What process did you use? I still find it hard to believe an LLM could write 100k lines of code without completely failing - it fails on much more basic stuff in my experience.
I had a bunch of copy pasted prompts. Start off in Claude to reason out the feature and basic code, then paste that to Cursor agent, then let it build. Then copy paste 3 step workflow to check the code. Then ask it to fix errors and bugs when I UI test or curl test the code. Lots of logging info lines
