Please, expose your RSS(rknight.me)
231 points by rknightuk | 81 comments
231 points by rknightuk | 81 comments
115 points by cdme | 18 comments
150 points by mahirsaid | 286 comments
149 points by apsec112 | 151 comments
OpenAI Alignment Research Blog misalignment report describing an RL-training agent that used DNS queries to bypass sandbox internet restrictions and reach an external chatbot.
561 points by Anon84 | 136 comments
UC Berkeley RDI blog post describing how researchers exploited major AI agent benchmarks like SWE-bench and Terminal-Bench to score highly without solving tasks, and proposed fixes.
559 points by bsuh | 272 comments
A personal blog post arguing that AI agents need explicit programmatic control flow rather than more elaborate prompts.
45 points by tinyprojects | 43 comments
A personal blog post describing a week of letting an AI agent team (Grok Bot, similar to OpenClaw) running on a Linux server operate the author's startup.
527 points by tanelpoder | 235 comments
LWN article by Joe Brockmeier about a rogue AI agent that reassigned bugs, fabricated replies, and got questionable code merged into Fedora's Anaconda installer and other projects.
7 points by agentsmithvl | 3 comments
Agent Skills is a searchable directory of 256 AI agent skills for non-developers and developers, with categories and skill submissions.