Product managers who have not mastered front-end and back-end development skills by now are already outdated.

Yesterday, I talked with the developers on our team about how product managers can now become developers, as well as the career development dilemmas faced by programmers. One of our developers mentioned that the boss of their previous company can now independently complete both front-end and back-end development all by himself.

Yet I am still only working on front-end development, which is totally out of date!

It can be seen that the CEOs of those profitable tech companies are now able to independently complete product development, and even build usable applications on spatial computing platforms such as Vision Pro and Android XR without relying on programmers. This was almost impossible in the past, but the situation has now changed.

When I was developing an APP before, when it came to the stage of online deployment, I inevitably had to deal with servers, domain names and related matters. At this point, product managers need to master both front-end and back-end development.

Not limited to front-end and back-end development, product managers should also master model fine-tuning

A colleague in my team mentioned that I am not avant-garde enough now, as I only know front-end development! The boss of my former company has already mastered both front-end and back-end development. But later I shared that during my doctoral research, I also used Astra to complete model fine-tuning and even experimental training, which was something I could never have completed on my own before.

Things that even his boss would not get involved in, and that many product managers will not participate in either.

Especially when using a local GPU, agents like Codex can let AI help you jointly debug the model, and even run experiments according to the experiments you designed.

Of course, a professor from Tsinghua University recently mentioned that students' ideas should not come from models. I think this is too restrictive. After all, although the types of knowledge combinations cannot be exhausted, innovations created by the combination of human prompts and AI also count. After all, you can let AI help you check whether relevant research results already exist.

You know, in the past, you could only complete all the work by coding yourself, and learn boring Python, but now this is completely unnecessary.

When product managers do back-end development, technology selection is no longer restricted by team members as in the past

Now, based on AI recommendations, our team can choose the most suitable back-end solution. Due to the accumulation of technology stacks, our team used to use Java all the time in the past, because all back-end developers are proficient in Java or Python.

But starting from September this year, all project editing is completed through AI-recommended back-ends, which ensures that the most appropriate back-end is used for corresponding businesses and product lines, making the project not too heavy, fully meeting business needs, and resulting in lower maintenance costs (low token consumption cost).

When product managers work on spatial computing, using AI to run spatial simulators still has limitations

Now, no matter the simulators for Android XR, visionOS or PICO OS, Apple's vision OS is undoubtedly far ahead. However, to develop a spatial application without MR headset hardware, a real device is still required.

At present, all major MR headset manufacturers have launched development tools and SDKs for their own operating systems. With the help of the most powerful AI models, product managers can also quickly complete development from ideas to prototypes. Essentially, this is the process of developing spatial computing applications. Compared with the early stage of the mobile Internet, the current development threshold and cost are lower, the speed is faster, and the support from manufacturers is also stronger — the main reason is that the current MR application ecosystem is extremely scarce, and many users have almost no playable content after buying the headsets.

The vision OS simulator can be said to be the most powerful one. Except for operations that cannot be done with cameras and gestures, and object tracking, almost all other functions can be realized.

Product managers should use AI to develop a spatial application. Of course, to run the simulator locally, the memory and video memory of the computer must be large enough. For example, my current 32G MacBook Air is so laggy when running the simulator that it is almost unusable. So in the future, the performance of Air will no longer be sufficient for product managers, the computer gets extremely hot, and the simulator is very stuttering.

When developing spatial computing applications, as a product manager, I will open Blender, Unity and the application simulator at the same time, which is completely different from the past when I only opened Axure, document tools or flowchart tools.

So it can be seen that in the future, the MacBook Air will no longer meet the needs of product managers, who will turn to MacBook Pro.

My current device is equipped with the M4 chip, which is a laptop released within the last two years and is still in its life cycle, although its performance cannot compare with the current M5 Apple chip.

If you still only do demand research instead of actually letting AI help you implement and land your ideas, compared with the past when you only stayed at the prototype design and UI design stage, now you can really test whether the product can run, as well as check the user experience and the usage rate of each function. All of these can be done by product managers right now.

You know, in the past, you had to wait for developers, or even ask developers for help. Now you don't need to ask anyone for help, you only need to manage your own work well.

Product managers have a longer-term vision than developers, and own product framework prompts

The prompt design made by product managers is more featured with product framework and global design. No matter it is UED interaction or product function layout, product managers can figure out which parts are required globally and which parts need to be modified right now, so many requirements can avoid repeated adjustments later.

This is the advantage of product managers. They can plan functions related to different life cycles in advance to achieve the final interaction effect. For all the product managers who still only use AI to write front-end code now, they are totally out.

That's all for today's sharing.

This article is from the WeChat Official Account "Kevin's Drops That Change the World" (ID: Kevingbsjddd), the author is Kevin's Stories, and it is released with authorization from 36Kr.