A from-scratch Linux distro that turns a phone into a standalone AI-agent device ("Jarvis phone edition"): voice in (STT) -> agent CLI -> voice out (Piper TTS), autonomous camera, per-agent memory, retro block-font UI on a Wayland compositor.
First dev target is x86_64 in a VM; the hardware target is the Redmi Note 12 Pro 4G "sweet" (2209116AG, a rebadged Redmi Note 10 Pro - Snapdragon 732G / SM7150-AC, Adreno 618), pending bootloader unlock.
See docs/ROADMAP.md for architecture decisions and milestones,
and MEMORY.md for the full design brief.
configs/ Buildroot defconfigs (neuros_x86_64_defconfig)
board/neuros/ per-arch kernel config, grub/genimage, post scripts
overlay/ rootfs overlay (branding, systemd drop-ins, networkd)
package/ NeurOS's own Buildroot packages (agentd, compositor, ...)
scripts/ helper scripts (make-vm.sh)
buildroot/ Buildroot 2026.02.3 LTS, cloned, git-ignored
output/ build output, git-ignored
make config # load neuros_x86_64_defconfig
make # full build -> output/images/disk.img (~30-45 min first run)
make vm # convert to VDI + create the "NeurOS-dev" VirtualBox VM
make run # start the VM
# after boot:
ssh -p 2222 root@localhost # password: neuros
tail -f output/serial.log # serial consoleBuild parallelism is capped (JLEVEL=8, nice/ionice) to keep the desktop
responsive; override with make JLEVEL=N.