[eropple]

I consult for companies from ten-person startups to a few thousand employees, and, from my perspective, it's all the same thing. While everybody pretends their particular organization size and type is supreme, it all comes out in the wash. Sure, startups move marginally faster. The time for reflection and self-analysis is lesser and most people seem to get less out of what they do, learn less from their mistakes, have generally poorer leadership that can't devote the time to helping more inexperienced employees learn the lessons they should learn from those mistakes and help them grow both as practitioners of a craft and as human beings.

And I'm gonna be honest: the assumption that a mentor has fear-based motives for ensuring that your work has greater benefits than the risk it poses is so utterly unrelated to reality in the general case, so monstrously projective and in itself pathologically fear-based, that I have trouble believing somebody would state it in good faith as such a universality. (To the newbie: your highly-self-regarded work is risky. Unknown commodities are always risky. If you are not analyzing your actions as a leader through the lens of risk, you are a bad leader. Many startups don't do this. They have bad leaders, too. But the understanding of and formalization of this is not something taught to entry-level employees at that move-fast-die-fast startup, because all too often, nobody else knows about it either.)

This is not to say that there are not startups where one can learn and thrive, because obviously they exist in some proportion. I would say, and the failure rates of startups and the it-was-my-first-job candidates I've seen ejected from the crashing wrecks strongly push me in this direction, that it certainly seems significantly harder to learn how to be directed and effective (as opposed to thrashy and high-effort) in such an environment. But it's certainly possible. Nor would I say that there isn't a very good time to leave a big company--I bailed two years to the day and probably stayed six months longer than I needed to in order to be a good engineer rather than a good coder.

But there is no generalization that holds, in either direction, as universally as your post asserts, and the assumption of such is foolish.