Hacker & Security News

hacker.at.thenote.app

12 October 2026 21:51

The Most Dangerous Memory in an AI Agent Is the One That Used to Be True. AI agents can remember facts that are no longer valid. Learn how memory validity, temporal context, provenance, and state versioning make agent memory safer. #hackernews #news

bmaltaverne.bsky.social

12 October 2026 21:50

Org knowledge now has 2 readers: people & #agents. People need context & pointers to experts. Agents need explicit definitions, constraints, owners, & dates. For #procurement, capturing everything matters less than keeping the right knowledge current. procurementtidbits.substack.com/p/groundhog-...

rain 🌦️

sunshowers.io

11 October 2026 14:46

Ah so OpenAI has withdrawn multiple math papers, some for containing critical errors that invalidate the entire argument. Basically mathslop without a human in the loop who is accountable for what they publish. Sad

Rhoman¢er🧌

angussoftware.dev

11 October 2026 14:44

It's kinda nutty to think a year ago most of us were either AI skeptics or still using AI coding auto-complete and copy pasta from ChatGPT dot com rather than agentic.

Jennifer Ouellette

jenlucpiquant.bsky.social

09 October 2026 16:02

Rogue AI or human error? The real story behind the OpenAI-Hugging Face incident www.scientificamerican.com/podcast/epis...