Cortex
A personal experiment.
Screen capture with surrounding context.
Cortex is a screenshot tool that tells you what to do with your screenshots. Instead of leaving hundreds of unnamed images in a folder, it reads the text in each capture, works out what it is (something to buy, a to-do, an idea, an article to read later, a piece of code) and files it with a title, a one-line summary and a priority.
It can also remind you to come back to a capture, so a screenshot becomes a task instead of clutter.
- 01
Capture from the tray
A desktop app (Electron) sits in the system tray. Quick Capture dims the screen so you can drag a box around any area.
- 02
Read the text
Tesseract.js runs OCR on your device to pull the text out of the image.
- 03
Classify it
An AI model, through Puter.js, sorts the text into one of seven categories: Buy Later, To-Do, Idea, Read Later, Inspiration, Code or Ephemeral. It suggests a title, a summary, a priority and when to remind you. A simple built-in classifier takes over when the AI is not available.
- 04
Review and be reminded
Captures are saved on your computer and appear in a dashboard with stats. Desktop notifications bring back the ones that matter.
Captures are stored on your own computer. Only the text read from an image is sent to the AI for sorting.