Would you mind running /feedback and sharing the id here next time you see Opus do something weird? Would be very helpful for debugging
SAFETY
-
Quantum Computing Threat: Harvest-Now, Decrypt-Later Attacks
By
–
Adversaries are already collecting encrypted data today, betting they can decrypt it once sufficiently powerful quantum systems arrive. This is called "harvest-now, decrypt-later." And it's happening right now. Here's why no industry is exempt from what's coming:
-
AI as Civilization’s Last Hope for Survival
By
–
I think these look like the last decades of our present civilization. AI might give us a chance to continue
-
Humanity’s Last Gasp: AI’s Impact on Human Future
By
–
[AINews 14 Apr 2026] Humanity's Last Gasp https://
latent.space/p/ainews-human
itys-last-gasp
… -
Functionality Over Power: Group Coordination in AI Systems
By
–
This isnt about more power but being functional as group
-
CIMC Salon: M3gan 2.0 AI Alignment Movie Discussion
By
–
This Thursday 6pm in SF: it's time for another little CIMC salon; we will start with the wonderfully questionable AI alignment movie M3gan 2.0
-
Double-Agent Defender for Belief Steering via Theory of Mind
By
–
Playing Along Learning a Double-Agent Defender for Belief Steering via Theory of Mind paper: https://
huggingface.co/papers/2604.11
666
… -
Anthropic’s Automated Alignment Researchers Outperform Human Scientists
By
–
2026 already holds some surprises. And it doesn't look like that will change anytime soon. Anthropic's automated alignment researchers are already outperforming human scientists and "discovering ideas that humans would not have considered, thus broadening our exploration space
-
Neurosymbolic Implementation for Deterministic AI Behavior
By
–
This is great work but I hope we have neurosymbolic implementation under the hood to ensure deterministic behaviour where needed. @pierrepinna https://t.co/SIUjAtKeUP
— AI (@DeepLearn007) 14 avril 2026This is great work but I hope we have neurosymbolic implementation under the hood to ensure deterministic behaviour where needed. @pierrepinna