Today I vibe coded a webapp that is simultaneously capable of being hosted for free on github pages, uses a 22mb model (small enough to run in browser) to do semantic search across any 2 pdf's uploaded, and only took 1 working day to build.

Are LLM's a bubble?