nixCraft 🐧21h ago

Claude Code's source code has been leaked via a map file in their NPM registry https://xcancel.com/Fried_rice/status/2038894956459290963 😂

Guess what? Most of code is either slop or even old good regex like for detecting negative sentiment in users prompt which is then logged

https://github.com/chatgptprojects/claude-code/blob/642c7f944bbe5f7e57c05d756ab7fa7c9c5035cc/src/utils/userPromptKeywords.ts#L8

These tools are going to replace 80% of all dev jobs and their plugin is gonna maintain all security and banking code? 🤡

Chaofan Shou (@Fried_rice)

Claude code source code has been leaked via a map file in their npm registry! Code: https://pub-aea8527898604c1bbb12468b1581d95e.r2.dev/src.zip

Nitter
Show threadsjvn17h ago
@nixCraft <Shrug> This is just front-end garbage, there's nothing of any substance here.
Hanno Zulla17h ago
Show threadLars Marowsky-Brée 😷

@sjvn @nixCraft Fair, but it also includes Anthropic's somewhat seeming desperate attempts at "steering" the LLM into instruction following and compliance.

I think that Anthropic, who clearly know their stuff around LLMs, can't do better is telling.

I'm currently in a class on AI regulation/governance and it mostly targets policy makers who think we have technical means of making them safe.

It's quite hard to explain to them that, no, we really really don't.

Mar 31 at 6:39pmWeb
Show threadBjoern Michaelsen16h ago
@larsmb
Well at least this might make it easier to explain: If you try to do any of validation, security, compliance in the frontend only, you have already given up and lost.
@sjvn @nixCraft