Hey thanks for trying out some searches. Depending on the intent of your search, we try to predict the best source of knowledge to find an answer. Computational or simple knowledge graph queries like those will often get routed to Wikipedia or Wolfram, or even DDG's Instant Answers API. Based on API costs and intent categories, Wolfram queries are about 0.25% of queries, and Wikipedia is about 17%. We're an alpha with only a small user base, but I suspect your questions were a little unrepresentative of typical use.
Other searches use vertical searches or specific APIs (weather, time, computation, organizations etc), or are use our own indexes (especially news and navigation).
More complex questions will be more likely to trigger our question-answering model, which works a bit differently by going out and searching for content from live sources in real time.
Here are a few fun searches to try that will trigger different approaches.
Extracts from web content:
what US state grows the most potatoes?
how many starlink satellites are in operation?
Complex answers (these take longer as they are searching through content in real time:
how many starlink satellites are in operation?
why is ethereum moving to proof of stake important?
what are the first steps towards consolidating your graduate student loans?
What is the frequency of birth defects in babies born to men who had a type 2 diabetes diagnosis who are taking metformin?
(this is really overwhelmed right now so the cluster is dropping about 30% of requests - so it will may fall back to keyword searching until we figure out handling the extra load).
As you can see, there is kind of an inverse pyramid here, where the more complex your queries, the more complex the handling to answer the question.
Wolfram is great for knowledge graph and computation queries, but you can see the approach is very different with those last questions.