A paper that fails to change how people think fails for science. New in @science.org with Casey Petroff and Gary King: why surprise is the currency of scientific impact, and how to bring framing and audience choice inside the formal theory of inference. doi.org/10.1126/scie...

Out today in @science.org: with the amazing Haochuan Cui, Yiling Lin, & @lingfeiwu.bsky.social, we analyzed 3.6 million scientists publishing 1960–2020. The findings reshape a century-old debate about age and scientific creativity.

New paper with the amazing Nadav Kunievsky, just accepted to ICML 2026 (in Seoul, Korea), uniquely measures the extent to which an LLM tracks user intent or merely superficial phrasing. The paper: arxiv.org/abs/2506.16584

New paper led by @junsolk.bsky.social w/ Shiyang Lai, Nino Scherrer & @blaiseaguera.bsky.social: What happens inside reasoning models like OpenAI's o-series, DeepSeek-R1, QwQ when they think? They don't just compute longer. They spontaneously generate internal debates. 🧵

Excited to share new research with the inimitable Di Tong and Lingfei Wu, just out in Nature Communications. 🧵

Our paper is out today in Nature! 🎉 Collaborated with the amazing Qianyue Hao, Fengli Xu, and Li Yong from Tsinghua & Zhongguancun Academy to analyze how AI is reshaping science. 41.3 million papers. Four decades. One paradox.

New research with the brilliant Likun Cao, out just today in Technological Forecasting & Social Change: For decades, innovation scholars have debated whether deep technological search (intensive work within a domain) or broad search (combining distant knowledge) drives greater impact.

Awesome discussion/podcast with the BigBrains team at UChicago about the future of AI (over two sessions!)-- news.uchicago.edu/big-brains-p...

Are we making AI too human?, with James Evans

Data scientist explains how we are limiting models like ChatGPT in training them—and how that’s narrowing research

news.uchicago.edu

Kudos to Google. Gemini 2.5 Pro tops the latest leaderboard. www.thealgorithmicbridge.com/p/google-is-...

Google Is Winning on Every AI Front

Neither OpenAI nor Anthropic have a chance at this point

thealgorithmicbridge.com

Why do scientists disagree? With the amazing @justinsulik.bsky.social, NakWon Rim, Elizabeth Pontikes, and @glupyan.bsky.social , we show in @naturehumbehav.bsky.social how cognitive traits and dispositions align with persistent scientific divides (www.nature.com/articles/s41...).

New in @ScienceMagazine: "Large AI models are cultural and social technologies". Working with the brilliant @himself.bsky.social, @alisongopnik.bsky.social, and Cosma Shalizi, we offer an alternative to the prevailing narrative about AI models as autonomous agents. www.science.org/stoken/autho...

Large AI models are cultural and social technologies

Implications draw on the history of transformative information systems from the past

science.org

Come to the 4th International Conference on the Science of Science and Innovation in Copenhagen, June 16-18 ('25). ICSSI explores the dynamics of scientific research and innovation, featuring science and technology policy environments, opportunities, and challenges in Europe and around the world.