He isn't suggesting letting them deliver a triple-scoop of their enchanting story, he's suggesting that you filter it through your own logic & experience and try to poke holes in it.
Elon Musk was once asked in an interview "How do you judge whether a highly-technical expert really knows what they're talking about?" His reply was along the lines of "You ask them questions of progressively increasing levels of detail. Someone who's actually an expert will understand a subject on many levels; they'll be able to give you the ELI5 version, but then will have good answers every time you probe a level deeper."
Larry & Sergey's elevator pitch was "We built the best search engine in the world", but if you asked them "What's it running on? How many machines? How many pages does it index? How do you handle duplicates? How many queries/day do you get? What are people looking for? How do you rank pages?" etc. you'd get good answers for all of them.
Elizabeth Holmes's elevator pitch was "We're going to revolutionize blood testing with finger-stick tests", but then if you asked "Which tests do you run? How many molecules of __ are in a drop of blood? How much variance is there between people?" you would very quickly run into either evasions or outright lies. Once you start asking questions on top of a lie their story usually falls apart pretty quickly, because few people can remember everything they just made up 5 minutes ago and still make up some logically coherent detail on top of it. (This, BTW, is why improv theater is so hard, and why most improv shows are comedy, where it's just funny if you end up with a non-sequitor.)
The same technique works well for reference checks, but you can't blindly trust whatever the reference tells you. Most people cherry-pick their references; it's easy for a charismatic charlatan to find some people who are still super-enchanted by them. Many people are also pretty unwilling to talk (particularly about litigation-happy scam artists like Holmes), for fear of defamation lawsuits.
Elon Musk was once asked in an interview "How do you judge whether a highly-technical expert really knows what they're talking about?" His reply was along the lines of "You ask them questions of progressively increasing levels of detail. Someone who's actually an expert will understand a subject on many levels; they'll be able to give you the ELI5 version, but then will have good answers every time you probe a level deeper."
Larry & Sergey's elevator pitch was "We built the best search engine in the world", but if you asked them "What's it running on? How many machines? How many pages does it index? How do you handle duplicates? How many queries/day do you get? What are people looking for? How do you rank pages?" etc. you'd get good answers for all of them.
Elizabeth Holmes's elevator pitch was "We're going to revolutionize blood testing with finger-stick tests", but then if you asked "Which tests do you run? How many molecules of __ are in a drop of blood? How much variance is there between people?" you would very quickly run into either evasions or outright lies. Once you start asking questions on top of a lie their story usually falls apart pretty quickly, because few people can remember everything they just made up 5 minutes ago and still make up some logically coherent detail on top of it. (This, BTW, is why improv theater is so hard, and why most improv shows are comedy, where it's just funny if you end up with a non-sequitor.)
The same technique works well for reference checks, but you can't blindly trust whatever the reference tells you. Most people cherry-pick their references; it's easy for a charismatic charlatan to find some people who are still super-enchanted by them. Many people are also pretty unwilling to talk (particularly about litigation-happy scam artists like Holmes), for fear of defamation lawsuits.