From having seen what happened during the development of FirefoxOS (Boot2Gecko), from a developer point of view. My take is that the biggest hurdle are not the development of the core applications, but the approval and certifications from the various partners.

For example, some company did not like the fact that the call stack could be reverse engineered … yet they would not help you build a prototype at the pace of development (and this was back in 2011)

Then once you have a working OS, the second hurdle was to demonstrate a very high number of active users, such that they dare to even consider your new environment as a potential target. This a bootstrap challenge which is hard to achieve. To get traction from App developer implied getting traction from users, which implied having apps the users care about …

While you goal of removing the application layer sounds like a nice way to work-around the bootstrap problem, I fear you will still have it at the API level.

I wish you good luck, and I hope the landscape has changed in the last decade.