Sure, the startup "circles" regard an idea as a short description of a sentence or two to a family member, neighbor, or prospective user/customer of how the new product/service will look to them.
But, then what about how to make that product/service work? Ignore that? That part should be crucial: Crucial for having the first good or a much better product/solution, a technological advantage and barrier to entry, etc. That "idea" could take original research, be regarded as valuable intellectual property and protected as a trade secret, etc. Sure, the user/customer need not be aware of that "idea" maybe in the internals of the product/service.
Then to say
"The very best ideas usually don't seem worth stealing."
ignores the crucial, core ideas, say, the details of the coating (likely highly classified and from a lot of R&D) and shape (again from a lot of R&D, and Maxwell's equations computations, likely challenging and classified) that were key to stealth.
Or, here's an "idea": A safe, effective, cheap one pill taken once to cure any cancer. Sure that "idea" is just pie in the sky and worthless. But to create such a pill will take one heck of a good idea. To me it is just such good ideas that should be central in information technology startups. It sounds like Sam is neglecting that any such ideas are possible, relevant, valuable, or important. Or, to me, an idea for an algorithm that shows that P = NP is astounding, and astoundingly valuable -- implement it on a server and behind a lot of security and just sell the results, say, anyone who wants to send in any integer linear programming problem, that is, any of the thousands of important real world examples of NP-complete problems.
My favourite pure good idea with enormous value has to be the Kelly criterion [1].
I'm trying to figure out why I didn't know this.
The Kelly criterion seems to not be well known - why I can't say other than there are a lot of groups that benefit from people being ignorant of it.