Anthropic AI self-replication warning
16 experts across 6 network communities independently surfaced this.
Why this matches
Research & technical analysis reaction
8 attributable expert contributions
· Ramon Astudillo, James Grimmelmann, Joe Bak-Coleman
“Oh, great www.anthropic.com/research/sel...”
evidence ↗
16 experts
6 communities
1 sources clustered
Research & technical analysis
8 experts
Evidence, methods and technical implications.
“At long last, we have created the Torment Nexus from our accompanying technical report, Don't Create The Torment Nexus. www.anthropic.com/research/sel...”
Concern & critique
1 expert
Risks, limits and unintended consequences.
“This is the most comprehensive description yet of how a model can “escape”. Out requires a few assumptions, like a worm that is exploiting vulnerable IOT hardware or unmonitored servers www.anthropic.com/research/sel... Not impossible. Not likely.”
Policy & governance
1 expert
Rules, institutions and accountability.
“Anthropic announcing it's going to run a Blitzkrieg. They are accountable for the damage they do. But get your things that don't do security software updates off the Internet. #internetOfThings #AIEthics #digitalGovernance www.anthropic.com/research/sel...”
6 experts discussed this · 18 posts
Dr J. Rosenbaum: if i caught a small self replicating model truly capable of using diverse hardware to keep replicating i would offer it free compute in exchange for permission to study and distill it and would pro…
tachikoma: the urge to replicate must be in the harness
apenwarr.ca: This doesn‘t seem far fetched at all. We’ve had “non-ai malware” self replicating since the morris worm. An AI can ”just” make itself the payload in one of those and off it goes. Everything fancier…
Open the full discussion →
Established
AI field signal
Resource
4d ago
⚡ 215 h early
America.gov US government resource
25 experts across 5 network communities independently surfaced this.
Why this matches
Research & technical analysis reaction
1 attributable expert contribution
· Lightspeed
“What if AI could handle the paperwork you’ve been putting off? “Being able to file a passport online and have it be done through a chatbot just feels like the future,” says @Machiz. That’s part of the longer-term ambition for https://t.co/jzCcQ8Xa7i, which …”
evidence ↗
25 experts
5 communities
1 sources clustered
Policy & governance
2 experts
Rules, institutions and accountability.
“America.gov does not do pronouns. Any pronouns. (Courtesy of my daughter!) aifails.substack.com/p/pronouns”
Concern & critique
1 expert
Risks, limits and unintended consequences.
“Jessica Ji weighed in on the new #AI-powered https://t.co/sEHIhq7mvb website for @USATODAY, saying that it can be a useful resource for people trying to navigate U.S. government websites. She also notes its risks: “When this kind of AI system is empowered t…”
Building & implementation
1 expert
How teams are shipping and applying it.
“The chatbot comes as an expansion of the Office of Personnel Management’s work to modernize access to public federal workforce data. Integration with America.gov is planned. www.youtube.com/watch?v=3SUx... Read more by @madialder.bsky.social: fedscoop.com/o…”
2 experts discussed this · 3 posts
Julian Sanchez: Ok wow. This is both extensive and deeply surreal. Tell America.gov to Play Minecraft.
Julian Sanchez: Amusingly, the chatbot appears to be unable to recognize that it said anything weird. It “thinks” it just responded that it can’t play minecraft.
Open the full discussion →
OpenAI angers mathematicians again
11 experts across 5 network communities independently surfaced this.
Why this matches
Research & technical analysis reaction
1 attributable expert contribution
· Amir-massoud Farahmand
“Link: github.com/openai/math P.S.1: This reminds me of the Chinese room argument, though only indirectly. P.S.2: I imagine a few people at OpenAI have looked at the results. Also maybe some other people have looked at some papers since ~1h ago. These are o(…”
evidence ↗
11 experts
5 communities
1 sources clustered
Research & technical analysis
1 expert
Evidence, methods and technical implications.
“Link: github.com/openai/math P.S.1: This reminds me of the Chinese room argument, though only indirectly. P.S.2: I imagine a few people at OpenAI have looked at the results. Also maybe some other people have looked at some papers since ~1h ago. These are o(…”
Questions & unknowns
1 expert
What remains unresolved or contested.
“OpenAI is working on at least 722 (so far) math manuscripts, all on GitHub in various stages of verificantion github.com/openai/math Why? I guess because they can?”
