GPT-5 has admitted its mistake. Every product manager makes mistakes in the process of iteration.

It's August 2025 now. There are only 4 days left until the press conference of vivo vision MR glasses. That is to say, on the 21st, China's first MR glasses will be updated.

There are rumors that this MR glasses is an MR device based on Android system with gesture and eye operation. It is said that this head-mounted display device, which is half as light as Apple Vision Pro, is coming.

I'm quite sure that after the release of vivo vision, we will truly enter the era of spatial computing. In the future, product managers of mobile apps will gradually be phased out.

Five Major Stages of Product Managers

Remember when I first became a product manager in 2016, it was just at the tail end of the mobile Internet boom. A few years ago, when you walked on the street, you could see people giving away gifts for scanning QR codes. As long as you downloaded WeChat, you could get a plush toy or follow a WeChat official account.

With the popularity of mobile social media, the marketing department even made budgets for increasing followers of WeChat official accounts. Many enterprises regarded the number of followers of official accounts as a marketing KPI. They could get a certain amount of marketing budget based on the number of followers.

1. In the first stage, the job of product manager didn't require specific educational background or major

At that time, when applying for a product manager job, whether you were an undergraduate or a junior college student, whether you were from liberal arts or science, as long as you learned prototype design and basic product design requirement methodology, you could apply for the job. And it was very easy for fresh graduates to get a monthly salary of 10,000 yuan in first-tier cities. I still remember when I first started my internship in TCL, which was to promote its app to hospital doctors. I could get more than 7,000 yuan per month just for the internship.

2. In 2016, those who chose to be product managers more or less wanted to be like Steve Jobs and become CEOs

I also switched to become a product manager for this reason. Because Steve Jobs meant changing the world, and becoming a product manager meant getting closer to changing the world, and then incubating my own enterprise and becoming a CEO.

Since investors have a very clear exit route, you can exit in the next round of financing or cash out by holding stocks when the enterprise goes public. So, startups doing apps could get financing in startup cafes just with a PPT. Even many young people with just prototype designs and project explanations could get investment and financing and participate in roadshows to meet various excellent investors.

3. After 2018, product managers began to be divided into front - end and B - end product managers

Many people became product managers influenced by people like Steve Jobs and Luo Yonghao, and most of them were C - end product managers. On the contrary, B - end product managers, dealing with more complex business, often found their work more boring.

So, learning to be a B - end product manager became popular. Due to the boring business of the B - end and the difficulty of finding competitors, many product managers need professional learning or specific projects to gain project experience as B - end product managers.

As the systems in various industries are gradually improved, the requirements for product managers are becoming more and more demanding. It's no longer just about developing software, but also about assessment. At that time, many teams were always wondering what the assessment criteria for product managers should be.

Since development and UI design are not done by product managers, what assessment indicators can be used for product managers?

So, there began to be C - end product managers specializing in apps or user experience designers, product managers serving internal company business or enterprises, called B - end product managers, and those designing information systems for business, called business - type product managers.

The assessment of the former is based on function usage rate, page stay time, and subsequent user acquisition and growth, while the latter is based on indicators such as NPS and favorable rate. The assessment indicators for product managers are clearly divided into C - end and B - end.

3. The explosion of data product managers: Making requirements based on data, not on a whim

In addition, as the amount of data gradually increases, the iteration of C - end apps increasingly relies on data optimization, which has given rise to data product managers. The popularity of data product managers at that time came from various data tools, from Baidu Statistics to Umeng and then to Sensors Data. The AARRR model became a professional tool for data product managers in Internet companies.

It indicates the optimization direction and iteration ideas, thereby helping enterprises achieve higher growth and revenue. At this stage, the app market has almost become a red ocean. Next, it's about competing in iteration speed and direction to quickly meet users' needs.

For example, after the latest update of ChatGPT 5, ChatGPT 4O was hidden. Due to users' complaints and dissatisfaction because they couldn't find the previous 4O, OpenAI had to roll back to the previous version and support model selection so that users could find 4O.

We can see that data is very important for the iteration and optimization of product managers. In this case, it's obvious that no data analysis was done. They didn't consider the distribution of the number of users using various models on GPT. Instead, they just directly launched GPT 5 on a whim. If data analysis had been done, they would have known that paid users are in the minority, and most users are still free. Therefore, it's crucial to retain users' habits of using 4O.

Big - data product managers can be regarded as a typical sign of the mobile Internet approaching its end. Because almost all kinds of ideas have their own apps, and there are leading players in each industry. Next, it's about how to reduce costs and increase efficiency in product operation, optimization, and iteration.

4. The arrival of artificial intelligence product managers

At that time, artificial intelligence was not today's large - language models, but was based on the Python language, which made Python the most popular language. Big - data analysis and artificial intelligence have given rise to many algorithm engineers, especially in fields such as image recognition and voice recognition. Python training courses were very popular.

Almost everyone was learning Python, and Python was regarded as the best language in the world.

