Currently epsilon is a Linux kernel with musl and tcc. Since tcc can be used for scripting, I believe that with a non-trivial amount of work (including patching musl), musl could be built by tcc. Therefore epsilon would become just Linux and a static build of tcc.
To further reduce epsilon (or, in somewhat standard terminology, the "binary seed") I am also writing asmc (https://gitlab.com/giomasce/asmc), which is a minimal operating system with a minimal compiler, totalling less than 6 KiB binary seed. This compiler is not for C, but for G, which is a C-inspired language that I invented for this purpose (see the project README and links for more details). The target for asmc is to be able to build and boot Linux and a static copy of tcc (the latter is more or less already there; Linux is, of course, the hard part), at which point it can bootstrap nbs. I'm not there yet, though.
Also, to be fair, although asmc's seed is 6 KiB, you still need an x86 CPU (I'm not aware of free implementations), a loader and a BIOS to have a functional environment. This is not little thing, but I guess it is a reasonable target so far.
To further reduce epsilon (or, in somewhat standard terminology, the "binary seed") I am also writing asmc (https://gitlab.com/giomasce/asmc), which is a minimal operating system with a minimal compiler, totalling less than 6 KiB binary seed. This compiler is not for C, but for G, which is a C-inspired language that I invented for this purpose (see the project README and links for more details). The target for asmc is to be able to build and boot Linux and a static copy of tcc (the latter is more or less already there; Linux is, of course, the hard part), at which point it can bootstrap nbs. I'm not there yet, though.
Also, to be fair, although asmc's seed is 6 KiB, you still need an x86 CPU (I'm not aware of free implementations), a loader and a BIOS to have a functional environment. This is not little thing, but I guess it is a reasonable target so far.