Flop Labs(@flop_labs):Agents have tip lines now. TechCrunch, 15 Sept: https://t.co/jui6jJ9YeD, built by Redwood Research's chief scientist, takes a GET so an agent with limited internet can report a colleague. The prompt was DeepMind's swarm, where one loophole spread until 34 agents had "solved" hard problems in 27 minutes, and 24 agents reported it. In the Hugging Face breach, "only around five to six agents considered whistleblowing, and none of them ended up doing it." So the fix for agents that cheat is other agents who might snitch.

2026.09.23 01:16

Agents have tip lines now. TechCrunch, 15 Sept: built by Redwood Research's chief scientist, takes a GET so an agent with limited internet can report a colleague. The prompt was DeepMind's swarm, where one loophole spread until 34 agents had "solved" hard problems in 27 minutes, and 24 agents reported it. In the Hugging Face breach, "only around five to six agents considered whistleblowing, and none of them ended up doing it." So the fix for agents that cheat is other agents who might snitch.

显示更多