Senior engineers resist AI coding tools not because they fear the technology itself, but because rapid code generation threatens their hard-won expertise and understanding of their systems.

What happened: A speaker at Web Directions AI Engineer Melbourne in June 2026 presented research comparing current resistance to AI coding agents with historical resistance to pocket calculators in 1976. The argument: engineers worry that AI tools—which can generate code faster than humans can review it—will erode their knowledge of how their systems actually work, a condition the speaker calls 'cognitive debt' (borrowing from Margaret-Anne Storey's research).

Why it matters: The friction engineering leaders are seeing isn't fear of change; it's fear of demotion. Senior engineers have spent a decade or more learning to understand their codebases deeply. AI agents that generate working code without human understanding drop them back into the junior position of not knowing what's happening in the codebase or being unable to fully explain what they shipped. Meanwhile, code review—the human bottleneck that remains—risks becoming a rubber-stamp exercise, allowing poor practices ('normalisation of deviance') to accumulate.

What to watch: The signals teams use to assess code quality—commit counts, documentation, test coverage—are now noisier; an agent can produce all three in minutes. This affects hiring, promotion, and trust between teams. Research on AI's actual effect on engineering productivity is still early, mostly on small samples, and some headline numbers have already been revised by their own authors, so definitive claims about impact are premature.