Press and hold the Action button, and Grok starts reading your screen: the demo is cool, but screenshots go to a third-party image host first

Elon Musk reposted a demo from Chris: the Grok bot can now see what's on your phone screen; hold the Action button and ask directly.

Chris's three examples are very specific: ask about the video you're watching; have Grok turn what it sees in a Reel into a plan for your business; have it break down the article you're reading. He added that the Grok bot can watch videos, read posts, and look up web pages.

Chris made it into a free iPhone shortcut and open-sourced it. Free entry point here: https://x.ai/bot/yc8yWwlDJLRwr0JuyHRhA

Someone asked what to do on an iPhone 12 Pro without an Action button. Chris replied that the shortcut is fully customizable and can also be changed to voice-reply while scrolling.

Privacy concerns came up quickly in the replies. TheSameCat pointed out an implementation detail: screenshots are passed to the routine as temporary imgbb links because base64 images in the webhook payload would be truncated. In other words, every screenshot is first uploaded to a third-party server. Some users said privacy has left the chat, and others worried that this data could be stolen or exploited.

Community reactions split several ways. Some asked for official native integration, some asked about phone call features in the EU, and others started playing with bot avatars and Groklet. Dominik Scholz said he built an avatar tool because you can't create new cute bot characters in Grok Bot, and tagged @poteto to ask for an official launch.

Some users also asked whether the Grok bot could diagnose PC problems.

This feature turns the phone screen into an input box, and the path is very short. The cost is also clear: system permissions weren't opened up, yet screenshots still go through a third-party image host and a shortcut. If you want to try it, you can use the free shortcut; if you want to be safe, wait for native integration.