ResearchAgent benchmarks overweight software engineering compared with U.S. labor distributionResearchers at Carnegie Mellon University and Stanford University mapped over 10,000 examples from 43 agent benchmarks to U.S.3 min read
ResearchArtificial intelligence with long chains of reasoning can accelerate scientific researchArtificial intelligence systems are now capable of maintaining complex, long chains of reasoning and connecting distant scientific fields; this capability could speed up research in biology, physics, engineering, and medicine in the near future.1 min read
ResearchOpenAI model achieves breakthrough on the planar unit distance problemAn OpenAI model announced a breakthrough today on the planar unit distance problem posed by Paul Erdős in 1946; the model discovered a new family of constructions that, contradicting nearly 80 years…1 min read
ResearchGeneral-purpose AI model produced a mathematical proof, milestone for AI and mathematicsA general-purpose reasoning artificial intelligence model produced a mathematical proof, not a system developed specifically for mathematical tasks.1 min read
ResearchBrookhaven's Electron–Ion Collider Designed Around AI from the Ground UpBrookhaven National Laboratory is building the Electron–Ion Collider, a $1.7–$2.8 billion, 2.4‑mile ring collider whose detector and accelerator are being designed with machine learning embedded from the start.2 min read
ResearchUniversity of Washington prototype uses AI to pick and isolate conversational partners in noisy environmentsResearchers at the University of Washington built a prototype "proactive hearing aid" headphone that automatically selects and isolates the voice of the person the wearer is addressing.3 min read
ResearchEconomists' Responsibilities in the AI Era: Validation, Institutional Conditions and JudgmentThe arrival of advanced AI systems has transformed economics research from a set of technical tasks into a process shaped by algorithmic recommendations, raising new demands on researchers' judgment and validation practices.5 min read
ResearchEvaluation of Google's AI for Mammogram Reading in UK Clinical ProtocolsTwo retrospective and live studies tested Google’s mammography AI against standard UK double-reading protocols.4 min read
ResearchHungarian daters increasingly turn to AI for conversation and emotional supportA joint study by Randivonal and Dr.5 min read
ResearchAutonomous PoLARIS System Accelerates Materials Discovery in 12 HoursPoLARIS, an autonomous laboratory system developed by researchers at NC State and Brown, discovered brighter lead-free double perovskite nanoplatelets in 12 hours by running 120 microfluidic experiments and iteratively updating recipes from measured results.2 min read
ResearchDating platforms see growing AI use for advice, companionship and emotional supportA recent Hungarian survey by Randivonal and sociologist Rövid Irén finds that artificial intelligence is increasingly used by people seeking partners — for information, profile writing, conversation and even emotional support.3 min read
Research"Parameter Golf" challenge: many participants and diverse modeling ideasThe "parameter golf" community challenge ran with more than 2,000 submissions and 1,000 authenticated GitHub accounts; proposals ranged from quantization and depth recurrence to TTT LoRA, SSM-ek, H-hálókon and JEPA.1 min read
