AI is often framed as a replacement for software engineers. In practice, it behaves more like a multiplier.

Over the past few years of day-to-day development with AI-assisted tools, the pattern I’ve observed in my workplace is consistent: engineers with strong fundamentals become significantly more effective, while others see limited improvement.

AI amplifies capability. It does not create it.

I still believe writing 20% of code is beneficial to my knowledge. Completely handing off will cause sufficient memory lapse for judging and guiding AI. It doesn’t boosts the creation process, it boosts the painful heavy repetitive workload polishing for the last mile in producing quality production grade work. (test coverage, profiling tools, refactoring) Now we can do much more and even something not we can in the past.