Apple one’s own Apple Intelligence The development progress seems to be behind expectations. So far, no breakthrough features have appeared, but through third-party models,iPhone AI automation workflows are still achievable — recently someone shared a tool built with the GPT 4.1 model. It’s not the kind of app you can install directly from the App Store, but it’s free and open-source on GitHub, so anyone who knows how to run it on an iPhone can give it a try.
Overseas developers have created an iPhone AI automation tool based on GPT 4.1.
A developer named Rounak Jain built this tool at the 2024 OpenAI hackathon, and starting this year, he has made the open-source project available on GitHub for everyone to access for free.
From the video he demonstrated, users can use this open-source tool to issue voice commands to have the AI carry out specified tasks across applications—for example, taking a photo with the camera first, then sharing it with others through another app, all without pressing any buttons.
Rounak Jain first demonstrates “taking a new photo and sending it to Ron with a haiku about the weekend”:
The AI then activates the camera to capture the current view, uses the share feature to find Ron and launch iMessage, generates a haiku in the input field, and sends it—all without him having to lift a finger.
It also has a built-in “always-on mode,” where the AI continuously listens for commands without needing to open the tool. Rounak Jain gave the following command: “Agent, message Ron that I’m flying DL1715 to Los Angeles, and call an Uber X to SFO.” The entire workflow then becomes: open iMessage to send the specified message to Ron, then launch Uber to arrange a ride to SFO.
The video indeed demonstrates the AI operating the entire process, with Rounak Jain not doing anything at all:
Opening Uber to hail a ride is pretty impressive, and what’s particularly interesting is that for the text “SFO,” the AI seems to have used paste:
AI also successfully selected the correct Delta Airlines by identifying the flight number DL1715:
In addition to working with multiple apps, this tool can also perform system-level operations.
Such as requesting, “Open Control Center and turn on the flashlight.”
Then, you can reply accordingly through the notifications it sends. Here, he is asking to turn off the flashlight:
Rounak Jain stated that running this tool only requires an OpenAI API key and Xcode—no jailbreak needed, so you can use it with peace of mind.
This is the complete video:
With the appearance of this project, it’s also clear that AI automation is starting to develop not just on computers but on mobile as well. Of course, some automation features can already be done through iPhone’s Shortcuts, but with AI backing in the future, not only will operations be simpler (no need to drag things around in Shortcuts), but there will certainly be much more you can do.
Source: KOCPC Chinese