Omarchy Archive

Search GitHub

Feed / idea

omarchy-mlx: Short Prefill 33% Faster Than macOS Metal

JoshuaSWarren reports Omarchy short-prefill now 33% faster than macOS Metal on the same M1 base, with long-context still catching up and broader Apple Neural Engine model support underway.

More like this

SiliconForested: Omarchy Mac + Aurora Silicon Update

@SiliconForested

SiliconForested posts an Omarchy Mac + Aurora Silicon Update: "We are bringing online new hardware at an incredible pace thanks to AI (SI?) and strong team work from both the Omarchy M Team and the Omarchy Mac community at large." They count 27 releases (19 stable), 553 commits, and 31 PRs (17 merged) "across iconidentify/aurora-linux + aurora-silicon/linux", with 24 people involved, in "5 days of Aurora (Linux on Apple Silicon) kernel work": one kernel for every Apple Silicon Mac now, with M1, M2, M3 and the MacBook Neo all running the same build; M3 Pro built-in display + GPU; in 12.0, M3 Pro external displays, USB-C on all 3 ports + HDMI at 4K, and three screens at once; one m1n1 boot loader for every Mac; Thunderbolt docks on M1/M2 Pro & Max, two monitors per dock, and Studio Display brightness; Touch ID on more M1/M2 Macs, plus reset/enroll fixes; Neural Engine on M1 Max & M2 Max, MacBook Neo Wi-Fi, and video decode on by default; and M3 Air bring-up underway with remote testers. They say it is all built on top of Asahi Linux and that the kernel is a preview of what's to come in their official release; the still is a dark card titled "5 days of kernel" and reading "Linux on Apple Silicon. One kernel for every Mac, 11.17 → 12.0.", with "27" releases, "553" commits, "31" pull requests, and "24" people, a list of version-numbered changes, and a "THANK YOU (GITHUB)" list of contributors.

tcballard: Today in Omarchy (4-Oct-2026) on Packaging the Apple Neural Engine

@tcballard

tcballard's Today in Omarchy (4-Oct-2026) says seeing Omarchy package the Apple Neural Engine rather than merely talk about Apple Silicon gets an M1 MacBook owner's attention, covering Joshua Warren's opt-in omarchy-mac-ml package, now live in edge for aarch64 with MLX for the Apple GPU, Vulkan support, the ANE driver, and the compiler for Core ML models, enabled by default only on the tested T8103 M1, M1 Max, and M2 Max systems, while the proposed Install > AI row is still a draft; it also covers OpenClaw 2026.9.6-3 restoring the previous package layout and Geert-Johan Riemer's fingerprint repair merging into Quattro; the still is a green pixel "TODAY IN" and "OMARCHY" wordmark on black.