AI Won't Replace Programmers. Not Yet.

In recent years, we've been constantly hearing: "AI will replace programmers," "Programmers will be out of work," "Neural networks write better code than humans." Let's examine if this is actually true and what the future holds for the industry.

Computers Don't Understand Your Code. They Just Execute It.

The only programming language natural to computers is machine code. Everything else was created not for machines, but for people. Programming languages exist primarily so that one person can explain program logic to another. Computers don't care about code readability. You could write an absolute mess of symbols, and if the syntax is correct - the program will work.

This is precisely why we have "best practices," architectural patterns, and frameworks - not for machines, but for programmers. Good code is code that's easy to read and maintain. Although sometimes developers' imagination leads to strange places, and we get gems (or should I say, "Perls") like this completely valid code:

$@ = (*ARGV)[$#{[()]}]

_{(Perl fans, try to convince us that this is more understandable than machine code!)}

LLMs Augment Programmers, Not Replace Them

Large language models (LLMs) were trained on existing code, so they learned to write programs in human-oriented languages. This makes development faster and more convenient, but doesn't eliminate programmers themselves. Similarly, the emergence of Fortran and Cobol increased the productivity of developers who previously wrote in assembly language.

"Vibe programming" is now actively discussed - the ability to describe tasks in free form, letting AI write the code. This approach resembles another wave of no-code development, but its main drawback is the difficulty of subsequent modification and maintenance. In serious projects that require long-term development and multiple iterations, this programming method is simply unsuitable.

Computer Code for Computers

If AI can already write programs in Python, Java, or C++, then the next step is to remove humans from this process. Why should a computer write code in a language convenient for humans? Our limitations lead to redundancy and complexity in programming.

The quality of code generation will eventually hit a ceiling, and the next logical step will be creating a programming language optimized for LLMs - an AI-first programming language.

This language will likely be:

• High-level and concise
• Without "syntactic sugar" and unnecessary abstractions
• Optimized for tokenization and transformer models

And that's when programmers will truly be at risk.

Future: Fewer Abstractions, Faster Programs

We currently live in an era of software "bloat." Computers are becoming more powerful, but software is getting slower. Complexity is growing, and to handle it, we create new abstractions, frameworks, and libraries. But AI has different limitations.

LLMs using an AI-first language could write code without redundant layers. We might see programs that don't require operating systems.

Conclusion

As a programmer, I anticipate AI-first programming languages with some apprehension, but as a user, I hope this will bring the qualitative leap in software that we've all been waiting for!

For now - no, AI won't replace programmers. At least, not in the next few years. Instead, it will become a powerful tool that enhances our capabilities. But the future... the future might be entirely different.

‍