💭 How to make a project: Notice a problem, look at if the problem resembles things that can be handled by code or llms or ml. Think about how a naive solution would work and notice any cringey inefficient parts, and what you'd do instead as a human. Decompose into parts, and scan for parts you've never seen solved before to make sure they're actually solvable. Make each part a file or function and give it a nice interface, comment what it needs to become. Try to implement them largely independently and use caching and fake inputs where needed to make dev faster.