There are already some attempts at creating new operating systems off of seL4. The most complete at the moment is Genode. There are also attempts to port existi

There are already some attempts at creating new operating systems off of seL4. The most complete at the moment is Genode. There are also attempts to port existing OSes to seL4, such as qubes. None of these are particularly mature. In terms of creating a Linux replacement, drivers are the biggest hurdle. There are some attempts to programmatically port the netbsd suite of drivers, which seems to be the most promising. Apart from that, drivers will likely need to be ported manually. The GNU userland is also unlikely to be ported easily en-masse due to the capabilities-oriented security model, although some projects would definitely be easier than others. Things like SystemD would be a monstrous task. I personally think that apart from drivers, which can use traditional microkernel approaches, it's probably time to start implementing some of the better OS ideas that have come out in the last 30 years, as opposed to just trying to replicate a posix environment. For example, NuShell has a ton of promise, but isn't particularly posix-y.