One of the threat models is not YC itself stealing your IP. Which they are to some extent, but them using vendors that themselves train their models on your IP. Or vendors that use vendors that...

>YC spends unimaginable amounts of money and effort each year to keep doing that, as it has done for two decades.

Yeah, that or they built their reputation early on and are now coasting on that. It's an industry whose explicit objective is to build companies (build reputations) and then exit and sell them, would you be surprised if YC itself did that?

>It doesn't even pass basic logic that YC would steal some gem of an idea from someone's transcript and hand it to someone else to build.

More likely is the opposite, they get 100k startup applications and they know what ideas people are pursuing and should be avoided, and which might be worth looking into. The former become would target demographics for shovel selling as well, so the 1% that actually gets into YC might setup shop selling to the other 99%. It's theoretical, but these are all attacks that emerge from the basic violation of sharing your source code. To be clear, I don't think that YC employees are starting startups, the threat model is that they act through their picks, and they strengthen their picks at the expense of their non-picks. It's like a russian roulette of startups, which isn't necessarily a bad idea.

>Everyone who is talented enough to build it is busy building products based on their own ideas,

Like, I get that, but also don't overcorrect into the naïve take that market intelligence is worthless or doesn't exist, especially when that take is 20 years old, if you treat that as your strategic advantage over the population, it's more likely that it's going to be used to take advantage of you.

Be moderate, share your idea, don't share everything. Share to individuals, don't share to massive data gathering systems that may use your data as training data for machine learning models. Basic stuff really.