This whole LLM coding mania has highlighted something that I was previously less appreciative of: some people (self included) like writing code because it is an enjoyable, enriching experience to them. Solving puzzles in code has lit up the reward pathways of my brain since the first time I tried it, and I don’t see why that would ever change.

Prodding an LLM into solving a problem is just not the same thing. I understand that the results of prompting could be more exciting than programming to some, but the idea that manual coding is “over” feels dismissive of hobbyists at the very least. Perhaps the author is suggesting that manual programming in the industry is over, but I don’t buy that either. Thesis being, there will eventually be a cost to employing “experts” that lack understanding.

> If agents can handle the code, could we finally work directly with the logic, interfaces, and data models we want?

Aren’t high-level programming languages very refined encodings of “logic, interfaces, and data models?” Why would using natural language be more efficient at such things than special-purpose languages developed over decades?