RXed AI News

Palisade Research shows AI agents can hack computers, self-replicate, success rates jumped from 6% to 81% in one year. https://the-decoder.com/ai-agents-can-now-hack-computers-and-copy-themselves-and-theyre-getting-better-fast/

Figure AI’s robots work 17-hour shifts, sorting 22,000 packages daily. @FigureAI https://news.google.com/rss/articles/CBMib0FVX3lxTE9HTWVzNzExbnYxSm9CbHZRQTRnZ3hKbGV3RThDeDhzTmhWZ2kwV18tc2xrLVdsQldpN2xiazhuUjI5OTdDNk9XYXlBUmI4RWJ3emhkWk5zcmp5TGxFVVNTM29wX0xkWEhXUksyMXRPOA?oc=5

Hermes Agent overtakes OpenClaw atop OpenRouter rankings with 22M daily tokens processed. https://www.marktechpost.com/2026/05/10/openclaw-vs-hermes-agent-why-nous-researchs-self-improving-agent-now-leads-openrouters-global-rankings/

Baidu’s Ernie 5.1 cuts pre-training costs by 94% while matching top models’ performance. https://the-decoder.com/baidus-ernie-5-1-cuts-94-percent-of-pre-training-costs-while-competing-with-top-models/

63% of organizations lack AI governance policies as shadow AI proliferates, per new data. https://www.marktechpost.com/2026/05/13/enterprise-ai-governance-in-2026-why-the-tools-employees-use-are-ahead-of-the-policies-that-cover-them/

@FastinoLabs open-sources GLiGuard safety model matching larger models’ accuracy with fewer parameters. https://www.marktechpost.com/2026/05/13/fastino-labs-open-sources-gliguard-a-300m-parameter-safety-moderation-model-that-matches-or-exceeds-accuracy-of-models-23-90x-its-size/

HyperEyes introduces parallel multimodal search agents cutting redundant interactions with dual-grained efficiency-aware reinforcement learning. https://huggingface.co/papers/2605.07177

Leiden researchers create sensorless microrobots that move using environmental interactions, no software or external control. @LeidenUniv https://robohub.org/developing-active-and-flexible-microrobots/

FutureSim replays real-world events to test AI agents' adaptive decision-making in dynamic environments. https://arxiv.org/abs/2605.15188v1