Elon Musk said that the AI singularity is coming.

Previously, people without technical knowledge faced a limitation when developing products: they lacked a programmer.

To be precise, they lacked a programmer who was patient, good - tempered, easy to talk to, worked quickly, and didn't mind changes in requirements.

Of course, such an ideal programmer is hard to come by in reality. However, things seem to be different now.

It's not that programmers have changed, but that AI programming has become much more powerful.

Let me put it this way. Currently, AI Coding has outperformed human programmers to some extent in most newly - built projects, and the pace of its evolution is accelerating.

David, the founder of Midjourney, publicly stated some time ago that he had completed more coding projects during Christmas than in the past 10 years combined.

He also said that although he could feel the limitations, he knew that everything had changed.

Elon Musk posted two consecutive messages on X, directly stating that 2026 would be the year of the singularity.

What they said actually points to the same thing: Claude Code.

I've previously written articles sharing the usage and capabilities of Claude Code and also presented some cases. Those interested can search for them in my past articles.

Recently, Claude Code has become very popular overseas, and many of my friends who are fond of AI Coding use it to complete coding and project creation.

In the technical field, the term "singularity" refers to the stage when technological growth enters an uncontrollable and irreversible exponential explosion phase.

In the past, we used "Moore's Law" to define the development pattern of computers. The number of transistors that can be accommodated on an integrated circuit approximately doubles every 18 - 24 months, and at the same time, the cost is halved.

However, this is still within the scope of control and predictability.

Now, the evolution speed of AI has clearly exceeded the scope of Moore's Law, and its capabilities are difficult to be constrained by a single qualitative framework.

Especially in the programming field, almost every release of a new model and iteration of an existing model claims an improvement in code capabilities.

Claude Code has always been popular, not just recently.

The Claude Opus 4.5 released two months ago really boosted Claude Code. On the LiveBench, a benchmark test list for large - language models (LLMs), Claude Opus 4.5 surpassed GPT - 5.1 Codex MAX and Gemini 3 Pro to take the top spot.

In a sense, this is currently the most powerful AI Coding model.

When I was a programmer, it took me about half an hour to write code to develop a simple login function.

If the logic was more complex, it could take two or three hours.

In the process, I needed to complete front - end page development, back - end logic development, database operation development, and then conduct debugging and debugging before finally publishing the product.

If we add the pre - work such as requirement analysis, prototype design, PRD writing, UI design, annotation, and image slicing, it would take at least one day to develop a simple login function.

Note that I'm talking about a simple function.

Once a bit of complex business logic is added, the overall workload increases significantly.

However, now, if we give the above tasks to AI with clear instructions, it may be completed in just a few minutes.

The key is that AI can achieve better quality, higher efficiency, and lower cost than humans.

During those days of staying up late to write code, I never thought that one day there would be something that could replace programmers in writing code.

So, can AI completely replace programmers now?

I don't think so.

It's not because AI's capabilities are not strong enough, but because it's difficult for AI to seamlessly integrate into existing operating systems.

AI can start from scratch in developing new products and systems and is capable of developing some small websites and App applications. It can indeed do better than humans.

However, few companies and teams would dare to let AI take over the code development of an existing complex system.

Most of the products of AI Coding you see now are either simple tools or tool - flows with only a small number of interface calls. There are almost no complete commercial - grade complex systems.

It's not that AI's code - writing ability is insufficient, but that it can't directly replace human programmers in terms of business logic, data processing, security strategies, and code quality.

Programmers won't be eliminated, but programmers who master AI capabilities will become more and more popular. And ordinary people finally have a chance to develop their own products using technology.

Just a few years ago, it was popular in the industry to learn Python and PHP to develop products, but now hardly anyone talks about it.

It's not that technology is no longer important, but that technology has become more accessible.

Readers with technical knowledge know that programming languages are actually computerized natural languages.

When we develop products, we first define requirements, conduct design, sort out logic, and then start development. The development process is to translate the existing logic into machine - understandable language using code.

Theoretically, natural language is actually a programming language.

We use plain language to describe what the interfaces of an App or website look like, explain in detail what the logic of a login function is, and describe what the processing flow is after a product order is placed.

All these are natural languages.

Previously, computers couldn't directly convert natural language into code, but now AI can.

Therefore, in the future, natural language will be the programming language.

That's why I say it's not that technology is unimportant, but that it has become more accessible.

In my opinion, in the future, there will be two types of people using technology to develop products. One type is professional programmers, and the other is ordinary people who can use AI for development.

More precisely, I would regard the latter type of people as product managers.

The former has a full set of knowledge and skills in the technical field and is mainly responsible for large - scale complex systems. The latter has strong structured thinking and expression abilities and can accurately describe their ideas to AI using natural language. They are mainly engaged in developing application - type products.

Someone once mocked the saying that everyone can be a product manager. Perhaps, the real era of "everyone is a product manager" is coming.

I'm very optimistic about this.

Some people may still think that current AI programming is not mature enough for commercial use. I understand this view, and it's quite normal.

However, since AI may be approaching the singularity, our past cognition and the rules we've mastered may also be broken.

Ray Kurzweil proposed the "Law of Accelerating Returns" in his book The Singularity Is Near. He believes that technological progress grows exponentially and predicted that the singularity would arrive around 2045, when humans could achieve integration with technology.

Now, it seems that this prediction may be advanced.

I wrote my first line of code in 2007, using the C language.

On the first day of learning technology and programming, every programmer starts by outputting "Hello World" in the command line.

Back then, the "World" for us was a computer world, but today's "World" is an AI world.

If writing code is approaching the singularity, what about jobs in other fields?

I have a hunch that in the future, many industries will have positions similar to "super product managers".

They understand business, users, and commerce. The key is that they are good at combining and using various AI capabilities to create value. This "one - super and multiple - strong" model will become the composite talents of the new era.

We are still in the best of times. Don't panic, don't be impatient, and don't wait.

Iterating ourselves is always the best solution.

This article is from the WeChat official account "Tang Ren" (ID: RyanTang007). The author is Tang Ren. It is published by 36Kr with authorization.