Hung out with the Aurora Silicon crew tonight We got 4k DisplayPort Video working over Thunderbolt Works on my CalDigitTS3, more testing soon They've forked Asahi and are going to take it to new heights We're fixing everything

I put Omarchy on my 9 year old's laptop (M1 Air 8gb) about two weeks ago. Now he's working directly with Claude to make games. I am not intervening Normally on a low key Sunday he'd be playing Minecraft, but right now it's all Omarchy and making games A sign of what's to come

🔥Omarchy Mac Users 🔥 The original maintainer didn’t want AI-generated contributions. Fair enough—we respected that boundary, forked the project, and got to work. The result after roughly 21 hours: • 29 PRs merged across two repos • 115 new commits beyond the inherited baseline • 7 PR authors contributing in parallel We now have working Apple Video Decoder acceleration for H.264, HEVC and VP9 on M-series Macs running Asahi/Omarchy, plus opt-in H.264 High 10. The work includes: • Chrome’s solid-green-video bug fixed • 10-bit HEVC blank-frame bug fixed • HEVC tile and wavefront hangs fixed • H.264 multi-slice firmware hangs fixed • VP9 state, color-range and submission handling fixed • Decoder lifecycle races and cross-context reference aliasing fixed • Multiple memory-safety, error-handling and malformed-stream bugs fixed • 15 AVD kernel patches • Automatic installation, recovery and kernel-update rebuilding • Sanitizer CI, guarded hardware testing and reproducible conformance infrastructure On real M1 hardware we’re currently bit-exact on: • HEVC: 144/147 streams • AVC: 73/135 • VP9: 216/305 • 864 additional generated hardware-frame comparisons Huge credit to the contributors: davefano, grudev, laihenyi, stefanakerwall, thebytorsnowdog and service-cyber—as well as the upstream and carried work from megi, sofus13, Igor Ryzhkov, Ante042, Aaron/aquarat and the Asahi Linux contributors. This is what agent-native open-source development can look like: humans set direction and validate reality; agents attack a shared roadmap in parallel; every claim is tied to tests and hardware evidence. We didn’t argue about whether AI belongs in open source. We forked it and demonstrated what it can do.