How Programmers Are Adapting to AI: Reviewing Machine-Written Code Instead of Writing It

On 12 July 2026, the Guardian published a report on how American software engineers are reshaping their careers around artificial intelligence: for many of them, the job has shifted from writing code to reviewing what AI generates.

What has changed for developers

Software engineering was one of the highest-paid professions in the US in 2022, but the arrival of AI has shaken the industry, followed by layoffs and underemployment. According to the Guardian, engineers' tasks increasingly consist of reviewing machine-generated code instead of designing architecture and solving problems by hand.

The contrast is sharp. As recently as 2022, software development was considered one of the most secure career paths in the American tech sector, and just a few years later some practitioners find themselves underemployed. Generative AI has not so much abolished the profession as redefined its day-to-day content, from creative engineering work to supervising machine output.

The report's central figure is an engineer the publication calls Matt, a pseudonym used so as not to put his employment at risk. Over the past six months, roughly since January 2026, his role has shifted from writing code, solving problems and designing architecture to checking what AI produces.

How engineers keep their skills sharp

To avoid losing his skills, Matt turns his four-hour daily train commute to the town of Pawling, outside New York City, into working time for a personal project: a browser-based video game in which he writes every line of code himself, with no AI prompts.

"I'm actively trying to keep the axe sharp," —

Matt, software engineer (name changed).

He deliberately avoids AI assistants, fearing that automation will dull his engineering skills. The sharpened-axe metaphor is not accidental: for Matt it means regular practice of fundamental programming, which keeps an engineer's thinking in shape regardless of which tools are available at work.

"I try not to rely on AI where I can do without it," —

Matt, software engineer.

Key facts from the report:

• 2022 — software engineering is among the highest-paid professions in the US
• The arrival of AI triggered a wave of layoffs and underemployment in the industry
• Over the past six months, the subject's work has shifted from writing code to reviewing AI-generated code
• A four-hour train commute to Pawling (New York) goes to a personal project, a video game
• Three adaptation models: new skills, a return to fundamentals, collective action

Why developers fear deskilling

Engineers' main fear is losing their skills to automation. When writing code on your own gives way to checking AI output, a specialist risks turning from the author of a solution into a supervisor who understands less and less about what exactly they are checking. That is why Matt deliberately programs by hand in his free time: it is deliberate training, not just a hobby.

The Guardian identifies three lines of behavior with which engineers respond to the pressure from AI. Some learn new skills and move into adjacent roles. Others return to the basics: foundational knowledge that does not go obsolete along with tools. Still others bet on collective action and union organizing, aiming to influence working conditions together rather than individually.

What this means

Matt's story is a particular case of a broader shift: generative AI is changing not only the number of development jobs but the content of the profession itself. Writing code is giving way to reviewing code, and some engineers resist this by hand, preserving their skills for the moment when live engineering expertise is in demand again. For the labor market, this is a signal: value is shifting from the ability to write code to the ability to evaluate it, and keeping a "manual" skill is becoming a personal insurance policy for the engineer.