Great read. One thing that would have been useful would be to know what sort of size dataset they were working with? If you're not working with millions of documents, is the search and link walking still viable?