AI will slow you down

.., until it doesn’t.

This study echos my own experiences. AI in new codebases, languages or contexts helps accelerate the productivity curve. In familiar, especially in familiar and large codebases, it is a significant tax. I won’t say “useless”, but nigh useless if you can already visualize the codebase and know what you are doing.

I think a good question to start asking then, is how much business value is the expertise you wish to have, as a humon. Then, if you want to be actually good (vs “senior”), scoping your career to places that actually need that.

Otherwise I would probably land on the business value of specialization being marginal outside of the *one* thing that business has value around. For non software firms, that likely means zero need for good devs. Their business, be it real estate, retail, travel, renting cars or houses, etc. doesn’t need the software expertise. It needs experts in its core business.