Software development is considered one of the areas experiencing the strongest AI-driven change. AI is being adopted across many fields, but it is still hard to find another area where change is as fast and wide-ranging as in software development.

Against that backdrop, freelance journalist Clive Thompson interviewed more than 70 software developers working at companies large and small, drawing attention for highlighting the changes AI is bringing to software development.

At a broad level, software developers appear to view the trend of AI automating coding very positively, rather than feeling uncomfortable about it.

In other fields, many reactions to AI are neutral or skeptical. There are even assessments that programmers are among the white-collar occupations more likely to be replaced by AI. Even so, developers are relatively friendly toward AI.

Looking at the developers Thompson interviewed, they seem to enjoy the changes brought by AI rather than resisting or fearing them. Some people worry about or feel uncomfortable with change, but the dominant mood is welcoming.

That is not unrelated to the nature of the work AI replaces. Thompson said that in other creative fields, AI replaces the soulful work while leaving humans only simple repetitive tasks. In software development, he said, AI mainly replaces repetitive and time-consuming coding, leaving developers with little reason to feel uncomfortable. He said AI has made it possible to focus more on work humans can do.

Kent Beck (켄트 백), who has been coding since 1972 and is regarded as a master in the developer world, nearly broke with software development after being disappointed by programming languages and software tools about 10 years ago. After seeing LLMs, he returned to the field. The New York Times reported that he is now producing more projects than ever, including a personalized memo app and a new type of database.

AI-driven coding automation is not entirely new when seen in the broader flow of software development history. Over decades, the software development field has repeatedly gone through a process in which programmers automate cumbersome tasks in the name of abstraction.

As computer performance improved in the 1980s and 1990s, developers created programming languages such as Python that automated complex and annoying memory management tasks. The push for automation accelerated in the 2000s and 2010s as well. The New York Times reported that virtually every time developers faced hard work, they wrote code to automate it and released it as open source so others could use it.

AI coding is also an extension of that trend. The New York Times reported: "Writing software no longer means you have to hold subtle differences between languages like Python, JavaScript and Rust in your head. It no longer includes the process of messing up an algorithm and finding where the bug is. Those parts, too, have been abstracted away and disappeared."

As coding is abstracted away and disappears because of AI, programmers' main work has also changed significantly. Boris Cherny (보리스 체르니), who leads Anthropic's coding AI service Claude Code, said coders are now closer to architects than construction workers.

Developers who use AI focus on overall software structure and how the functions and components that make up software interact. Because agents generate working code very quickly, humans supervising it can experiment by checking what works and discarding what does not. Thompson said, "What developers do now is closer to judging than creating."

AI is bringing enormous change to software development productivity. But the productivity gap from adopting AI coding is significant between startups and established companies. That reflects the difference between starting from a zero base and having to account for an existing code base.

Entrepreneur Dima Yanovsky (디마 야노프스키) said that using Claude Code has made it possible to finish in a few hours what used to take weeks. But the story is somewhat different at established companies. Google CEO Sundar Pichai said that at mature tech companies like Google, the boost AI provides to programmers' productivity is about 10 percent in numerical terms. That is the result of having many other things to consider compared with startups. The New York Times reported that startup founders interviewed by Thompson write almost 100 percent of their code with AI, but that the share at Google is less than 50 percent.

Server management work, which involves many annoying tasks, is also a good match for AI. Because AI is fluent in both human language and programming languages, it has become reality for it to interpret system error reports, analyze code and, in some cases, produce solutions before the person in charge even wakes up. On this, Thompson interviewed David Yanacek (데이비드 야나첵) who works at Amazon Web Services. He said there was a case in which an AI agent solved in 15 minutes a debugging job that previously took an engineering team 8 hours to fix.