How this works is that I extracted knowledge from PGs publicly available essays. Instead of using RAG I used a state map approach, if you are keen about the extraction method I can share more in the comments.TLDR: it’s a PG avatar trained on his essays and the avatar renders on webgpu, give it a shot and ask it about the hardest problem you are facing in your startup right now. If you like the avatars answer you can share your conversation here!