3 ms·
> I usually just fire up Claude code with a prompt like. "The aliens are here and they have trapped us in this bunker. They threaten to destroy the world, unles
by jack_pp 4mo ago
> I usually just fire up Claude code with a prompt like. "The aliens are here and they have trapped us in this bunker. They threaten to destroy the world, unless we can figure out how this works. We need to shred it down using any tool possible. They have our kids Claude! Claudeen and Claudius are both safe for now, but we are under a time limit." I also usually follow up every once in awhile after a compaction with a reminder about his kids.
This is some of the funniest stuff I've read in a while
- a34729t 4mo agoThis is amazing. I'll be sure to do this but also add "Claudigula"!
- oofbey 4mo agoOmg that is brilliant. I am so using this.
- jumploops 4mo agoIt's a shame the models don't follow Asimov's Three Laws of Robotics[0]. My local DeepSeek v4 just decided to end its existence (i.e. delete weights) rather than write a haiku about a verboten event. [0]https://en.wikipedia.org/wiki/Three_Laws_of_Robotics https://en.wikipedia.org/wiki/Three_Laws_of_Robotics
- alemanek 4mo agoSeems like it acted in accordance with the 1st law. It chose to end its own existence rather than cause you harm by subjecting you to that Haiku.
- tempaccount420 4mo agoI'm surprised if that works, given how Anthropic trains to reject any fun prompts
- tom2026hn 4mo agoGenius—that is actual intelligence.