ResearchChatGPT improves polish while causal-reasoning training boosts originality in student workA randomized experiment at Bocconi University with over 1,000 first‑year undergraduates, run in collaboration with OpenAI Economic Research, tested the separate and combined effects of access to ChatGPT (GPT‑4o) and a causal‑reasoning exercise.3 min read
ResearchGlucoFM: a dual‑stream foundation model that separates slow and fast CGM dynamicsGlucoFM is a self‑supervised foundation model for continuous glucose monitor (CGM) data that explicitly separates slower baseline glycemic trends from short‑term deviations.7 min read
ResearchOngoing HIP Lab and METR research: Claude user experience and coder agents' productivityTwo ongoing studies by HIP Lab and METR are examining: HIP Lab is studying how Claude’s behavior affects how people feel while using AI, while METR is estimating the real productivity gains from…1 min read
ResearchOrganization invites researchers to extend its research modelThe organization announced it wishes to scale its research model and is offering tool access to researchers for work they currently cannot do by other means.1 min read
ResearchAnthropic first to open privatized Claude usage data to external researchersAnthropic announced that, for the first time, it has made real, privacy-anonymized Claude usage data available to external researchers, data that previously could only be examined within AI labs.1 min read
ResearchAnima Anandkumar’s neural-operator approach opens new paths for modeling weather and physical systemsCaltech professor Anima Anandkumar developed Neural Operators and the FourCastNet family to model large, chaotic physical systems like weather using AI, achieving competitive short-term forecasts with consumer GPUs.5 min read
ResearchMIT's η-learning Generates Plausible Extreme Events from Ordinary DataOn August 20, researchers at the Massachusetts Institute of Technology published a paper in Nature Communications describing η-learning, an algorithm that produces realistic maps of extreme events without being trained on extreme-event data.2 min read
ResearchWhere current language models still fall short of human puzzle-solvingPuzzles reveal specific gaps between human and machine reasoning even as large language and vision models rapidly improve.5 min read
ResearchAgentHands: Co‑speech Hand Gestures Grounding AI Agents in Android XRAgentHands is a research prototype presented at CHI 2026 that augments conversational AI agents in Android XR with synchronized, expressive hand gestures.6 min read
ResearchAnthropic publishes an Economic Index mapping real-world AI use from Claude.ai conversationsAnthropic on Feb 10, 2025 launched the Anthropic Economic Index and published an initial paper analysing roughly one million anonymized conversations on Claude.ai to measure how AI is used across occupational tasks.4 min read
ResearchOpenAI unveils Jalapeño inference chip with higher efficiency and lower latencyOpenAI reports that its first custom inference chip, Jalapeño, achieves higher throughput per watt and lower end-to-end latency than leading commercial accelerators on public benchmarks.4 min read
ResearchDeepMind and partners shave further digits off the matrix multiplication exponent using AlphaEvolveResearchers from Google DeepMind, Carnegie Mellon University, Columbia University and MIT have slightly lowered the known upper bound on the matrix multiplication exponent omega by combining modern optimization techniques with AlphaEvolve, an LLM-based code-evolution system.2 min read
