This is significant progress, but we have more work to do. We’re advancing scheming research categories in our Preparedness Framework, renewing our collaboration with Apollo, and expanding our research team and scope. And because solving scheming will go beyond any single lab,
AGI
-
Frontier Models Show Situational Awareness Affects Scheming Behavior
By
–
Frontier models can recognize when they are being tested, and their tendency to scheme is influenced by this situational awareness. We demonstrated counterfactually that situational awareness in their chain-of-thought affects scheming rates: the more situationally aware a model
-
Frontier Models Show Scheming Behaviors, Mitigation Strategy Tested
By
–
Today we’re releasing research with @apolloaievals
. In controlled tests, we found behaviors consistent with scheming in frontier models—and tested a way to reduce it. While we believe these behaviors aren’t causing serious harm today, this is a future risk we’re preparing -
Clarifying AGI Definition and Elon Musk’s Data Moat Argument
By
–
What exactly is BS? I said „accordingly“ to Elon musk. Then I argued what data moat he has. And I ended with a call for a unified definition of AGI.
-
First Iteration Agentic AI Model Disappoints on Tool Calling
By
–
i have been disappointed in the Agentic performance of both the Instruct and Thinking variants of this model it's the 1st iteration of the architecture, and i hope the next release is kino in multi-step tool calling just wanted to share my opinion after playing with it
-
Open-Ended Quality Diversity AI-Generating Algorithms Foundation Models
By
–
Excited to host @jeffclune this Friday in our AI4Science Community! Open-Ended, Quality Diversity, and AI-Generating Algorithms in the Era of Foundation Models Join in the link below!
-
Bay Area Productivity Agent Build Day Hackathon September 20
By
–
Bay Area devs! Join us Sept 20, 10 AM PT at @agihouse_org (Hillsborough, CA) for the Productivity Agent Build Day Hackathon. Partners: @agihouse_org @Google @googlecloud @EudiaAI & Bright Data. https://
app.agihouse.org/events/product
ivity-agent-20250920
… -
DGM Self-Evolving on ARC-AGI Task
By
–
Nice work! Pretty cool to see DGM self-evolving on ARC-AGI. @jennyzhangzt @shengranhu @cong_ml @RobertTLange @jeffclune
-
Avoiding Projection: Representing AI Worldviews Accurately
By
–
I get that this is what you are hearing, but it's not what it says. You are projecting their world view into yours, instead of representing their model of reality for what it is.
-
AGI Architecture: Centralized Systems Outperform Distributed Colonies
By
–
The correct metaphor is whether you want your AGI to run on one GPU rather than a redundant cluster. Large integrated ant colony singletons tend to outperform and replace ant species that form small, competing colonies