Comment by devxpy

The main issue is that its very slow and expensive to browse the internet like this. The LLM will only perform well if you have it do chain of thought reasoning, and that has a latency hit because of a longer generation.