On December 20th, the 12-day launch event of OpenAI has entered its 11th day, which is the second-to-last day. The company has released a MacOS desktop application and its interoperability features with various applications. This will lay the foundation for the future of "Agentic AI" (or intelligent agents), making ChatGPT more powerful and seamlessly integrated into users' daily workflows.
On the 11th day of the launch event, OpenAI's Chief Product Officer, Kevin Weil, along with two colleagues dressed in Christmas-themed suits, discussed the company's newly launched MacOS desktop application. They emphasized the transformation of ChatGPT from a simple conversational assistant to a more powerful agent tool, meaning that ChatGPT can now represent users to perform more tasks, bringing unprecedented convenience to users.
1 Three Major Features Introduced
Currently, users can view and automate their ChatGPT work through the MacOS desktop application. Although more similar versions will be released by 2025, OpenAI has already introduced the following three major features:
First, with the "Work with Apps" feature, users can now integrate ChatGPT into more coding applications, including BBEdit, MatLab, Nova, Script Editor, TextMate, Android Studio, AppCode, CLion, DataGrip, GoLand, IntelliJ IDEA, PHPStorm, PyCharm, RubyMine, RustRover, WebStorm, Prompt, and Warp.
In the demonstration of the MacOS desktop application, OpenAI showed how AI can delve into an application, acquire and understand its context information. Once the user selects an application through the "Work with Apps" feature, ChatGPT can immediately integrate, gain insight into the application, and provide instant assistance.
Of course, ChatGPT is not just a simple viewing tool; it relies on a powerful AI model to perform various functions. In the demonstration of Warp, ChatGPT not only can capture the content on the user's screen but also can delve into the application and browse more information. For example, when processing long code, ChatGPT can achieve seamless scrolling, greatly improving work efficiency.
Compared to the Windows Recall feature, ChatGPT focuses more on real-time collaboration with applications, rather than just recording and building a memory library. In another demonstration, the OpenAI team closely integrated ChatGPT with XCode, allowing it to work within Apple's development application. Users only need to make a simple request, and ChatGPT can generate code or solve programming problems.
It is worth noting that OpenAI also demonstrated a new skill of ChatGPT: it can directly embed the generated code into XCode, a feature that is expected to greatly simplify the workflow. Although in the live demonstration, ChatGPT's code attempts encountered two failures, the OpenAI team successfully ran the code on the third try.
Second, for users who utilize ChatGPT for writing, OpenAI announced that the MacOS desktop application now supports Apple Notes, Quip, and Notion. In the live demonstration, the OpenAI team was browsing a document aimed at creating a guide for a hiking activity in Notion.
With this new feature, ChatGPT can seamlessly collaborate with Notion. The live demonstration focused on a specific text segment in the document and set the task to "supplement these talking points." Additionally, users can utilize ChatGPT's search function to generate responses. For example, in the demonstration, it generated talking points about "Emperor Norton (Norton I)" based on the selected text, including citations and sources.
Third, in addition to traditional text selection, copy, and paste operations, the MacOS desktop application supports advanced voice mode and can work with other applications. In this mode, users can set a "Holiday Party Playlist" in Apple Notes and consult Santa Claus through ChatGPT on the candidate songs. ChatGPT can even point out the user's mistakes, such as mistakenly writing the Christmas song "Frosty the Snowman" as "Freezy the Snowman".
These features have been officially released, and users only need to ensure they have the latest version of the MacOS application and subscribe to any of the ChatGPT Plus, ChatGPT Pro, ChatGPT Team, ChatGPT Enterprise, or ChatGPT Edu services to experience them immediately.
Regarding privacy protection, OpenAI emphasizes that ChatGPT will only interact with applications when manually triggered by the user. Once the feature is activated, the user will be clearly aware of which content will be appended to the message, effectively alleviating privacy concerns.
2 AGI Teaser Revealed Again
Since December 5th, local time in the US, OpenAI has launched an intensive new feature release cycle, planning to introduce new products and features through 12 live events over the next 12 days. Prior to this, OpenAI has already released a series of innovations, including the ChatGPT Pro plan, reinforcement fine-tuning technology, Sora, the interactive interface Canvas, advanced voice and visual features, Projects feature, ChatGPT search, the fully-powered o1 model, opening the o1 series large models to third-party developers through APIs, and interacting with ChatGPT via phone and WhatsApp.
As the launch event approaches its conclusion, people's attention to AGI (Artificial General Intelligence) is also increasing. At the end of the 11th day's event, OpenAI stated: "On the 12th day, we have prepared extremely special content, so don't miss it!"
In the corner of the demonstration screen, a folder named "AGI_Interface.swift" can be seen. This is not the first time such a surprise has appeared in the past 12 days. A few days ago, OpenAI also unveiled a calendar event Easter egg called "Super Secret AGI", which has undoubtedly further heightened people's expectations for these 12-day series announcements, and many are speculating whether these announcements are collectively painting a grand blueprint towards general intelligence.
OpenAI also revealed that the Windows application for ChatGPT will be released soon. But the more shocking news is that they have confirmed the existence of a new intelligent agent and expect to release it by 2025. OpenAI stated: "As our models become more and more powerful, ChatGPT will exhibit increasing autonomy."
A few weeks ago, there were rumors that OpenAI was developing an Agentic AI called "Operator", and the company only confirmed this plan during the 11th day's launch event. Perhaps this move was influenced by pressure from competitors.
Recently, Google announced Project Mariner, an intelligent agent that can navigate and perform actions on web browser tabs on behalf of users. Similarly, Microsoft has launched Copilot Vision, which can view content in the user's web browser and provide relevant information. Of course, Anthropic had previously released the Computer Use feature, which is ahead of other similar tools in terms of timing.
Now, with only the last day of OpenAI's 12-day series of events remaining, they seem to have saved the most exciting part for last - a brand-new and powerful frontier model is about to be unveiled. We will wait and see what new product OpenAI brings, and how this new model differs from the previous o1 model.
It is worth mentioning that some benchmark tests have already shown that the o1 model is one of the most powerful AI models to date, even surpassing Claude 3.5 in coding tasks. Recently, a user on the X platform reportedly discovered the GPT-4.5 model, although the model currently only provides limited preview functionality.
Now, all eyes are on OpenAI, and everyone is eagerly awaiting to see what surprises they will bring on the last day of the launch event.
This article is from the WeChat public account "Tencent Technology", author: Tencent Technology, translator: Jinglu, published on 36Kr with authorization.
