Seasoned software engineer Namanyay Goel has raised alarms about the latest generation of junior developers, claiming their heavy reliance on artificial intelligence tools is leaving them without a fundamental grasp of how their own code operates. In a recent blog post titled “New Junior Developers Can’t Actually Code,” Goel described how many entry-level programmers now run AI assistants like GitHub Copilot, Claude, or ChatGPT around the clock, producing working code at unprecedented speed but struggling to explain the reasoning behind their solutions.
“Every junior dev I talk to has Copilot or Claude or GPT running 24/7. They’re shipping code faster than ever,” Goel wrote. Yet when pressed on why a particular approach was chosen over an alternative, he said the response is often silence. “Ask about edge cases? Blank stares,” he added, noting that the foundational knowledge traditionally gained through trial and error appears to be missing.
Goel’s critique is not a blanket rejection of AI in programming. Instead, he likens the situation to the introduction of calculators in mathematics classrooms—a tool that eventually became standard, but one that initially raised concerns about over-reliance. His main worry is that AI serves as an easy crutch, preventing newcomers from engaging in the struggle that is essential to mastering a craft.
From StackOverflow to Chatbots
In the past, junior developers seeking guidance would turn to forums like StackOverflow, where they had to read through multiple expert discussions to piece together a full understanding. “It was slower, but you came out understanding not just what worked, but why it worked,” Goel noted. Now, he says, many simply copy and paste error messages into chat interfaces, receiving answers without the context needed to synthesize alternatives or think through problems independently.
Goel’s observations align with a recent study by researchers at Microsoft and Carnegie Mellon, which found that increased reliance on AI tools correlates with diminished critical thinking skills. The study, which relied on self-reported data, suggested that as trust in AI answers grows, users’ cognitive engagement decreases—similar to how an unused muscle weakens over time. While the study has limitations, the concept of “cognitive offloading” is widely recognized in psychology.
Beyond the concern for skill development, there is also the issue of AI accuracy. Large language models are known to hallucinate, and some evidence indicates that while they can speed up coding, they may also introduce more errors into the final product. This tradeoff, Goel argues, is a hidden cost that could outweigh the short-term gains in productivity.
Goel does not advocate for abandoning AI altogether. “The future isn’t about where we use AI—it’s about how we use it,” he wrote. But he warns that the current trajectory, which prioritizes quick fixes over deep understanding, could have long-term consequences. “We’re trading deep understanding for quick fixes,” he said. “We’re going to pay for this later.”
As AI becomes increasingly embedded in the software development workflow, the challenge for the industry will be to find a balance between leveraging these powerful tools and preserving the educational foundations that enable programmers to truly understand their craft. Whether that balance can be achieved remains an open question, but Goel’s critique serves as a reminder that the tools we use shape the skills we develop.
