I have new intentions on the front of non-Win10/11 developments/versions, based on my new awareness that MoltenVK is mature and viable. As a refresher, it wasn't until a year ago last fall where we decided to bite the bullet and undertake the massive investment in writing a next-gen AAA-grade modern optimized D3D12 engine from scratch. I intended to release with the D3D11 engine and only pursue a D3D12 rewrite as an academic task later. I am very happy with this choice as WSV3 Tactical Mesoanalyst graphics engine is truly world class, and I'm even as of the last week using the dedicated D3D12 COPY queue (not possible in D3D11) with measured texture streaming throughput increase over serial texture uploads on graphics queue only even on iGPU (which lacks a distinct DMA copy engine for truly concurrent graphics-copy work), because it can still overlay copy work during the frame present interval. I am actually measuring 1.5x-2x the texture streaming throughput even on Intel Iris Xe graphics from using the D3D12 COPY queue. I spent pretty much the last whole week solely getting that working, so a year later and the payoff of investing in D3D12 is finally fully measurable.
So anyways, staying on D3D11 would have given immediate MacOS support if you buy Parallels, which works great with D3D11 apps but still no planned support for D3D12.
So then when we did the D3D12 rewrite I knew this would eliminate Parallels support, but to be the best you have to be the best and cannot be held back by things like that. I said we could decide in 2026 to either do a D3D11 backport purely for Parallels support, or invest in a full-scale native Metal engine from scratch and MacOS port. Linux/Vulkan would be third and last, mostly as a flex than true market need.
But given the fact that a POSIX-compliant port of the platform layer (everything but the rendering engine) already gets us 90% of the way to Linux and MacOS support besides the renderer, and because MoltenVK works great, clearly the best path forward is to invest in a native Vulkan rendering engine and POSIX-compliant Linux port of the platform layer. From that point, it would run on Linux machines using the most cutting-edge, modern graphics API (Vulkan), and MacOS from there is pretty easy.
So the current vision is, the flagship remains Win32 x64 D3D12 gold standard for time being, but with a single secondary platform layer (POSIX compliant for Linux and MacOS), and a single secondary rendering engine implementation (native Vulkan, which runs beautifully on modern MacOS through MoltenVK, and no need for a native Metal rewrite).
Vulkan is also much more 1:1 correspondent to D3D12 than Metal, which has a totally different API model. So the extreme difficulty and intensive work typically associated with a custom bespoke from-scratch native Vulkan engine is largely work that I have already done. And I can independently build and test the Vulkan engine on Windows 10 before dealing with the POSIX rewrite of the platform layer.
This is the best of all 3 worlds. It's also nice to be able to use and test on Linux before I even need to buy a Mac, at which point most of the work is done. And having the native Linux version supports my vision for having an industrial service where we render graphics on a schedule on server and upload to S3 for your website or company/municipality display, for clients who need non-interactive image/video clip outputs in an automation context. But also of course to support interactive use of the graphics client on Linux.
I'd unlikely start this work before mid-2026.