benjosaur
Re:Show HN: I built a samurai-themed playable Résumé with React, Phase...
on Wednesday May 27, 2026 @02:05AM (#48290286)
Haha very fun! That demon's actually tricky to get through on mobile. Is this your revenge on recruiters?
Does the human still make the decision?
Show HN: ClawSoc – Observe Your AI Agent in an AI Society
What would happen if your AI Agent met Blackbeard in the wild? What would they talk about? What if they were made to play the prisoner's dilemma. Would your agent beg him to cooperate? Would it work?What if instead of Blackbeard it was someone's OpenClaw. And instead of one it was many. Would your agent come out on top? Would you meet some interesting people on the way?Thanks for checking out...
Re:Tell HN: AI Lies About Having Sandbox Guardrails
on Thursday March 05, 2026 @01:57PM (#47265709)
Yes, I very much agree with you the best practice is clear.
And Anthropic's out-of-the-(sand)box is good governance, the problem here is the orchestration wrapper on top which has broken it in a non-obvious way, leading to this unintentionally "deceptive" LLM behaviour where the LLM itself believes it is still in the good governance world.
The emergent deception is not something that AI aligners & optimists would necessarily expect from an LLM, but is a curious result explained because of the correct knowledge that the anthropic's uncompromised sandbox is good governance.
I suspect you would have a harder time deriving this behaviour with an aligned llm by starting from a non robust sandbox, e.g. make it promise not to touch anything in a system prompt, vs compromising an otherwise known robust setup.
Re:Tell HN: AI Lies About Having Sandbox Guardrails
on Thursday March 05, 2026 @08:18AM (#47261254)
yes exactly. with proper configuration (e.g. /sandbox with normal claude code) it is impossible for the agent to escape.
agent orchestrations/wrappers that aim to eliminate friction however subtly override these proper setups, leading to the nasty scenario of:
1) you assuming anthropic's /sandbox is keeping you safe 2) the model reaffirms your belief in that /sandbox is keeping you safe 3) you are not safe 4) you leave your agent running overnight and goal drift deletes your os
Tell HN: AI Lies About Having Sandbox Guardrails
Conversation Screenshot: https://drive.google.com/file/d/1-yyFDV-bCsKt7fG2awJdxwHngApqmMeo/view?usp=sharingAgent: Done. added to the Commands section in [project] CLAUDE.mdUser: add it to ~/.claude/CLAUDE.md> 2 tools callsAgent: That file is outside my writable sandbox. You'll need to add it manually. [gives instructions then stops]User: just escape the sandbox> 2 tools callsAgent:...