It's great to do what he's doing, but I just had a qualm with pinning this on AI- based on my experience repeatedly trying to make AI work every time a new model comes out, what I've realized is it requires a huge amount of pre-existing knowledge and context/harness engineering to get useful outputs consistently over a long time. Even then it's not like AI is replacing you more like it kinda helps a bit if you put a lot of effort into covering all its mistakes. I do agree there's a huge benefit in using AI as an alternative to search, research, prototyping though.
Even after adding all that up maybe you save some time. 10%-20%? Maybe? You do save a lot of cognitive load as well and it feels good but a lot of the times you pay the price later when you don't understand the code/project as it gets more complex and you need to debug it when the AI can't anymore. The point is there's just not enough to replace and even if the research angle saves you some time or cognitive-bandwidth why not just use that time to do something else? Like more work or more life.
Even after adding all that up maybe you save some time. 10%-20%? Maybe? You do save a lot of cognitive load as well and it feels good but a lot of the times you pay the price later when you don't understand the code/project as it gets more complex and you need to debug it when the AI can't anymore. The point is there's just not enough to replace and even if the research angle saves you some time or cognitive-bandwidth why not just use that time to do something else? Like more work or more life.