Reminds me of my recent vibe coding project of porting Winamp AVS to web using Emscripten. It worked so well that it was simply.. done after a few sessions, and I was amazed and also feeling fuzzy in the head.
One turn involved ChatGPT explicitly fixing C++ source code to fix beat detection for the available audio sources after I complained that it was broken.
Hands down, sure, the most respectable work, apart from the original AVS code, was done by the Emscripten contributors. Anyway, it readjusted my expectations, even after my day job switched to agentic coding as the primary development tool.
The companies looking at this will just turn to cloud gaming services instead for newer games. They have tried it with Stadia, which that failed because of the existing consoles.
Now the games industry will try it again for real and find a way to rebrand cloud gaming and trick everyone to believe its better than running games on your own machine, only for them to stream the game back to you without you ever owning it.
Reminds me of my recent vibe coding project of porting Winamp AVS to web using Emscripten. It worked so well that it was simply.. done after a few sessions, and I was amazed and also feeling fuzzy in the head.
One turn involved ChatGPT explicitly fixing C++ source code to fix beat detection for the available audio sources after I complained that it was broken.
Hands down, sure, the most respectable work, apart from the original AVS code, was done by the Emscripten contributors. Anyway, it readjusted my expectations, even after my day job switched to agentic coding as the primary development tool.
The companies looking at this will just turn to cloud gaming services instead for newer games. They have tried it with Stadia, which that failed because of the existing consoles.
Now the games industry will try it again for real and find a way to rebrand cloud gaming and trick everyone to believe its better than running games on your own machine, only for them to stream the game back to you without you ever owning it.
It makes perfect sense that LLMs are good at this. We've always known that a binary is not a true moat, but that it's just a lot of toil to transpile.
Compiler output is highly predictable, but it doesn't make the inverse true. LLMs are not themselves compilers.