Don't worry. Why do we say that AI can never eliminate real programmers?

God Translation Bureau is a translation team under 36Kr, focusing on fields such as technology, business, workplace, and life, and mainly introducing new technologies, new ideas, and new trends from abroad.

Editor's note: When AI programming tools are hyped again for "replacing programmers", history has repeatedly proven that technology never replaces skills but elevates them to a higher level. If carpenters are equipped with CNC machines, guess who can make better furniture? This article is from a compilation.

From "No-Code Tools" to "AI-Assisted Programming"

Every few years, some shiny new technologies emerge, claiming to make programmers unemployed collectively. Media headlines always follow the same old routines: "Programming is dead", "Everyone is a developer", and my favorite, "Your kids can write code before they can read."

Bosses get excited, consultants swarm in like sharks, PPTs start to grow wildly, and budgets shift significantly.

Then reality always hits hard.

The truth is never about replacement but transformation. Tools that claim to eliminate technical barriers ultimately give rise to brand - new high - paying positions. The no - code movement didn't eliminate programmers; instead, it created no - code experts and backend integration engineers. Cloud computing didn't eliminate operations, but it transformed system administrators into DevOps engineers with doubled salaries.

Now, AI programming assistants are repeating this cycle. The promise of "AI writing code for you" is evolving into "requiring engineers to command AI systems" - essentially the same people, but with upgraded skills and higher asking prices.

However, this transformation hides a deeper change. Different from previous technological revolutions, AI programming exposes a truth that has always existed in the software industry but has been ignored by everyone:

The core value of programming has never been writing code but designing system architectures.

And this ability is precisely what AI is least likely to replace.

The Carousel of Replacement Promises

How many rounds of similar hype have we experienced? Let's count them:

The No - Code/Low - Code Revolution

When drag - and - drop interfaces claimed to enable business people to build their own applications, their slogans were so appealing: "Everyone can develop. Why hire expensive programmers?"

The reality is that these tools have created brand - new problems. Specialists are still needed to design underlying data models, connect with existing systems, handle exceptional cases that visualization tools can't solve, and upgrade with demand iterations.

As a result, the number of developers has increased instead of decreasing, and "no - code experts" who understand both business and technology have emerged. Guess what? These people's salaries are even higher than those of the programmers they were expected to replace.

The Cloud Computing Revolution

"You don't need operations when you move to the cloud!"

It's as if infrastructure can be automatically managed when placed on someone else's server. Cloud technology has never eliminated system expertise but has reshaped skills and greatly expanded their boundaries.

Operations engineers have collectively transformed into DevOps engineers, with fancy titles and doubled salaries. The work hasn't disappeared but has evolved into infrastructure - as - code, automated deployment, and distributed system management.

As I said in a LinkedIn discussion about microservices: "I've seen too many teams spend months splitting perfectly good systems, only to find that they've replaced old problems with more expensive ones." The complexity brought about by cloud services ultimately requires system experts to handle it at a higher level.

The Offshore Development Wave

"Overseas development is cheap. Why hire local programmers?"

The cost dream quickly hit the hard wall of communication barriers and quality defects. People finally realized that efficient software development requires in - depth business knowledge and continuous collaboration.

It has finally evolved into a more refined model: Distributed teams need to clarify the boundaries of rights and responsibilities, strengthen architectural specifications, and the total cost - unsurprisingly - far exceeds the initial estimate.

The AI Programming Assistant Revolution

Now it's AI's turn to promise to write code for us. "Just talk, and the code will be automatically generated!"

Early signs in reality have emerged: The code generated by AI seems reasonable but hides errors, and senior engineers spend a lot of time correcting them. In the "ambient programming" phenomenon, experienced programmers are better at extracting value from AI than beginners. Systems built purely by AI often have chaotic architectures.

If carpenters are equipped with CNC machines, guess who can make better furniture?

The pattern is clear again: Technology never replaces skills but elevates them to a higher level.

Why This Time Is Different

The fundamental misunderstanding of those who claim that "AI will replace programmers" is that code is not an asset but a liability. Every line of code needs to be maintained, debugged, secured, and will eventually be replaced. The real asset is the business capability implemented by the code.

If AI makes writing code cheaper and faster, it actually accelerates the creation of liabilities. When liabilities can be accumulated at an unprecedented speed, the ability to strategically manage and minimize liabilities becomes extremely valuable.

This is especially true in the AI era - AI is good at local optimization but poor at global design. It can improve a single function but cannot judge whether a service should exist or how it should interact with the macro - system. When the implementation speed soars, architectural errors are made long before you notice.

This may be okay for one - time marketing websites, but it will be a disaster for core systems that need to evolve continuously.

The pattern of technological transformation has always been the same - operations become DevOps, backend development becomes cloud architects - but AI accelerates everything. The ability that will ultimately stand the test has never been writing code.

It's the ability to design system architectures. And this is exactly what AI can never do.

Translator: boxi.