It wouldn't be either, really. The smartest people I've worked with are usually not engineers at all. I started my career working on classified defense and inte

It wouldn't be either, really. The smartest people I've worked with are usually not engineers at all. I started my career working on classified defense and intelligence projects and the research scientists driving new capabilities are still the only people I've ever encountered in any line of work doing things that I legitimately felt it would take me a decade to understand. Think for instance of everything Claude Shannon did while he was at Bell Labs. Of the two, though, obviously larger companies are more likely to even be able to fund truly novel research compared to a startup. There's only so much you can do under the constraint that anything you build has to be a component of a usable product you can plausibly sell to a defined market in the next six months.