The article argues that while AI coding assistants promise to boost productivity, they may also undermine the formation of deep expertise. It cites a quote from Sam Altman of OpenAI, who described intelligence as a utility that users can purchase on a meter, suggesting a future where AI handles much of the cognitive load. The piece notes that developers with years of experience tend to benefit most from these tools, because their existing knowledge allows them to steer, audit, and verify AI‑generated output effectively. In contrast, newcomers who lack that background are placed in a difficult position: they are encouraged—or even required—to use AI assistants that presuppose expert‑level judgment to be used responsibly. This creates what the author calls an “expert novice” scenario, where beginners need the very expertise the tools are meant to supplement in order to use them well. A JetBrains study of junior and novice developers is highlighted as evidence. Participants who relied heavily on AI assistance often skipped crucial planning stages, finished with an “illusion of competence” rather than true understanding, and became lost when the generated code contained errors they could not diagnose. Conversely, those who limited AI use developed what the study terms “negative expertise”—the ability to dismiss unhelpful AI suggestions—and were able to use the models to accelerate work they already intended to produce. The article describes an “inverted learning” dynamic, where less experienced users attempt to guide the AI but are instead misled by its pattern‑based outputs, lacking the judgment to ask the right questions. It references a 2025 UPenn study that found students using large language models to learn mathematics without guardrails performed 17% worse than peers using only textbooks, reinforcing the claim that unfettered AI reliance can harm learning. The author concludes that the friction inherent in debugging, trial and error, and repeated practice is essential for building developer intuition—what Germans call Fingerspitzengefühl—and that removing this friction through AI may prevent the formation of that intuition.
