The point of decomps/recomps was never for some altruistic reason related to skill or art.
The point was people wanting to play games. This just adds to that. Yes - the people who did it before were skilled, smart and deserve recognition. But these ports are exactly that - ports so that people can play their old games or relive them.
I'm thinking of Matrix here, just with games instead of chicken. Is it actually the game itself, or is it a machine's assumption of what the game is supposed to be?
With the decomp/recomp/mod/RE in general scene, the people involved were at least highly knowledgeable of what they were doing and they knew where their limits were and what needed to be done.
The chatbot however will yield a happy "done, did the needful" and deliver a pile of garbage instead.
When you automate the process, the only interesting thing about the whole operation dies. I already knew you could reverse engineer C and I already knew emscriptem exists, it's the manual work that made it worth checking out.
The Mona Lisa is an interesting piece of art people come to visit France for. MonaLisa.jpeg is a good wallpaper for about a week at best.
Before LLM-assistance there were barely any fully playable recomps.
How many recompilation projects have you helped with? I get annoyed with non-contributors complaining about the people that contribute.
Decomps or decompilation is about recovering source code that can bit for bit recover the same binary asset. This gives you back the source code from just the binaries. This requires the same toolchains, sdks, etc.
Recomps or recompilations basically inline and unroll say a PPC interpreter back into HLL code. This gets you something that modern toolchains can work with, while you replace all the IO layers so you can update the graphics, controller and networking stacks.
Neither of these are transpilation, which normally refers to a source to source compiler that translate the codes laterally in the same semantic strata.
What most folks are doing are recomps and there are existing SDKs for this (pre-ai) but more and more tooling is becoming available to the point at which I think there will be an agentic harness that runs the game under and emulator for realtime delta debugging and visual diffing.
Thanks for the interesting "erm-akshully" on the distinction!
I'm tired of these terribly made AI projects in this space. It's just noise and low quality spam. I miss when these were done legitimately.
Some script kiddie ran this through a chat bot once and probably doesn't even understand the code.
For those who don't understand, watch this segment of DF Retro Super Show - https://www.youtube.com/watch?v=A6gJTdk_BkI&t=165s