Harper Reed gives a nice overview for how he is using LLMs to write software. I like his simple breakdown of using successive steps to (1) brainstorm, (2) plan, and (3) execute. Execute is where I still am learning. Right now my process pretty much just looks like copying pasting a lot of code between my IDE and Gemini.
I like his overall take on the current optimism and skepticism surrounding LLMs in general:
A lot of my friends are like “fuck LLMs. They are terrible at everything.” I don’t mind this POV. I don’t share it, but I think it is important to be skeptical. There are an awful lot of reasons to hate AI. My main fear is about power consumption and the environmental impact. But… the code must flow. Right… sigh.
If you are open to learning more, but don’t want to dig in and become a cyborg programmer - my recommendation is not to change your opinion, but to read Ethan Mollick’s book about LLMs and how they can be used: Co-Intelligence: Living and Working with AI.
Book purchased. Look forward to reading!