This talk examines the evolution of web search experiences over 20 years, and their impact on the underlying architecture. Early web search represented the adaptation of methods from classic Information Retrieval to the Web. Around the turn of this century, the focus shifted to triaging the need behind a query - whether it was Navigational, Informational or Transactional; engines began to customize their experiences depending on the need. The next change arose from the recognition that most queries embodied noun phrases, leading to the construction of knowledge representations from which queries could extract and deliver information regarding the noun in the query. Most recently, three trends represent the next step beyond these "noun engines": (1) "Queryless engines" have begun surfacing information meeting a user's need based on the user's context, without explicit querying; (2) Search engines have actively begun assisting the user's task at hand - the verb underlying the noun query; (3) increasing use of speech recognition is changing the distribution of queries.