The direct quote is:
>Thiel described the argument Zuckerberg finally came down on like this: "[Yahoo] had no definitive idea about the future. They did not properly value things that did not yet exist so they were therefore undervaluing the business."
Yahoo's market capitalization in July 2006 was $42.51 billion. A 22 year-old presumed they were stupid, and he was right. 
Today FB has a market cap of $264.91B and Yahoo? Down to $35 billion after 9 years of growth.
 by the way to get the market valuation at the time, I did this search: http://www.wolframalpha.com/input/?i=what+was+yahoo%27s+mark... I can't believe it worked! I used wolframalpha because this is the kind of search they promise they can answer - and they were right, they actually delivered. Nobody else on the face of the planet does this, and it shouldn't even be possible. But it is. If you think something is possible, JUST DO IT. If you think your competitors are stupid (compared to what you think you can do), you're probably right. (or you wouldn't have that thought.)
... except for the 50 people in the world who know something that Nike doesn't. Those 50 people who do have an answer to that question are the people who can topple them by doing something better.
Nobody delusional can explain something clearly, and answer all your questions about it - if someone clearly explains something and you think they're delusional for it, likely you're just not qualified to judge. You may be the kind of person who, if you got back sent 1,000 years into the past, would learn to ride a horse and learn some trade, because it's obvious that recreating anything from the future is "impossible", or someone would do it. Nothing wrong with that. Just doesn't apply to startups.
That advice in the context of startups is simply completely wrong, irrelevant, and inappropriate. Couldn't disagree with it more.
Finding someone who happened to ask the exact same question by doing an Internet search is a different problem entirely. Wolfram Alpha can instantly answer questions that require certain levels of deduction or calculation, and which have never been asked in the history of the world. Google can do that too: for mathematical equations like 453442433 * 234523432 + 3485478347434879. Applying this to questions of a general type that require real-world data is miraculous.