I'd actually characterize the problem the opposite way as you: there aren't 10x ReactJS/Angular2/MongoDB/TensorFlow developers, because when we say "10x", we really mean "They delivered the same revenue with 1/10 the code/time/money" or "They delivered 10x the revenue with the same amount of code/time/money."
That usually means thinking of out-of-the-box solutions that reframe the problem in a way that's dramatically easier to solve and often more useful for the end user. For example, they may realize that your customers don't want a web UI that they have to visit separately at all, they just want to receive a text message when certain events happen, and so they can replace your entire UI team with a single Twilio API call. Or they may realize that all of the complicated algorithms you have a whole backend team slaving over can be replaced by a deep-learning network that's mostly pre-trained over ImageNet, and just needs some fine-tuning for your problem.
Usually this requires deep knowledge across domains, not deep knowledge within a domain.
Tendency to think out of the box is a major problem when trying to be hired, if you are a software developer. Somehow most companies have strictly shaped role-"holes", so developers should develop and do not ask questions, managers should manage and do not code, and we already have smart guys, thank you.
There isn't that much room for out-of-the box thinking in software development (not enough for a 10x boost).
Developers who wander off the yellow brick road don't usually end up in a nice place. The term "lone wolf" is a derogatory term used to describe the kinds of developers who think too far "out of the box".
The best technical ideas are usually derivatives of much older ideas which require deep niche domain expertise to understand.
People say this, and they make career decisions that box them into a certain career path, and then they assume that because that's the only career path available to them, that's the only career path. Meanwhile, other people make career decisions on the assumption that they will get to do the work they really want, and they usually have to work a lot harder to make it happen and face some nerve-wracking moments, but a good many of them actually do end up getting to do what they want.
That usually means thinking of out-of-the-box solutions that reframe the problem in a way that's dramatically easier to solve and often more useful for the end user. For example, they may realize that your customers don't want a web UI that they have to visit separately at all, they just want to receive a text message when certain events happen, and so they can replace your entire UI team with a single Twilio API call. Or they may realize that all of the complicated algorithms you have a whole backend team slaving over can be replaced by a deep-learning network that's mostly pre-trained over ImageNet, and just needs some fine-tuning for your problem.
Usually this requires deep knowledge across domains, not deep knowledge within a domain.