Sources: OpenAI's Daybreak pages and Greg Brockman's post (23:16 Europe/Rome, 12 Nov 2026). Early access. Engineers review patches before merge.
https://t.co/OIPWewRwf4
https://t.co/GfoCGZcqcP
https://t.co/elbhyP9WNd
#AISecurity
Announcing two new cyber products:
- Daybreak Defense Factory (https://t.co/OdlGbgPWp8)
- Daybreak Red Team (https://t.co/VQTAjDfduz)
Both make it much easier to put frontier cyber capabilities to work, helping defend companies in new ways.
OpenAI just announced Daybreak Red Team and Defense Factory: agents that test web apps and draft patches.
Engineers still review before a merge. Isolated environments and Guardian review limit what they can reach.
Would you let agents draft the patch if a human still merges?
Introducing Claude Meetings: add Claude to your team’s video call and it builds docs, whiteboards, and other artifacts with you while you’re still talking.
Starting to roll out in beta on Team and Enterprise plans.
Sources: Anthropic's articles on self-hosted Managed Agents (12 Nov 2026) and Claude Meetings, plus the Meetings Help Center for plan defaults.
https://t.co/QxXu4w3L0V
https://t.co/oyFuAfehTc
https://t.co/x4yWiy4h68
#AISecurity
I went to sleep on self-hosted Claude agents. This morning I woke up reading Claude Meetings.
On Team plans it can join Google Meet by default, including calls with outside guests, unless an owner turns that off. Transcripts go to https://t.co/qCr6Y81rUB.
Would you leave that default on?
An agent with confirmations off ran pkill -f "cat" -n on @deedydas's Mac. He says every app closed.
-f matches the whole command line. "cat" is in "Applications".
My rule: skip-permissions only in a throwaway sandbox.
What would you never let it run?
https://t.co/kTuX7IeJHc
Had my first crazy dangerously-skip-permissions moment.
It ran `pkill -f "cat" -n` to kill a previous stray `cat` process, but it matched to everything in `Appli[cat]ions` and killed every single running app on my Mac.
Had my first crazy dangerously-skip-permissions moment.
It ran `pkill -f "cat" -n` to kill a previous stray `cat` process, but it matched to everything in `Appli[cat]ions` and killed every single running app on my Mac.
Days apart, two coding-agent stories.
Karpathy: Claude turned 43M HN comments into a 6,454-video recommender in ~2h.
Deedy Das: an agent with permissions skipped ran pkill -f "cat" -n. He says every app on his Mac closed.
Your setup: confirm each command, allowlist or skip?
This is f*cking insane. The engineer who built Claude Code just dropped a 28-minute video on how to write prompts that actually work.
I've seen $600 courses that don't cover what he shows in the first 5 minutes.
Watch it, then read the guide below on the Claude features 99% of users never find.
Six projects from the last few weeks that caught our attention:
A wildfire analysis prototype that maps the fire’s spread over time and tracks hot spots, made with Opus 5.5.