People SHOULD feel like making a useless Mario Kart clone isn't worth the effort anymore. They should, instead, be trying to figure out how to actually use these models to make something that doesn't feel like a useless Mario Kart clone.
Software will be more like a "proof of work", where people would still pay $100 for good software that took $10k tokens to build.
Fast at first especially to a PoC but then you hit a time when the arch just gets so grimy it falls over, you ship the wrong thing and you end up in rework hell. This isn’t guaranteed but when it’s teams of people slinging slop it gets worse and worse, the best devs burn out from trying to manage the ones who believe AI will save them.
Its like the problems of outsourcing all the time
My philosophical take is that there are different types of coders, the ones who did it because they wanted to see their design be implemented and those who love the technical puzzle (with obvious overlap). The technical puzzle is taking a hit currently, but I think it will still exist in some form; the designer trying to get something implemented still loves the design work primarily which is sorely lacking in current LLMs, but if they don't have enough technical knowledge they will still get slapped in the face by technical gaps and bugs that LLMs make.
For corporate stuff, I think bureaucracy and issues with churn reduce the effectiveness of using an LLM because the most effective way might be anathema to corporate fiefdoms and structure. The churn is a problem for everyone though...
People will get "starter" projects that they can tweak themselves, and that come with AI instructions for development.
Imagine wanting to create a high quality RPG game. You could spend days prompting and hundreds of dollars in tokens, or you can buy a "RPG starter kit" that contains all the systems already impmemented and tested and markdown files on how the AI can add features, test, deploy, etc.