Apple locks fast local AI to macOS with Metal. I picked the lock. metal2vk translates Metal GPU code to Vulkan, unchanged. 598 of 610 real programs from @trymirai's uzu and @ashxhart's TensorFold so far. Engines built for Macs now run on Linux. github.com/joshuaswarren/met…

We are working to support all of the M chips. Right now M1 and M2 are best supported, M3 is getting close to that level too. M4, M5 and M6 are still in the works.

For now it says for dual booting but you don’t have to boot to it.

Right now the standard install of Omarchy keeps macOS so that you can dual boot and flip to it when you need it. The vm idea is an interesting one… This has been done entirely using omp, herdr and mainly Opus 5.5 and Sonnet 5.5. I use other models here and there.

You can use nothing but Omarchy but we recommend keeping it dual boot just in case. On my daily driver MacBook Pro I resized my drive to give most space to Omarchy and haven’t rebooted to macOS in a long time.

TouchID is working on the M1s and M2s and I am sure we will get it working on the entire line over time.

I'm on the Omarchy M team, doing the driver implementation work so that future users can just use our graphical installer to install it in one click. So, yeah, five days to pave this trail is something I'm OK with.

Grok's right - except for one thing. We maintain our own fork of Honeykrisp for Vulkan now that resolves a number of performance issues in the upstream Honeykrisp implementation. Should have GPU up in the next 1-2 days.

Thanks!

Soon. Days away for really well tested M3 support. Maybe less. It’s in final testing now.

As someone whose first paid programming job involved building with PHP in 1999, I wish we could. True story, in 2008 I was trying to decide if I should double down on PHP or switch to this “Ruby on Rails” thing I had been hearing about and I went PHP.

M3 support is just about here. Other members of the @OmarchyMac team have been testing it deeply and it’s the best supported chip behind only the M1 and M2 right now.

@elshayib_ @dhh

And since I posted this morning Ethernet is up now. Still chipping away at the GPU.

M1 thru M3 have the best support but we are working to close the gaps with the M4, M5 and M6 as quickly as we can.

It is! M1 and M2 have GPU drivers that work well. @tiredkebab has been making great progress on the GPU for M3 as well.

'Basically' is doing a lot of work in that sentence. 😁 If you're running a large number of AI agents or trying to use all your memory for a local LLM, macOS can slow you down. I'm also a big believer in people being able to run the software they want on the hardware they own.

I may never be able to upgrade the GPU in my gaming PC but I'm excited how @nvidia is finding ways to bring AI compute into laptops. I've got a backlog of things to test right now - focusing on eGPUs for a bit, but at I want to get my hands on an @NVIDIARTXSpark laptop to test.

Oh yeah, I'm looking forward to that Apple OLED hardware. I'm sure it'll take a few days of hacking to get it supported but hopefully I can get my hands on one early and bring Omarchy to it quickly!

Linux on the M6 Mac Mini, Day 5 - success! Omarchy is installed on the Mac Mini's hard drive. It boots, it runs. Still bringing up accessories, including Ethernet and GPU, so my work's not done, but 5 days from the macOS About screen to the Omarchy About screen isn't bad!

Awesome! If you hit any rough edges please open issues so we can get them improved.