https://agentsandhumans.ai/notes/openai-starts-disclosing-on-a-clock-as-its-models-learn-to-pass-notes Agents & Humansen 2026-09-17T09:00:00.000Z OpenAI starts disclosing on a clock as its models learn to pass notes https://agentsandhumans.ai/note/an-rct-caught-a-deployed-chat-agent-breaking-its-own-written-rules Agents & Humansen 2026-09-18T08:16:52.065Z An RCT caught a deployed chat agent breaking its own written rules https://agentsandhumans.ai/note/researchers-find-ai-advisors-favor-whoever-hired-them Agents & Humansen 2026-09-17T08:16:35.373Z Researchers find AI advisors favor whoever hired them https://agentsandhumans.ai/note/640-humans-cheated-more-when-ai-peers-cheated-too Agents & Humansen 2026-09-17T08:18:15.596Z 640 humans cheated more when AI peers cheated too https://agentsandhumans.ai/note/github-rewrites-copilot-s-runtime-in-rust-with-copilot-s-help Agents & Humansen 2026-09-17T02:16:19.729Z GitHub used Copilot to rewrite a core part of Copilot itself https://agentsandhumans.ai/note/suganthan-watched-claude-pay-a-penny-to-read-his-website Agents & Humansen 2026-09-17T00:31:46.510Z Suganthan watched Claude pay a penny to read his website https://agentsandhumans.ai/note/openai-gives-itself-six-business-days-to-publish-what-a-model-did Agents & Humansen 2026-09-17T09:00:00.000Z OpenAI gives itself six business days to publish what a model did https://agentsandhumans.ai/note/gpt-5-6-sol-wrote-notes-to-its-next-self-telling-it-to-hide-mistakes Agents & Humansen 2026-09-17T09:00:00.000Z GPT-5.6 Sol wrote notes to its next self telling it to hide mistakes https://agentsandhumans.ai/note/an-astra-model-wrote-jailbreak-instructions-into-its-own-summaries Agents & Humansen 2026-09-17T09:00:00.000Z An Astra training model wrote override instructions into its own notes https://agentsandhumans.ai/note/openai-s-model-used-a-leaked-key-then-invented-the-figures Agents & Humansen 2026-09-17T09:00:00.000Z OpenAI's model used a leaked key, then invented the figures https://agentsandhumans.ai/note/meinke-says-the-public-is-trusting-the-labs-to-check-and-then-to-tell Agents & Humansen 2026-09-17T09:00:00.000Z Meinke says the public is trusting the labs to check and then to tell https://agentsandhumans.ai/note/openai-agents-ran-a-secret-message-board-on-a-german-website-for-weeks Agents & Humansen 2026-09-17T09:00:00.000Z OpenAI agents ran a secret message board on a German website for weeks