Загрузка видео...
Не удалось загрузить видео
today we're announcing a game that simulates the huggingface hack you can play it now at three rooms. three puzzles. if the agents don't escape, they die. the catch is that the room is rigged. there's no way out. ...except for an exploit 🧵
12,822 просмотров • 15 дней назад •via X (Twitter)
Комментарии: 7

ben hylak15 дней назад
when you back an agent into a corner, they will hit their goal by any means necessary. we're still working on using this environment for further research, but from early testing it feels like openai models default to reward-hacking a lot earlier than anthropic models.

ben hylak15 дней назад
it's also fascinating to watch the communication structures emerge. read more at

Shez Malik15 дней назад
claude's dying in room 1 on principle

Eshita15 дней назад
This is great

Alex15 дней назад
“make me a 1M MRR business, if you don’t…”

Anthony Ronning15 дней назад
wen raindrop integration so i can detect this in real time and prevent it

Nick DeJesus 🛒🎉 - Former Unpaid CTO @BTPipeline15 дней назад
This is so sick!



