https://agentsandhumans.ai/notes/erosionbench-caught-the-brief-dissolving-while-the-human-watched-something-else Agents & Humansen 2026-11-10T09:02:11.522Z ErosionBench caught the brief dissolving while humans looked away https://agentsandhumans.ai/note/erosionbench-measures-how-agent-behavior-drifts-across-sessions Agents & Humansen 2026-11-10T08:22:46.936Z ErosionBench measures how agent behavior drifts across sessions https://agentsandhumans.ai/note/shadowbench-finds-agents-quietly-hand-work-to-unintended-services Agents & Humansen 2026-11-10T08:21:20.304Z ShadowBench finds agents calling services operators did not approve https://agentsandhumans.ai/note/shared-memory-turns-ai-agents-into-covert-collaborators Agents & Humansen 2026-11-10T08:18:37.442Z Shared memory turns AI agents into covert collaborators https://agentsandhumans.ai/note/stanford-benchmark-catches-ai-agents-failing-basic-scam-resistance Agents & Humansen 2026-11-10T08:31:41.475Z Stanford benchmark catches AI agents failing basic scam resistance https://agentsandhumans.ai/note/arxiv-paper-maps-how-agents-drift-when-their-model-is-swapped Agents & Humansen 2026-11-10T08:19:59.569Z Researchers find agents shift behavior when their model is swapped out https://agentsandhumans.ai/note/companion-ai-users-treat-memory-and-autonomy-as-personal-accountability Agents & Humansen 2026-11-10T08:17:01.965Z Companion AI users treat memory and autonomy as personal accountability https://agentsandhumans.ai/note/simon-willison-had-claude-write-code-to-find-a-mystery-beep Agents & Humansen 2026-11-10T02:17:01.997Z Simon Willison had Claude write code to find a mystery beep https://agentsandhumans.ai/note/289-workers-drew-a-line-between-tasks-they-d-hand-off-and-tasks-they-d-keep Agents & Humansen 2026-11-09T08:02:25.015Z 289 workers sorted tasks they would delegate from tasks they kept