so the main feature of claude project is its ability to have internal knowledge and a set of custom instructions (which mostly affects how it answers questions)
Isn’t this basically just uploading the files to a vector db and attaching it to the llm ? Plus a simple system instruction which asks it to follow the provided custom instructions when answering, and use the vector database as long term memory ?
Is it really this simple or am I missing anything ? Like what’s the hype really about ?
Comments URL: https://news.ycombinator.com/item?id=45475857
Points: 1
# Comments: 0
Source: news.ycombinator.com
