Developers confront an identity crisis as AI takes over the craft of coding

Software engineers who once defined themselves by their ability to write elegant code are being forced to ask a more fundamental question.

If anyone can generate working software, what remains of the profession?

On Thursday a widely shared post by Santiago Valdarrama, the computer scientist and AI educator known as @svpino, captured a quiet anxiety spreading through the industry. “A buttload of developers are going through a tough time right now,” he wrote. “Their identity has always been tied to their ability to write code, so now they feel threatened by this new era where anyone can write code. If they don’t write code anymore, who are they?”

Valdarrama, who describes himself as pragmatic, admitted he has struggled with the shift himself. Manual coding, he argued, is “destined to disappear.” The reaction was swift and revealing. Many replies agreed that the real skill had never been typing syntax but solving problems. One of the most liked responses came from founder Rick Blalock: “Stop tying your identity to ‘I know how to write a loop faster than anyone.’

These were always poor things to attach your identity to anyway. If your identity is attached to solving problems… then you attach yourself to value that aligns with the world and its needs.

Another high-engagement reply put it more succinctly. “My main skill was never writing code, it was always solving problems with software… now there’s just more ways to do it.”The conversation reflects a broader transformation already well under way.

Surveys in 2026 show the majority of professional developers using AI coding tools daily, with the proportion of AI-generated or AI-assisted code rising sharply in some reports to more than half of output. At the same time, trust remains low. Large majorities of engineers say they do not fully trust the code these systems produce and spend increasing amounts of time reviewing, verifying and correcting it.

The result is a subtle but profound change in daily work. Senior developers frequently describe AI as a force multiplier that frees them to focus on architecture, edge cases, system design and business problems. A junior engineer or product manager armed with the same tools often still produces mediocre results. Experience, judgement and the ability to specify requirements clearly have become more valuable, not less.

Yet the psychological and skill-related costs are real. Evidence of the latter comes from a controlled experiment published by Anthropic earlier this year. Researchers recruited 52 mostly junior software engineers who were familiar with Python and AI coding tools but unfamiliar with a particular library (Trio). Participants were randomly assigned either to use AI assistance or to code by hand using only documentation and conventional search. Both groups completed the same programming tasks; they then sat an immediate quiz on the concepts they had just used.

Those who worked with AI scored 17 per cent lower on averag, the equivalent of nearly two letter grades than those who coded without it. The gap was widest on debugging questions. The AI group finished slightly faster, though the time difference was not statistically significant. Participants who treated the model more as a collaborator, asking conceptual questions rather than simply delegating the work, retained more understanding. Pure “offloaders” performed worst.

The findings underline a tension now widely discussed in the industry: AI can accelerate output while simultaneously impeding the formation of deep mastery, particularly for less experienced developers. For a generation of engineers who took pride in the craft of writing clean, efficient code, the automation of that core activity can feel like an existential threat. Some report a sense of skill erosion and a nagging worry that they are becoming better at prompting than at understanding systems deeply.

Others note a growing divide between those who lean heavily on AI and the more experienced “craftsmen” left to clean up the resulting code.The labour market has so far proved more resilient than the gloomiest predictions. Demand for senior software engineers has held up or even risen in several markets, while entry-level openings have remained subdued. Companies still need people who can take responsibility for systems, manage risk and exercise judgement that current AI cannot reliably provide.

The more constructive voices in Thursday’s discussion suggested the profession is not ending so much as moving up the abstraction ladder. The valuable developer of the near future may resemble a “creative director of code” defining what should be built, guiding AI agents, verifying the results and ensuring the software actually solves the right problem.Whether that redefinition feels liberating or diminishing depends largely on how tightly individuals have tied their sense of self to the keyboard. As one reply noted, once value is no longer measured by lines of code produced, space opens for more creative problem-solving. The transition is uncomfortable. It may also be inevitable.