As a long-time RE this does not surprise me. It feels like an eternity ago that the MCPs for various reverse engineering tools dropped, but from the onset it seemed like a perfect fit. While we've had ML-assisted solutions before, nothing of this scale comes close, and as the recent Linux kernel presentation mentioned, a lot of it boils down to a massive effort-by-exhaustion and pattern-matching. This is the golden road for RE - if you know your compiler versions, processor arch, and you have something trained on countless permutations of different complexities, you're going to hit the right notes.

Yesterday's news on the mathematics breakthroughs were a foregone conclusion - there is a lot of hidden overlap between it and RE. RE is just the more obscure of the two, but among my circle of peers, many people have already undertaken (and succeeded) at very ambitious projects. Obvious standouts are things like Modern Warfare which gets mentioned because it is highly popular, but many games have already been reversed by the quiet majority.

I don't see many options for PS/Nintendo. Some form of bespoke-hardware-based lockdown reminiscent of the Cell processor, perhaps. But that is a lot of R&D to spend on something like this. Will have to wait and see.

So, what to make of this? Machines of the future or stochastic parrots? Of course. I'll take two.