Does AI coding help junior developers actually get better at the craft, or just ship faster in the moment?

This explores whether AI assistance helps less experienced developers become more capable, rather than just faster. The corpus separates the two, and that separation matters. On output, the evidence favors juniors: a study of more than 5,000 customer support agents found AI assistance raised productivity about 15% on average, and the least experienced workers gained the most in both speed and quality Does AI assistance help less experienced workers most?. Experienced people benefit less and sometimes lose ground. In that same study the most experienced agents saw slight quality declines. In a randomized trial with seasoned open-source developers working in codebases they knew well, early-2025 AI tools made them 19% slower, even though they had expected to be 24% faster Do AI coding tools actually speed up experienced developers?. So on productivity, AI does narrow the gap between novices and experts.

On skill, the picture flips. A randomized trial of developers learning a new library found that AI use hurt their conceptual understanding and debugging ability Does AI assistance actually harm the way developers learn?. The key detail is that how people used the AI mattered more than whether they used it. Developers who handed work off or pasted answers scored 24–39% on later quizzes. Developers who asked for explanations and checked their own understanding scored 65–86%. The tool doesn't decide whether a junior learns. The habit of use does. A junior who closes the output gap by delegating may be widening the skill gap underneath it.

There is a second, quieter problem: juniors may not get the work they need to learn from. Interviews with South Korean software engineers describe generative AI pulling routine entry-level tasks into workflows run by seniors with AI. That removes the hands-on struggle through which juniors used to build expertise, and seniors and juniors don't see the problem the same way Does generative AI prevent juniors from getting entry-level work?. Seniors feel it too. Anthropic's engineers report large productivity gains but worry that letting Claude handle routine coding erodes the practice they need to catch its mistakes Does AI assistance erode the skills needed to oversee it?. If people who already have the skills fear losing them, juniors who never built them are in a weaker spot.

You might hope that working well with AI is itself the new skill juniors could build. The corpus is skeptical. Patterns in developers' conversations with coding agents did predict outcomes beyond prior ability, but those patterns weren't stable or transferable enough to count as learnable skills Can conversation patterns predict coding outcomes better than prior skill?. Hiring may hide all of this. Recruiters gave candidates who listed AI skills 8–15 points more interview invitations, and barely checked whether those skills were real Do AI skills help candidates get more job interviews?.

The short answer is that AI closes the gap in output, but it closes the gap in skill only when the junior uses it as a tutor and not as a substitute. The corpus has no long-term studies following junior developers over years, so whether early gains in output turn into lasting expertise is still open.