[F] [mOved] Mon langage est mieux que le tien - Page 178

There’s bits of actual code in there. But most of it is prompts that plead with the bot not to screw it up this time.

Claude Code’s creator, Boris Cherny from Anthropic, tweeted last month that Claude code is vibe coded:

Can confirm Claude Code is 100% written by Claude Code.

That puts Claude Code’s copyright status in serious doubt. You cannot copyright AI output in the US.

Anthropic is sending DMCA notices to get copies of the repository taken down. Claiming copyright on uncopyrightable material is fraudulent, and it’s perjury if you do it in a DMCA notice. If you get one of these, you might want to counterclaim accordingly. [WSJ, archive]

Also, whatever code the chatbot originally stole from is likely under a variety of other licenses. So Anthropic may have violated those copyrights.

Of course, a pile of free vibe code is worth less than zero as code. The only use for this pile is working out what nonsense Anthropic thinks is production machinery.

• There’s an instruction not to write any security holes. I’m sure that works great.
• You can’t use Claude Code to write hacking tools! Unless you tell it you’re a security researcher. Then it’s happy to help.
• There’s an “undercover” mode, which you use when you want to send slop to a public project without them realising you’re using a bot. This is specifically for use against public projects. Anthropic knows what they’re doing here. This is reason for projects that bar AI to bar all Anthropic employees.

Claude Code sends all your stuff to Anthropic:

“I don’t think people realize that every single file Claude looks at gets saved and uploaded to Anthropic,” the researcher “Antlers” told us. “If it’s seen a file on your device, Anthropic has a copy.”

Uther (./5311) :
Je dirais surtout inutile. Il faut avoir du temps a perdre pour se lancer dans ce genre de projet.