Andrej Karpathy raises an interesting question about the productivity of "10X engineers," those perceived to be significantly more effective than their peers. He suggests that the gap in productivity might widen as generative AI becomes more prevalent. Karpathy believes that while some engineers can be ten times more productive than the average, the impact of AI could shift the profession's dynamics. If generative AI enhances the capabilities of top engineers, the traditional notion of who qualifies as a "median engineer" might change, potentially redefining productivity metrics. Karpathy also proposes a shift in focus from comparing top engineers to the median to looking at the productivity difference between the 99th percentile and those slightly below, like the 95th or 98th percentiles. There’s a possibility that AI will help excellent engineers minimize errors and improve design, which could narrow the performance gap among high achievers. However, if the focus shifts to the 99.9th percentile, he believes the gap could widen significantly again. He emphasizes that a significant portion of engineering effort often goes into fixing avoidable problems. Many resources are wasted on maintaining poor designs or addressing issues stemming from inadequate abstractions. While AI promises quicker development, the importance of creating maintainable designs and avoiding common pitfalls tends to get overshadowed in discussions about productivity. Finally, Karpathy warns against assuming that the skills of today's top engineers will be the same as those needed in the future. While some qualities, like work ethic, will remain important, new skills will emerge as the industry evolves.