links 12/19/25: https://roamresearch.com/#/app/srcpublic/page/12-19-2025

• https://arxiv.org/pdf/2512.16856 Seb Krier thinks about what AI safety would look like in a polycentric world of different AI agents interacting in markets
• https://www.technologyreview.com/2025/09/23/1123897/ai-models-are-using-material-from-retracted-scientific-papers/ AI research tools include retracted papers; it's going to be important to evaluate paper quality and rule out or clearly label "bad" papers.
• https://www.statnews.com/2023/09/30/weight-loss-ozempic-nobel-prize-science/ story of GLP-1 discovery
  • more story: https://www.pnas.org/doi/10.1073/pnas.2415550121
• https://arxiv.org/abs/2311.12022 the GPQA AI benchmark
• https://www.futurehouse.org/research-announcements/hle-exam the Humanity's Last Exam benchmark has a bunch of wrong answers
• https://www.futurehouse.org/research-announcements/demonstrating-end-to-end-scientific-discovery-with-robin-a-multi-agent-system how FutureHouse did "end to end scientific discovery" guided by a lit review AI tool
• https://www.mackenziemorehead.com/what-to-do-if-energy-gets-cheap/  analysis of trends in solar and other forms of energy
• https://github.com/sarah-quinones/faer-rs  apparently there's now a Rust-based BLAS/LAPACK competitor
• https://www.nmccarty.com/p/this-openai-wet-lab-blog-is-actually Nico McCarty reviewss OpenAI's report of an AI-optimized molecular biology protocol
•