I would like to see a search engine like Google, except without paywalled content, without PDFs, without whitepages.com type results in name searches, and without store product pages for items that are out of stock. A lot of that is repackaged public content, so the fact that they're duplicative should be a very strong negative signal. Relegate those pages to a DMOZ/Yahoo model. I think that would take care of 90% of…
What's wrong with PDFs? Often when I am looking for something related to Maths or Science, I search, and I get nice short 8 ish pages PDFs from good unis, authored by a professor for some course. Those PDFs are really nice.
Cool, then it can point you to scholar.google