Several generative AI tools were tested for their image-generation capabilities, including Gemini and ChatGPT. Not a systematic review, but fascinating nonetheless. www.washingtonpost.com/technology/i...

Researchers analyzed higher-ed course grades and peer ratings to assess how LLMs affect learning outcomes. Writing a first draft and using an LLM for revisions compared favourably to the 'no LLM' baseline in terms of final grades.
www.frontiersin.org/journals/art...

Minecraft as an AI gym. Great article on how multi-agent tools get developed and tested. www.sciencefocus.com/future-techn...

Check out our December 8 issue of @NEJMClinician. clinician.nejm.org/edition/2025...

How artificial intelligence (AI) can revitalize clinical research by building "down" from patient-level data to hypothesis-generation and in-silico testing, contrasting with the current approach of building "up" from molecular parts. by Ziad Obermeyer @nejm.org www.nejm.org/doi/10.1056/...

Stroke AI imaging support in England: more than 450,000 stroke admissions were analyzed. AI use doubled thrombectomy rates, cut transfer delays, and improved outcomes. Not subtle. www.thelancet.com/journals/lan...

How AI scribes transcribe encounters. So much can be lost. Authors call for keeping the raw data which makes total sense. @NEJM www.nejm.org/doi/10.1056/...

Great article in STAT, including a table summarizing AI scribe clinical research, including recent trials published in NEJM AI. Briefly: meaningful impact on burnout, less so on measurable EHR-related metrics.
www.statnews.com/2025/12/03/a...

Regulating AI is very similar to assessing and licensing doctors. Great article in JAMA IM, and similar to my 2025 NEJM AI perspective. jamanetwork.com/journals/jam...