1. I generally prefer that an LLM not search the web. The top N results are often either SEO spam, excessively long articles created solely to rank well, or long-established websites that gained authority years ago, when Google's crawler and ranking algorithms were less sophisticated.
2. Web search by LLMs is likely here to stay, so I'm curious whether there's an agent-friendly web format. For example, when an RSS reader visits a website, the site responds with an RSS feed. I think we need something similar for agents - an open standard that all websites would support. This could reduce processing overhead and potentially improve the accuracy of the information retrieved. Thoughts?