The Legend of Zelda: Ocarina of Time has reportedly been dumped on its Switch 2 launch day, with its original release and a 23GB update already circulating online.
Its not really AI so much as developers can copy/paste and vibe code but serious projects still require proper effort. LLMs help developers who use them discerningly but every successful project has more developer than LLM still.
Bloodborne hasn’t had a successful emulation or port since release, then it seems like someone just used Claude to make it happen quickly. I dunno much about it but just what I’m noticing.
You’re not wrong: Decompiling (turning a compressed program like an .exe back into readable code) is a task that AI excel at. That’s exactly what’s boosting development speed of all these crazy new mods and ports.
Good points! Although close. Its not decompilation, LLMs are actually bad at this due to limited context windows. Most articles I see suggest LLMs support coders recreating known outcomes using templates. Devs still do tonnes of work.
Steam argued its mostly due to single mechanics getting recreated, suggesting devs spend more time fleshing out new mechanics as LLMs are really just helping devs recreate workflows.
“Rather than relying on a single mechanism that someone can easily copy, you must continue to grow your game by adding features, personality, and creativity that stem from your own perspective as a developer.”
The cloner cloned used AI to build templates and workflows to copy mechanics. He still did a lot himself. Also because the indie game used asset packs the cloner could use the same.
LLMs can’t recreate complex assets too well either so they still need to steal models and rigs etc for big games.
True, I should have been more specific. LLMs are a great reverse engineering tool as they excel at semantic recovery and readability vs “traditional” decompilation tools. Naturally most work is by an actual dev but LLMs are a huge productivity multiplier.
Its not really AI so much as developers can copy/paste and vibe code but serious projects still require proper effort. LLMs help developers who use them discerningly but every successful project has more developer than LLM still.
Bloodborne hasn’t had a successful emulation or port since release, then it seems like someone just used Claude to make it happen quickly. I dunno much about it but just what I’m noticing.
You’re not wrong: Decompiling (turning a compressed program like an .exe back into readable code) is a task that AI excel at. That’s exactly what’s boosting development speed of all these crazy new mods and ports.
Good points! Although close. Its not decompilation, LLMs are actually bad at this due to limited context windows. Most articles I see suggest LLMs support coders recreating known outcomes using templates. Devs still do tonnes of work.
https://www.gamespot.com/articles/entire-games-are-being-ripped-off-by-ai-and-things-are-only-going-to-get-worse/
Steam argued its mostly due to single mechanics getting recreated, suggesting devs spend more time fleshing out new mechanics as LLMs are really just helping devs recreate workflows.
https://www.pcguide.com/news/valve-addresses-steams-ai-game-problem-as-devs-face-threats-well-copy-your-game-with-ai-and-release-it-first/
LLMs make solving known patterns easier. But devs still do most of the work.
Here is a recent example:
https://www.reddit.com/r/Games/comments/1xoishj/my_game_was_cloned_with_ai_what_do_i_do_game/
The cloner cloned used AI to build templates and workflows to copy mechanics. He still did a lot himself. Also because the indie game used asset packs the cloner could use the same.
LLMs can’t recreate complex assets too well either so they still need to steal models and rigs etc for big games.
True, I should have been more specific. LLMs are a great reverse engineering tool as they excel at semantic recovery and readability vs “traditional” decompilation tools. Naturally most work is by an actual dev but LLMs are a huge productivity multiplier.
You put that so well! I will re-use your good description.