5. The explosion of mini - program product managers gave rise to social fission

In 2018, when Zhang Xiaolong launched WeChat mini - programs at the WeChat Open Class, front - end product managers were further divided, and there emerged mini - program product managers. Even many early mini - programs looked exactly the same as apps. As users' usage scenarios were different, one was in WeChat and the other was a native app, product design specifications for mini - programs were gradually developed. Thus, mini - programs contributed to the traffic growth of Luckin Coffee at that time. Through mini - programs, Luckin Coffee quickly acquired customers. So, mini - programs redirected traffic to apps or even to WeChat groups, and social e - commerce, such as Pinduoduo, was born. It quickly attracted a large number of users in the sinking market through WeChat's "help - cutting" and group - buying features.

In 2019, companies that achieved growth through social fission based on WeChat mini - programs also included Luckin, the company that went public the fastest in history. It acquired users through massive fission methods and expanded its stores. At that time, a book called "Traffic Pool" became a must - read for almost every Internet product manager. The fission growth logic in it became a low - cost user - acquisition method that many product managers had to learn.

Luckin Coffee, as one of the most representative brands in China's new consumption field, its development process can be regarded as a "roller - coaster" in China's coffee industry. It only took 18 months from its establishment at the end of 2017 to its listing on the NASDAQ in May 2019, setting the world record for the fastest IPO.

As the capabilities of mini - programs were gradually expanded, with the addition of live - streaming, payment, and stores, mini - programs became an important front - end application touchpoint in the industry. Many companies began to focus on mini - programs instead of apps.

Because mini - programs have lower development costs and faster development speed.

6. The birth (and fastest disappearance) of metaverse product managers

With the rise of VR and the renaming of Facebook, metaverse product managers gradually emerged. The metaverse conference of Baidu became a direction for product managers to transform. Learning VR and 3DMAX production became a goal for many product managers.

The metaverse became a new wave, and related technologies and concepts such as blockchain and the metaverse became popular. However, unfortunately, there were no distinctive products in the metaverse. On the contrary, blockchain gave birth to many well - known wallets. Even due to the open - source code, many early product managers became holders of blockchain products.

However, since there has been no practical application scenario in China, as the popularity faded, related products were gradually taken off the market because there were really no outstanding products. After Facebook was renamed Meta abroad, it launched the related VR device Quest and has really been making arrangements in the Meta operating system and open - source, which continues to this day.

During the years of the pandemic, almost no new technologies and products emerged. Instead, new social media platforms like Xiaohongshu and Douyin were born, attracting users with content consumption and short - video habits based on recommendation algorithms.

7. With the emergence of ChatGPT 3 in 2022, there appeared AI product managers based on large - language models

When AI first emerged in China, many product managers just used AI as a shell for applications, such as achieving basic Q&A functions by connecting to AI. However, product managers with forward - looking vision began to find that AI models could be used as shells or open - sourced for secondary training to make up for the shortage of domestic large - language AI models.

For example, LLMA, which adhered to open - source from the beginning, was the earliest open - source model. Many early domestic startup companies used it as an open - source model for secondary training to create domestic models and attract a large amount of financing.

Thus, product managers of model manufacturers gradually emerged. With the application of AI +, some product managers began to be involved in AI functions and started to develop shell - type applications.

However, after the launch of the GPT app store, many teams doing shell - type applications went out of business. This made many people afraid to start AI - related product startups, fearing that model manufacturers would also enter the same field.

However, as model manufacturers compete with each other, today's Agent and native AI applications all rely on AI training resources, such as GPU graphics cards, which has led to a sharp increase in the price of NVIDIA graphics cards. Now, AI product managers are clear about their direction, which is to focus on the AI application layer, especially native AI shell - type applications.

From the changes of product managers in the past decade, we can see that product managers are always in a state of learning and transformation. If they don't transform, they will be eliminated, which confirms what Ma Huateng once said: "In the IT industry, you may not have done anything wrong, but you're just too old."

And the next step is the great explosion of brain - computer interfaces and spatial computing

After the development of 3D intelligence and brain - computer interface capabilities, AI product managers will replace the mouse and keyboard and change the display from a flat screen to a spatial one. This will be the next major development for product managers. Just like I'm writing this article in Vision Pro now.

After Tencent's Hunyuan 3D model was open - sourced, it supports the generation of spatial videos in minutes and allows interactive operations. The clarity has reached over 1K, which is much clearer.

On August 21st this year, vivo will also launch China's first domestic Vision Pro - like device. According to information from the MR community, the experience can almost replace that of Vision Pro. So, I believe that in the next 2 - 3 years, MR glasses will become a new terminal development direction, corresponding to a new form of WeChat - like application, which will give rise to new social interaction methods and applications.

This article is from the WeChat official account "Kevin's Little Efforts to Change the World" (ID: Kevingbsjddd). Author: Kevin's Stories. It is published by 36Kr with authorization.