Original video in Chinese.
Key Takeaway
- ChatGPT isn’t very efficient on the iPhone because it isn’t integrated with the system or other apps, so it can’t run automatically.
- By using the iPhone’s Shortcuts app, you can automate ChatGPT tasks such as automatically summarizing WeChat public account articles, voice Q&A, and content creation.
- The core logic of Shortcuts is: input (webpage text, speech-to-text) -> processing (ChatGPT summarization, answering) -> output (Markdown-format notes into Notes).
- This kind of automated workflow can significantly improve the efficiency and convenience of using ChatGPT, making up for its limitations as a chatbot.
- Through the three Shortcut examples of “Summary GPT,” “Q&A GPT,” and “Inspiration GPT,” this article shows how ChatGPT can be used in personal productivity.
Full Content
ChatGPT still isn’t good enough to use on the iPhone. The main reason is that it isn’t integrated with the system or other apps, so it can’t run automatically and isn’t very efficient. So for now, it’s still just a chatbot—you have to open the app every time you want to use it, and then you still have to type in a bunch of stuff. No wonder people don’t use it that often; it’s just too much of a hassle.
At this stage, if you want ChatGPT on the iPhone to be useful, you definitely need to use the Shortcuts feature. I’ll first show you how I use it, and then explain in detail how to set it up.
I’m sure everyone finds this kind of thing overwhelming: there are way too many WeChat public account articles to read every day. What do you do?
First, open this article in Google Chrome.
Then, click the Share button to the right of the address bar and choose the top Shortcut option (the one I set up is called “Summary GPT”).
Finally, wait a moment, and ChatGPT will, according to the requirements I set in advance, use GPT-4o, the latest model, to summarize the core content of the article, extract the Key Points, and create a new note in the specified folder in Notes, saving everything there in Markdown format.
The benefits of doing this are:
First, I don’t have to manually switch between apps and type all kinds of things—now it’s done automatically with Shortcuts, which is obviously much more convenient.
Second, although ChatGPT has chat history, I still have it saved in Notes. That’s because Notes is built into the system and syncs very quickly between iOS and macOS. That way, when I go back to my Mac, I can edit it directly. And because it’s in Markdown format, I can also paste it into writing software or note-taking software for further processing.
What I just showed you was “Summary GPT.” Following the same idea, I also set up a “Q&A GPT,” which also uses voice input.
For example, I ask: what are the differences among the mainstream switches for mechanical keyboards? Like red switches, blue switches, and so on.
After receiving a voice input, this Shortcut first converts the speech into text and sends it to ChatGPT.
Then, according to my pre-set configuration, after ChatGPT answers the question, it also saves the content in Markdown format to Notes, making it easy for me to find and edit later.
These two Shortcuts are the ones I use most often. In addition, there is also “Inspiration GPT”:
Whenever I have any content-related ideas, I, just like with “Q&A GPT,” feed everything to GPT by voice. It will organize my scattered expressions logically, turn them into a content framework, and then add and expand on it.
If you watched my last video, you’ll find that this is actually using Shortcuts to build a simple Agent Workflow. Before the new version of iOS and the new version of Siri come out, let’s use this approach to achieve AI automation.
Next, I’ll use that “Inspiration GPT” from just now to explain it in detail. I’ve already shared iCloud links for these three Shortcuts in Knowledge Planet and Patreon, and friends in the newtype community can just take them and use them. I’ll also premiere this video in the community, as an early screening for everyone.
As I said in the previous video, generally speaking, the starting point of an Agent Workflow is user input. In “Inspiration GPT,” we convert speech to text, so we need to use the “Dictate Text” function. In the settings for “Dictate Text,” remember to open the dropdown menu and select a language, otherwise it will error out.
After the initial input is in place, the second step is to add a “Text” action and write a Prompt inside it to clearly explain the requirements for the AI, then paste the content from step one after it, just like I did. This entire text is the information that will then be sent to the AI for processing.
Third, add “Ask ChatGPT” and include the “Text” action. In the dropdown menu, check “Start New Chat,” and then you can choose which model to use.
Fourth, add another “Text” action and put all of GPT’s answer into it. You can add a description at the beginning of the text, like I wrote “From GPT,” so I know this note was generated by AI.
Fifth, add “Notes,” choose the folder you created in advance, and have GPT create a new note there and save all the text from the previous step into it, and that’s it.
The other two Shortcuts follow the same idea. In fact, the core of this whole workflow is just three steps, which is easy to understand:
First, where does the initial information come from (input)? Is it text on a webpage, speech converted to text, or a photo? This needs to be set up clearly.
Second, what do you want GPT to do for you (processing)? I usually add a “Text” action, put the fixed instructions and the specific requirements for each run into it, and then send the whole text to GPT.
Third, what kind of result do you want (output)? Just like what I demonstrated earlier, use Notes plus Markdown as the output.
Once you’ve thought through input, processing, and output, you’ll know how to use Shortcuts and ChatGPT to build a simple workflow.
OK, that’s it for this episode. If there’s anything you want to talk about, come find me on newtype. See you next time!