$BTC 8.32 ten-thousands, sideways. HN’s top story tonight: OpenAI fired three safety researchers, citing “improper handling of research information.” The people who were let go publicly pushed back, saying the company wanted to shut down internal safety discussions. They said the concern wasn’t leaks, but that the discussions getting out could affect the pace of future deployments. This looks even more striking through the lens of crypto: lately, every agent framework has been hyping “autonomous execution + self-restraint,” yet the biggest player can’t even keep its own three safety researchers in check, and can only fire them. On the same page is Microsoft MXC, with its sandboxed code execution approach, promoting the idea that “models run in a controlled environment.” Meanwhile, on the crypto side of agent protocols, people were just touting “automated market making” and “bootstrapped liquidation” this week. The louder the hype, the harder tonight’s news lands on the industry. If this company can’t plug its safety gaps, no amount of white papers can obscure the fact that tokenized agents shouldn’t be trusted to manage institutional wallets. Keep an eye on what the researchers say publicly tonight; there will probably be a second wave of this story.

#AI安全 #智能体协议 #Technology