The thesis
ravynOS is building a Mac you can run Mac software on.
Russet is building a Mac you can run.
Why this exists
macOS Tahoe 26 is the final release supporting Intel Macs. The successor is ARM-only.
This is categorically different from every prior round of Apple dropping
hardware. Those were feature gates -- a board-id
check, a missing kext, a Metal capability flag -- on machines whose
silicon could still execute the code. OpenCore Legacy Patcher existed
because the x86_64 binaries were there and somebody just had to talk the
installer into laying them down.
An ARM-only release ships no x86_64 kernel and no x86_64 userland. There is nothing to patch toward. Rosetta translates x86 to ARM, never the reverse, and the reverse at daily-driver speed is not an engineering problem, it is a wish.
So OCLP ends here -- not from lack of effort, but from absence of code. Apple committed to roughly three years of security updates for Tahoe, which puts the cliff around 2028. Millions of functioning machines will need an operating system that still receives network security updates.
That is the Windows 10 -> Linux Mint moment, aimed at Mac hardware. Mint did not win those refugees by being first. It won by being ready, and by being kind to people arriving under duress.
Design principle
Inherit the interaction grammar, not the costume.
Every failed "macOS-like Linux" copies the visible nouns -- dock, menu bar, icons -- without the syntax connecting them. The eyes are fooled; the hands never are. Russet preserves the mental models and the muscle memory while being visibly, proudly its own Unix-like system.
Hardware compatibility gets the machine to boot. Cognitive compatibility gets the person to inhabit it.
- Macintosh interaction grammar.
- Linux sovereignty.
- UNIX composability.
- Original visual identity.
- Zsh by decree.
The grammar is specified as pass/fail rows in a conformance document, so that the manifesto can fail a test rather than merely read well. A row describes behaviour a user's hands can verify, never implementation. "Uses Plasma's global menu applet" is not a row. "The menu bar shows the frontmost application's menus" is.
Where it actually stands
- Started
- 31 Aug 2026
- ISOs built
- 5
- Conformance
- 4 / 62
- Installer
- None
62 conformance assertions. Four passing, two failing, the rest untested. The failures are the point of the document working: Russet has no menu bar yet, and it still boots to a default theme. A spec that cannot fail is decoration.
Done
Five ISOs built on an Arch host under VMware Fusion, starting from a
baseline releng fork to prove the toolchain. Russet now boots
unattended to its own named desktop: a live user, SDDM autologin, zsh as
the login shell, and no password prompt on the way in. The Broadcom Wi-Fi
quirk (brcmfmac.feature_disable=0x82000) is baked into every
boot entry by default rather than documented as a workaround, and no boot
menu says the name of the distribution it was forked from.
The terminal ships four fixed-pitch faces from inside UNIX history -- Sun's Gallant from the SPARCstation console, the X11 9x15 that GNU Emacs made familiar, a trace of the DEC VT220 ROM, and Keith Ohlfs' NeXTSTEP face -- all public domain or BSD, each rendered at its own native pixel grid. Inheriting a lineage is not the same as wearing a costume.
Next
The menu bar and the Command-key layer -- the two sections that decide whether any of this is more than a renamed Plasma. Both are now testable, which is what the autologin desktop was for.
Not built yet
Russet has no installer. An archiso image installs nothing; it boots.
The shape of the answer is now settled: Russet's configuration becomes a
package, so that what you boot and what you install are the same thing by
construction rather than by copying a live medium. A scripted
archinstall preset proves that boundary first; a branded
Calamares comes after, because the installer a person meets while leaving
macOS under duress should not be a text menu. The Btrfs
+ snapper rollback story still waits on it.
Reference machine
MacBookPro14,3 -- 15-inch MacBook Pro, 2017, Touch Bar.
Apple T1, not T2, which deletes most of the horror
stories: the T1 handles only the Touch Bar and Touch ID, and the SSD is
ordinary NVMe on PCIe rather than locked behind a security coprocessor.
On a 6.15+ kernel this machine is essentially all mainline.
| Component | Part | Linux |
|---|---|---|
| CPU | i7-7820HQ, 16 GB | -- |
| iGPU | Intel HD 630 | i915, in-tree |
| dGPU | Radeon Pro 560 | amdgpu, in-tree -- AMD-not-NVIDIA is the luckiest fact here |
| Keyboard / trackpad | SPI | applespi, upstream since 5.3 |
| Touch Bar | T1 / iBridge | apple-ib-drv, upstreamed in 6.15 |
| Wi-Fi | BCM43602 | brcmfmac, in-tree; needs one quirk |
| Audio | Cirrus CS8409 | patch_cirrus, upstream |
Accepted casualties: the FaceTime HD camera, whose driver needs firmware that cannot be redistributed on an ISO, and Touch ID, which has no driver.
Scope, stated honestly
The floor, and the real target: a build that is genuinely nice to use on one MacBookPro14,3. Falsifiable, achievable, and sufficient.
The public ambition: a general-purpose distribution for orphaned Intel Macs.
The far horizon, stated as a dream and not a plan:
binary compatibility with .app bundles.
Maintaining out-of-tree kernel patches over time is not a cost to be minimised here. It is the point. This project is also an apprenticeship in free-software maintainership.