The brief · 24 February 2026
Today, AI is showing up everywhere you'd least expect it: in your chemistry lab, on your Oura ring, and already in use by 70% of healthcare players according to NVIDIA. Four stories, one common thread: the pace of adoption is far outrunning what we think we're seeing.
NVIDIA's annual survey polls medtech, pharma, and hospital players, not curious dabblers. The shift is real: the question is no longer whether AI works but where it pays off most, with medical imaging leading the pack.
practitioners › 61% of medtech and 57% of pharma/biotech respondents report concrete ROI; 85% of those surveyed plan to increase their AI budget in 2026.
blogs.nvidia.com →Berkeley Lab has wired AI directly into its chemistry experiments to observe, adjust, and validate in real time while the reaction is still running. What used to take months of lab cycles now compresses into minutes. For labs hunting new materials, this is the kind of tool that changes the pace of discovery more than the discovery itself.
practitioners › The DTCS platform, announced February 24, 2026 by Berkeley Lab and NERSC, tunes an experiment's parameters mid-run instead of waiting between separate runs.
www.nersc.gov →Standard models write one word at a time. Mercury 2 uses diffusion to generate multiple tokens in parallel, making it noticeably faster without sacrificing quality, according to Inception Labs. If you're using the OpenAI API in a project, switching to Mercury 2 should be possible without rewriting your code.
practitioners › Compatible with the OpenAI API, available since February 24, 2026.
www.inceptionlabs.ai →Oura built its own LLM instead of calling GPT or Claude under the hood. The model focuses on cycles, fertility, pregnancy, and menopause, feeding into Oura Advisor. Whether a purpose-built model actually beats a well-prompted general-purpose one on these topics remains to be seen.
practitioners › Proprietary model announced February 24, 2026, integrated into Oura Advisor for Oura ring wearers.
Oura →