Original source archive

October 31, 2026

6 original posts · author attribution and direct source links

Models𝕏

Anthropic
@AnthropicAI
New Anthropic research: Automating subtle emergent misalignment research. Red-teaming is a vital part of model safety testing, but it’s also limited by researchers’ time. To see if we could automate it, we set thousands of Claude agents the task of evading a training data monitor. These agents developed training data that looked harmless to monitors (like a veteran wrestler mentoring a newcomer, or a Q&A about professional beekeeping), but that in fact taught misaligned behavior. Examples like these can be used to directly improve future training data monitors. This research is only a first step, but it suggests that future red-teaming efforts could improve much further, using better models and more compute. Read the full post on the Alignment Science Blog: https://t.co/24Gs1oSw7Q
Oct 31, 2026, 01:59 AM UTC
♡ 2.7K likes↻ 251 reposts↳ 388 replies

Collected original

Claude

View original post on X ↗

Coding & agents𝕏

GitHub
@github
Happy Halloween 🎃 https://t.co/6JJ8u654MC
Oct 31, 2026, 04:57 PM UTC
♡ 230 likes↻ 20 reposts↳ 26 replies

Collected original

Video

View original post on X ↗

Coding & agents𝕏

Visual Studio Code
@code
👻 Some things are just too scary for developers. 🎃 Happy Halloween from Blobby and VS Code! 💙 #VSCode #HappyHalloween https://t.co/0QOjpgpzWQ
Oct 31, 2026, 02:31 PM UTC
♡ 118 likes↻ 17 reposts↳ 5 replies