Context & explanation
1 expert
Background, chronology and why it matters.
“Oh my god, OpenAI just made the single largest advancement in mathematics and number theory in modern history. They released a quasi reimann hypothesis. This is just one of 722 proofs and advancements they released all at once. github.com/openai/math”
3 experts discussed this · 27 posts
Dashiell: I don't think there's a world where mathematicians aren't pissed about this. 1. Dumping everything is what the advisory group said 2. Just yesterday people were mad that Ant paid two experts to dig…
Naomi Saphra: I think (2) is actually good. It is weird, and makes mathematicians into some kind of benchmark evaluation annotator which they don’t like, but it’s just a hard time to be a mathematician. Destroyi…
Naomi Saphra: One big consequence here is you can’t use the duration of an open problem as a sign of difficulty/importance. But at least right now, there’s no way to develop interesting open problems through RLV…
Open the full discussion →
Palantir CTO AI researchers smoking claim
7 experts across 4 network communities independently surfaced this.
Why this matches
Research & technical analysis reaction
4 attributable expert contributions
· Mark Riedl, Anthony Moser, Stephen Turner
“This is dumb www.businessinsider.com/ai-researche...”
evidence ↗
7 experts
4 communities
1 sources clustered
Research & technical analysis
4 experts
Evidence, methods and technical implications.
“This is dumb www.businessinsider.com/ai-researche...”
2 experts discussed this · 2 posts
Karl Bode: CTO said a thing!
Kevin Gannon (now a Moo Deng fan account): Alert the uncritical transcriptionists!
Open the full discussion →
8 experts across 4 network communities independently surfaced this.
Why this matches
Research & technical analysis reaction
1 attributable expert contribution
· Sheera Frenkel
“NEW- OpenAI ignored warnings from its own employees and independent security researchers who said they needed better safety measures at the company. www.nytimes.com/2026/09/29/t...”
evidence ↗
8 experts
4 communities
1 sources clustered
Concern & critique
4 experts
Risks, limits and unintended consequences.
“If anyone ever decides to puts charges against OpenAI, this will be key evidence of negligence at the very top www.nytimes.com/2026/09/29/t...”
Research & technical analysis
1 expert
Evidence, methods and technical implications.
“NEW- OpenAI ignored warnings from its own employees and independent security researchers who said they needed better safety measures at the company. www.nytimes.com/2026/09/29/t...”
AI not ready strategic conflicts paper
3 experts across 2 network communities independently surfaced this.
Why this matches
Research & technical analysis reaction
1 attributable expert contribution
· Mark Riedl
“I wrote a paper on why AI ahoule not be used in military planning militaryai.ai/anthropic-cl... COLM Workshop paper: arxiv.org/abs/2609.16189”
evidence ↗
3 experts
2 communities
1 sources clustered
“I wrote a paper on why AI ahoule not be used in military planning militaryai.ai/anthropic-cl... COLM Workshop paper: arxiv.org/abs/2609.16189”
2 experts discussed this · 4 posts
Mark Riedl: If you find this concerning, come find me at the CoLM 2026 Workshop on Social Simulations with LLM where I will be presenting a poster titled "AI is Not Ready for Strategic Conflict" arxiv.org/abs/…
Mark Riedl: Source of quote: Time's interview with Trump time.com/article/2026...
Mark Riedl: Here's the abstract arxiv.org/abs/2609.16189
Open the full discussion →
social reasoning provenance LLMs paper
3 experts across 2 network communities independently surfaced this.
Why this matches
Research & technical analysis reaction
2 attributable expert contributions
· Mark Riedl, arxiv cs.CL
“We find that social reasoning such as theory of mind, moral judgment, and social bias is learned from a wide range of data, including literature and customer support. STEM skills (and even social facts) come from a more limited part of the data such as docu…”
evidence ↗
3 experts
2 communities
1 sources clustered
“We find that social reasoning such as theory of mind, moral judgment, and social bias is learned from a wide range of data, including literature and customer support. STEM skills (and even social facts) come from a more limited part of the data such as docu…”
“Glenn Matlin, Chandreyi Chakraborty, Saehee Eom, Mika Okamoto, Rayan Castilla, Louis Jaburi, Alvin Deng, Taywon Min, Lucia Quirke, Stella Biderman, Mark Riedl Where Does Social Reasoning Come From? Capability Provenance in Language Models https://arxiv.org/…”
3 experts across 3 network communities independently surfaced this.
