Engineer explores reluctance to adopt AI tools through self-interview, weighing productivity gains against loss of learning and accomplishment

Tom, a mechanical engineer specializing in integrated design, has barely used LLMs until recently despite years of reading about them, citing concerns about missing out on learning and the sense of accomplishment from completing projects himself.

When he tested GPT-4 on a fluid-flow program in December 2023 and Copilot on a blog post around January 2026, both produced usable results—but he found himself unwilling to use them, partly due to frustration that a Microsoft subscription cost had jumped 35% to cover AI features he hadn't opted into.

He distinguishes between AI as pure acceleration (which he resists because it strips away learning and self-confidence) and AI as assisted learning or capability-expansion (which he found valuable when translating a Python program to JavaScript using Claude Haiku 4.5, where he could tag specific functions and felt he was managing a predictable process).