> Without a natural feedback loop, interviewing mostly runs on myth and survivor bias.
It is a law of human nature that a chaotic environment breeds magical thinking and ingroup allegiance, and there are few business environments more chaotic than Silicon Valley, where startups succeed or fail seemingly at random. The noise calls for reason, and because there are no true known causes for success our mind demands some explanation. How many times have we read the line -- often written by some very smart people -- "correlation isn't causation, but we've found a strong correlation between....". What follows the but in that sentence is what is known as superstition. Correlation doesn't equal causation, but no buts follow. Actually, correlation plus confirmation bias equals magical thinking, and "data driven" correlation (with no confirmation bias) equals bigotry. Failing to realize that correlation without causation provides us with exactly zero predictive abilities (it might be "predictive" on unknown present data, but not on future data) is the root of a lot of evil. In fact, "data driven" correlation -- because it disguises itself as knowledge -- creates (or, usually, reinforces) a reverse causation.
> Whatever else one can say about the Mirrortocracy, it has the virtue of actually working, in the sense that the lucky few who break in have a decent rate of success.
This is the worst of all fallacies belying the "SV logic". Even supposing it were true, a large-number statistical observation says little or nothing about the behavior of a single random variable, or a single startup in this instance. Yes, the startup system "works" (for whom is another question) on the whole, but the vast majority of individual startups still fail. Learning back from correlative observations on the large system and implying the so called lessons to an individual company has little grounds in any rigorous reasoning.
> You can protest your logic and impartiality all day long, but the only honest statement is that we're all biased.
Understanding this is one of the keys to progress. How many times have we seen posts discussing sexism in tech deemed "controversial" here on HN? But the truth is that the chance a member of a society imbued with biases for millennia is not sexist (or otherwise biased) is extremely slim. The only way to fight this bias is to seek it out and see it (because it's there).