Used AI assistance to bring my dead hardware back to life

Developer fbritoferreira published a userspace macOS driver for the original Elgato HD60 capture card, streaming 1080p60 H.264 video and audio into OBS without a kernel extension, Rosetta, or Elgato software at runtime. The developer wrote the driver with AI assistance to reverse-engineer the hardware, which Elgato never properly supported on macOS, particularly on Apple Silicon. The developer framed the project as an example of AI making a stubborn hardware problem achievable in a weekend rather than replacing the developer's work.

I had an old Elgato HD60 lying around. It can capture HDMI at 1080p60, which is plenty if you just want to use an external camera as a webcam with your Mac. The problem is Elgato never properly supported it on macOS, especially not on Apple Silicon. So I made it work. I ended up writing a userspace macOS driver for the original HD60. It gets the card up and running and streams 1080p60 H.264 video and audio into OBS. No kernel extension, no Rosetta, and no Elgato software needed at runtime. Repo: https://github.com/fbritoferreira/hd60-macos https://github.com/fbritoferreira/hd60-macos People keep saying AI is going to take developer jobs. This felt like a better example of what it’s actually useful for: I used it to help reverse-engineer and get a stubborn hardware problems working. It didn’t do the job for me, it just made a weekend project much more achievable.