disk image and drag ai;dr to Applications.

Or run from source

git clone https://github.com/ronreiter/aidr
cd aidr
poetry install
./aidr