For medical information, general AI frontier models (Google, OpenAI, Anthropic) outperformed specialized @EvidenceOpen and @UpToDate as assessed by 12 US clinicians, randomized and blinded to which model and extensive testing/benchmarks. This was not anticipated. @NatureMedicine nature.com/articles/s41591-0…

There are few in this world who were such forces for good as @atulbutte. His dazzling intellect, relentless curiosity, generosity of spirit, boundless charisma, and infinite optimism left a mark on everything he touched and everyone lucky enough to know him. An enormous loss.

yay !!!!!

Do you or someone you know have #IBD and are interested in getting #pregnant soon? Consider participating in WIsDoM, an online-only registry that will help answer how IBD & IBD treatments (including surgery) affect #fertility. lightscameracrohns.com/2025/… 🙏@nataliesparacio

A small randomized trial of generative #AI for diagnosis again (as seen in a few previous studies) shows higher performance for #AI than physicians AI. May indicate that physicians need to be trained on how to incorporate AI. jamanetwork.com/journals/jam…

We’re excited to share that @CMSGov has granted ADLT status for our xT CDx test. xT CDx is the first FDA-approved #NGS assay to perform matched normal sequence analysis to provide tumor mutation profiling for patients w/ solid organ neoplasms. Learn more: tempus.co/4eMcFkM

@UCSF_DOCIT and @UCSFHospitals will develop the first continuous, autonomous monitoring platform for #AI tech in healthcare. Made possible by a $5M donation, this project aims to prevent undetected adverse outcomes for patients. #healthtech #patientcare tiny.ucsf.edu/R1PfmP

Exercise may be the single most potent medical intervention ever known. Its benefits in prevention outstrip any known drugs: 50% reduction in the risk of cardiovascular disease, 50% reduction in the risk of many cancers, positive effects on mental health, pulmonary health, GI health, bone health, muscle function. You name it. Exercise helps. In fact, the ability to exercise over long distances was likely key to our evolution as a species because the availability of densely caloric foods due to persistence hunting allowed our energy-avid brains to enlarge. And yet, we have had very little insight into the molecular basis of these magical effects...until now! Published in yesterday's Nature and featured on the cover was work from our consortium that represents the culmination of a couple of decades of pitching ideas to the NIH, forming a consortium, planning experiments, executing those experiments, and analyzing data at unprecedented scale, all aimed at enhancing our understanding of the molecular transducers of exercise. It was a major effort from so many in our consortium (playfully named MoTrPAC) and is the first landmark paper of many more to come. This first paper focused on the multi-tissue, multi-omics of treadmill exercise in rats. Specifically, we report the effects of eight weeks of treadmill running on the transcriptome, the epigenome, the proteome, the metabolome, the lipidome and the immunome of a broad range of tissues (in fact, 9,466 assays across 19 tissues, 25 molecular platforms, and 4 training time points). The result is the most comprehensive molecular map of exercise ever created. At Stanford, my colleague @MWheelerMD and I co-lead the bioinformatics center and it was our team's duty and privilege to ingest the data, QC the data, help analyze the data, and make the data available to the world. Various tools available at our data hub allow you to explore the data, visualize it, and download it for your own use. Have fun! And stay tuned for human data that will be coming. So many people to thank who made this possible (see the paper for details). Special shout out to the primary analysts and authors: David Amar, Nicole Gay, & Pierre Jean Baltran. Paper: nature.com/articles/s41586-0… Data hub: motrpac-data.org/

The first randomized trial of medical #AI to show it saves lives ECG-AI alert in 16,000 hospitalized patients 31% reduction of mortality (absolute 7 per 100 patients) in pre-specified high-risk group nature.com/articles/s41591-0… @NatureMedicine

For the first time in history, the NCAA women's basketball title game drew more viewers than the men's.

IBM & Meta are launching the AI Alliance to advance *open* & reliable AI. The list of over 50 founding members from industry, government, and academia include AMD, Anyscale, CERN, Hugging Face, the Linux Foundation, NASA.... ai.meta.com/blog/ai-alliance…

Happy Thanksgiving! Thank you @karpathy

Currently 80 job openings across the @UofCalifornia @QB3 startup ecosystem! Academia→ discovery→ jobs… let’s keep Bay Area entrepreneurship alive! Start here: jobs.qb3.org.

🌟🌟🌟Hiring a postdoc candidate, research scientist or research assistant🌟🌟🌟 (retweets appreciated) Exciting times! @KMagudia and @ecalabr are recruiting a candidate to join us here at @DukeRadiology @duke_dair Apply here: careers.duke.edu/job-invite/…

Colon cancer cases are rising among younger adults, according to a study from the American Cancer Society this month. Here's what to know about early signs, screening and prevention. nyti.ms/3FBPXfc

I can't recall a @UCSF grand rounds that was more fascinating or mind-blowing than this far-ranging discussion of #ChatGPT's implications for healthcare – clinical, research, and education – by @atulbutte, @AaronNeinstein, @SaraMurrayMD & @LowensteinMD. youtube.com/watch?v=j-aOCuzf…

This is a game changer. Segment 100 structures in any whole-body CT image in 2 minutes using TotalSegmentator in @3dslicerapp All free, open-source software. Runs on any computer, no GPU is required. See more information at discourse.slicer.org/t/26710

How can we prepare our clinical trials infrastructure to rapidly execute clinical trials in case of public health emergency? This is about biodefense, preparedness, equity, diversity, and reestablishing trust in science.

Should @elonmusk give $ 1billion directly to the poorest people in the world via @GiveDirectly vox populi, vox dei

A great resource (I took a similar class with lectures by @jcjohnss when he was at Stanford).