LLMs ruined coding for me, so here is how i fix it

What if i write a solver like coq but for geometry?

What a nice little project idea, i thought. What would anyone just a decade ago do? - Probably open up vim, pull up sdl2 docs, and start prototyping? They would get a working proof of concept in just a weekend, learn a lot in the process, and enjoy it. But what would someone in 2026 do? Hey [insert llm name here], write a geometry solver like coq in rust They would get code, but not code.