nx_nishi_usb_image.nx

nx_nishi_usb_image.nx -- x86 ladder R9-FUSE: the REAL bootable Nishi USB image, proven by execution. Until now three proofs sat SEPARATE: nx_nishi_bootimg (R9, a 512B MBR that only prints+halts), nx_emu_x86_stage2_test (R9-4, an INT 13h multi-stage loader, but only as an IN-MEMORY test), and the desktop kernel (which lived in no bootable image at all). This organ FUSES them into ONE persisted multi-sector artifact -- knowledge/status/nishi_os.img, the file you would dd if=nishi_os.img of=/dev/sdX to make a bootable USB -- and PROVES it boots end-to-end by EXECUTING THE ON-DISK BYTES on the sovereign 16-bit real-mode + BIOS-INT emu: sector 0 (MBR) --INT 13h AH=02--> loads sectors 1+2 to 0x8000, then jumps to stage2 sector 1 (stage2 loader) prints the NishiOS banner over INT 10h, then transfers to the kernel sector 2 (kernel) prints 'NISHIOS KERNEL' over INT 10h, then HLT A 3-stage chain (MBR -> loader -> kernel) is a genuine step past the 2-stage stage2 proof, and it is the REAL PERSISTED artifact: written to disk, re-read, then executed -- not an in-memory image. KAT: (T1) persisted byte-faithfully + 0x55AA signature; (T2) INT 13h actually copied BOTH stage2 and the kernel off disk (0x8000 and 0x8200 went 0 -> 0xBE); (T3) the LOADED kernel ran (console contains 'NISHIOS KERNEL'); (T4) clean HLT. NEG CONTROL (T5, liar-kill): a sibling image whose MBR loads only ONE sector never reaches the kernel banner -- so the banner in the good case MUST have come from the sector that was loaded off disk, not from the emu or a hardcode. HONEST SCOPE: every stage here is 16-bit real-mode. The next rung is the mode-transition handoff -- stage2/kernel set up GDT + CR0.PE + PAE + EFER.LME, enter long mode, and load the real 64-bit desktop kernel (nx_nishios_boot_gui) instead of a banner printer. This rung delivers the loader chassis + the real flashable artifact that that kernel will ride. NEVER-BRICK (Rule 26): writes a FILE (knowledge/status/nishi_os.img); models INT 13h *reads* only; touches no /dev and no host firmware -- by construction, not by promise. The 16-bit + BIOS-INT13h/INT10h emu below is inlined verbatim from the proven nx_emu_x86_stage2_test (a _test's main() blocks `import`, so we inline -- the same DRY exception the codebase already uses). expect_exit: 0 license_tier: ORIGINAL