ChatGPT Atlas browser is here, and product managers are no longer needed... (Proficient in prototyping, Xiaohongshu operations, development, and project management)

Just this morning, ChatGPT started to roll out an update, allowing users to download it for browser use. A pop-up window appeared on the ChatGPT conversation page, and I could download it, so I did.

I completed the download and installation right away.

Here is the download and installation interface.

After the download and installation are completed, you can see its mouse logo, named Atlas. "Atlas" is a name from Greek mythology, meaning "the one who supports the heavens and the earth". It can be seen that this browser truly lives up to this meaning.

Through the following task examples, I believe you can understand the Agent mode of the AI browser, which is like the future's Jarvis and an office mode. Based on my daily scenarios as a product manager, I conducted several tests on this browser, including automatically writing official account articles, prototype design, project management, and opening common web pages in the browser.

1. Posting on Xiaohongshu

Before opening a third - party application, the first - time use will prompt a security risk, indicating the relevant risks of authorizing and logging in to ChatGPT.

Authorize and log in to Xiaohongshu:

Start posting on Xiaohongshu:

Log in to the Xiaohongshu creation platform:

Complete the post on Xiaohongshu:

You can see that it helped me write the steps of the automated operation manual for automated posting on Xiaohongshu. Meanwhile, the entire task was completed through conversations. Each conversation would trigger the next operation through the visual model, and the completion status of the task was determined.

During the task, it will prompt whether to let ChatGPT take control of Xiaohongshu:

It supports button locking and one - click typesetting. You can choose the typesetting option, and let the AI click a certain button to complete the operation and publish the post.

Here is the final successful publishing effect. You can see that by using Xiaohongshu's official automated typesetting, the AI browser completed the publishing after clicking the typesetting button.

During this process, the user can choose to continue being taken over or pause the takeover.

However, at present, there are still defects in inserting pictures into the notes. The user needs to prepare the pictures themselves for the AI to insert. The Agent cannot generate pictures on its own and then upload them.

Let's try an upgraded task. Let the AI help me open my Xiaohongshu notes with relatively high data volume. Here is the task filtering for the notes.

The browser Atlas opened the browser and found my notes with high reading volume (presumably analyzed by the visual model). It can help sort them out, click and filter for me.

Above are the notes with high reading volume sorted out for me. You can view the corresponding URL links of the thumbnails in the dialog box.

2. Automated Twitter Posting

Here, the same task is to complete the automated posting on Twitter. After entering the task command, it starts to open Twitter and write a message.

After the successful posting, you can click to view the message and see its content. As shown in the following picture, after completing the corresponding task, you can see that the automated posting of overseas products is very smooth and can be published more quickly.

3. Automated Prototype Tool master GO, Search for UI Design Diagrams

Here, I started the automated office task test. I accessed the prototype tool masterGO, which I often use as a product manager. After authorizing and logging in to master go, I can view the corresponding projects and the UI files in the projects.

Like Xiaohongshu, the operations related to login and registration will be left to the user. Then, it needs to open the corresponding prototype page design and view the UI design diagrams.

Here, the AI browser in Agent mode reaches the corresponding UI design diagram page to start viewing the corresponding data pages and searching for data pages. What surprised me the most is that there are multiple vision PRo projects in the project, and it can open each project step by step for retrieval.

As shown in the above picture, it is the retrieval process and steps. You can see that it can almost help me quickly locate. In the past, when collaborating on UI design, as a product manager, I often couldn't find the corresponding design diagrams and project addresses. Now, with AI, it can almost completely help you locate. It can search for UI design diagrams, and even if the UI design diagrams are not named, it can be found through image recognition.

4. Automated Project Management Tool teambtiion

To open teambition, you need to register and log in first. After completing the authorization, you can log in. Here is the corresponding task creation and card progress monitoring.

Through the Agent mode of the AI browser, you can quickly complete the task creation and status tracking of each card. Even if there is no text for creating the card, only text for editing, the AI can still recognize it and complete the card creation.

5. Automated Use of figma maker

Similarly, I started using this browser to use figma. Different from the prototype tool, this time I used the AI browser to use the AI function, that is, using AI with AI. Here is the process of opening it.

What's