Entirely vibe coding an app called "rigma" (react + figma)

born out of desire to iterate on designs mostly in code, but liking the infinite canvas of figma for making many many copies and being able to see them side by side as the changes incrementally build up

so far I've written maybe 5 lines of this codebase. I probably wouldn't have tried to build something like this without LLMs because it'd take too long, so this is a fun test of seeing how far we can get before the vibes run out

https://github.com/mcgingras/rigma

I feel like I'm riding down a humungous hill on a bicycle with each new LOC the llm writes. It's easy, carefree, I don't have to pedal. But there's a sinking feeling that at some point, eventually, I'll have to climb back up the hill and make sense of the mess

working on ambush and /noms :: nouns stan

be sure to review the code constantly. idk which model you’re using, i prefer claude 3.5 still, have been trying out Gemini. 3.7 does way too much

occasionally have it review itself and ask if it can consolidate the code or split some of it into util files to keep things clean and understandable