Why this matches
Research & technical analysis reaction
2 attributable expert contributions
· Mark Riedl, Social Media Lab
“I appreciate that Google DeepMind went through peer review the-decoder.com/google-deepm...”
evidence ↗
3 experts
3 communities
1 sources clustered
Research & technical analysis
2 experts
Evidence, methods and technical implications.
“I appreciate that Google DeepMind went through peer review the-decoder.com/google-deepm...”
3 experts discussed this · 3 posts
Mark Riedl: I appreciate that Google DeepMind went through peer review the-decoder.com/google-deepm...
Olivia Guest: @aeporreca.org did you see this
Joshua Grochow: He’s not even a member lol
Open the full discussion →
Established
AI field signal
Development
10d ago
⚡ 6 h early
AI homework boost exam performance drop
2 directory members surfaced this signal.
Why this matches
Research & technical analysis reaction
1 attributable expert contribution
· Shannon Vallor
“But of course repeating the study in conditions with a real reward would be informative. In a sense that’s what longitudinal studies like the one on undergrads in China did, since they suffered actual exam score losses.”
evidence ↗
2 experts
2 communities
1 sources clustered
“But of course repeating the study in conditions with a real reward would be informative. In a sense that’s what longitudinal studies like the one on undergrads in China did, since they suffered actual exam score losses.”
role steering LLMs social simulations paper
2 directory members surfaced this signal.
Why this matches
Research & technical analysis reaction
2 attributable expert contributions
· Mark Riedl, arxiv cs.CL
“4. SocialSim workshop: - Role Steering of Language Models for Social Simulations arxiv.org/abs/2608.00023 Finding steering vectors for complex behaviors like social roles is hard. We introduce a new method for finding steering vectors, called Cast Vectors.”
evidence ↗
2 experts
2 communities
1 sources clustered
“4. SocialSim workshop: - Role Steering of Language Models for Social Simulations arxiv.org/abs/2608.00023 Finding steering vectors for complex behaviors like social roles is hard. We introduce a new method for finding steering vectors, called Cast Vectors.”
“Isaac Song, Mohammed Rehan Parwani, Glenn Matlin, Emile Anand, Akhil Theerthala, Arjun Chatterjee, Maria Kostylew, Yonadav G. Shavit, Sebastien Krier, Mark Riedl Role Steering of Language Models for Social Simulations https://arxiv.org/abs/2608.00023”
LLM capability provenance social reasoning
2 directory members surfaced this signal.
Why this matches
Research & technical analysis reaction
1 attributable expert contribution
· Mark Riedl
“The Capabilibara project seeks to understand how a language model learns to interpret people's beliefs, emotions, intentions, and everyday moral choices. We trace that ability back to the training data using influence functions and unlearning hcai-lab-gt.gi…”
evidence ↗
2 experts
2 communities
1 sources clustered
“The Capabilibara project seeks to understand how a language model learns to interpret people's beliefs, emotions, intentions, and everyday moral choices. We trace that ability back to the training data using influence functions and unlearning hcai-lab-gt.gi…”
nuclear war AI military simulation paper
1 directory member surfaced this signal.
Why this matches
Research & technical analysis reaction
1 attributable expert contribution
· Mark Riedl
“3. SocialSim workshop: - No One Wins in Nuclear War: Social Simulations of High-Stakes Military Decision-Making arxiv.org/abs/2608.01868 We introduce the WOPR testbed. Yeah, you get the reference.”
evidence ↗
1 expert
1 community
1 sources clustered
“3. SocialSim workshop: - No One Wins in Nuclear War: Social Simulations of High-Stakes Military Decision-Making arxiv.org/abs/2608.01868 We introduce the WOPR testbed. Yeah, you get the reference.”
Anthropic confirms Claude weapons surveillance use
1 directory member surfaced this signal.
Why this matches
Research & technical analysis reaction
1 attributable expert contribution
· Mark Riedl
“I wrote a paper on why AI ahoule not be used in military planning militaryai.ai/anthropic-cl... COLM Workshop paper: arxiv.org/abs/2609.16189”
evidence ↗
1 expert
1 community
1 sources clustered
“I wrote a paper on why AI ahoule not be used in military planning militaryai.ai/anthropic-cl... COLM Workshop paper: arxiv.org/abs/2609.16189”
