>They likely contain stolen or copyrighted code.
How would that be possible when it's trying to byte-match the original binaries
How would that be possible when it's trying to byte-match the original binaries
It's not only possible but I would argue highly likely that any LLM has absorbed that proprietary code into its training data making effectively any decomp not a clean room effort if they incorporate assistance from LLMs or coding agents into their tooling. If Nintendo wanted to be particularly onerous they could easily claim that their codebase makes use of proprietary tooling/knowledge and there would be zero recourse for the devs.