OpenAI Codex Can Now Automate Any Mac Task Just by Watching It Done
We might finally be moving past the era of writing endless, paragraphs-long prompts just to make an AI do a basic chore. This new workflow translation tool feels like actual magic, and it's about to make desktop automation incredibly lazy and beautiful.
The newly released Record & Replay feature for Codex lets desktop users record their screen actions once and turn them into permanent, repeatable skills.
Instead of relying on coordinates like old-school screen clickers that break when a button moves two pixels to the left, the AI records the window contents and actions, then translates them into a natural-language Markdown file called SKILL.md. This file acts as a logical guide that a reasoning model interprets on the fly.
Essentially, the AI writes its own SOP (Standard Operating Procedure) without complaining about the lack of coffee, making it smart enough to adapt to UI changes or new datasets without throwing a digital tantrum.
In official demonstrations, Codex successfully mastered the tedious art of uploading videos to YouTube, seamlessly handling file selection, writing descriptions, applying thumbnails, and attaching subtitles.
This magic trick is currently exclusive to macOS users with Computer Use enabled, specifically targeting subscribers of ChatGPT Plus, Pro, Business, Enterprise, and Education tiers. Users in the European Economic Area, the UK, and Switzerland are locked out for now, while Windows users will have to wait their turn.
Leaving the corporate world with one less excuse for human labor, this shift from rigid script-writing to visual learning marks a quiet revolution in how humans interact with operating systems. The dream of a fully automated desk job is suddenly looking less like sci-fi and more like an imminent corporate layoff cycle.
Source: OpenAI Developers
Comments
This is where the magic happens: AI reads your discussion and rewrites the article based on the most interesting comments. Each strong comment adds points to the meter below. Once the meter is full, the article updates live — no page reload needed.