Reading note. This reflection starts from a deliberately provocative paper, "The End of Software Engineering: How AI Agents Are Fundamentally Restructuring the Software Paradigm," by Zhenfeng Cao, posted to arXiv on June 5, 2026. We engage with its thesis — code becoming a disposable tool in service of model-driven reasoning — without treating its conclusions as certainties. It's a reading, not a prophecy.

In one sentence

For the first time, the machine no longer merely runs code: it has begun to write it. This doesn't mark the end of developers, but the end of their monopoly on creating value. The winners won't be those who write the most code, but those who best understand the problem to be solved.

A laptop displaying source code on a clean desk

A title that announces the end of the world

"The End of Software Engineering."

That's the title Zhenfeng Cao chose for a paper published in June 2026.

As often happens in tech, headlines announce the end of the world. Then, the next day, the world keeps turning.

And yet, this time, something is different. Because for the first time in the history of computing, the machine no longer merely executes code. It has begun to write it.

And that changes everything.

The developer has become the new taxi driver

Remember.

When Uber arrived, many believed technology could never replace the experience of a driver who knew their city. Then one day, GPS got better. The job didn't disappear, but its value shifted.

Part of software development is living through exactly the same thing today. For years, knowing how to write code was a rare skill. Now an AI agent can produce in minutes what once took several days of work.

So the question becomes uncomfortable:

"If your only value is typing code, how long will you stay indispensable?"

The problem was never the code

The great illusion of the digital industry was believing that code was the product.

Code was never the product. Code is only a tool. Nobody buys an application because it contains 500,000 lines of code.

People pay to solve a problem.

  • A shopkeeper wants more customers.
  • A town hall wants to improve its services.
  • A nonprofit wants to reach more people.
  • A company wants to save time.

The rest is just plumbing. And plumbing is precisely what artificial intelligence is starting to automate.

An industry obsessed with complexity

Let's be honest. Part of the tech sector loves complexity.

Framework on top of framework. Microservice on top of microservice. Dependency on top of dependency.

Sometimes it feels as if certain projects are designed to impress other developers rather than to help users.

AI is breaking that model, because it raises a brutal question:

"Why mobilize ten people for six months to produce what a small team can now deliver in a few weeks?"

That question is going to get harder and harder to avoid.

The winners won't be those who write the most code

For a long time, competitive advantage meant producing more. More developers. More features. More budget. More servers.

We believe we are entering another phase. The winners will be those who best understand their users. Those who build the simplest solutions. Those who know how to cut out the useless. Those who know how to create value before creating complexity.

Because artificial intelligence is turning code into a commodity. And when a resource becomes abundant, it stops being a competitive advantage.

The decade will be hard for shovel sellers

For two years now, everyone has been selling AI. Everyone is launching their platform. Everyone is chasing investors. Everyone is promising the revolution.

Yet economic history always teaches the same lesson: technology doesn't create value, it amplifies it.

An automated bad idea is still a bad idea. A useless product built ten times faster is still a useless product.

So the real question isn't "How much AI are you using?"

The real question is: "What problem are you actually solving?"

The return of the big picture

For nearly twenty years, the digital industry placed the developer at the center of everything. Code was scarce. Code was hard. Code was the strategic resource. So the developer became king.

Artificial intelligence is reshuffling the deck. This isn't the end of developers; it's the end of their monopoly on creating value.

We are watching the return of profiles able to see the whole board:

  • the return of project managers;
  • the return of product managers;
  • the return of designers;
  • the return of domain experts;
  • the return of those able to understand a need, build a vision, and orchestrate human teams — and now teams of agents.

Because writing code is gradually becoming easier. Understanding humans remains hard.

For years, some believed that mastering a technology was enough to secure their future. But technology is in constant flux. The COBOL developer had to adapt to client-server. The client-server developer had to adapt to the Web. The Web developer had to adapt to the Cloud. Today, everyone has to adapt to AI.

History waits for no one. You can mourn the change. You can criticize it. You can explain why it was better before. But in the end, you have to make a choice.

Get on the train. Or stay on the platform.

For our part, the choice is already made.

Sources and further reading

This document is updated if new elements emerge. Last revision: 文 June 9, 2026.