I mean, for a “clean room implementation” you could offload the work to Claude and GPT, you have one model describe the assembly into a spec file, you pull down

I mean, for a “clean room implementation” you could offload the work to Claude and GPT, you have one model describe the assembly into a spec file, you pull down open source programs from Planet Source Code that SHOULD compile, and then you have it ensure it compiles 1:1. Then you watch and let it run through the process in a few weeks. If you dont care about clean room then you plainly disassemble it and translate it. I live in the US where the second option might not be legal.