⚡ Around the Wards
AI scribes moved the burden instead of removing it: Dr. Ambreén Zaman Riaz, a physician and two-time founder who now builds clinical decision-support AI, says the interface “has shifted from typing to vigilance.” Full take below.
A West Virginia hospital in a strip mall was built for about $18 million: open since 2021, with 10 ED beds, 10 inpatient beds, no OR, and specialists on a Teladoc cart.
Navina hired Salesforce Health Cloud’s creator as CMO: Dr. Joshua Newman’s mandate includes “enterprise health system and payer executive relationships.” Navina says 20,000+ clinicians and care team members at 1,300+ clinics use it.
😤 Haters: “A CMO who sells?” That’s most vendor CMO jobs now. Read the posting before you apply.Frontier models invented findings when a photo contradicted the X-ray: hallucinated findings rose from 25–37.5% to 45–57.5%. More in the Evals Desk.
🩻 Real or AI?
Two photos of an open mouth with a red, bumpy tongue, both labeled “scarlet fever.” One is real. The other was generated by AI. Which is which?
A
B
Answer at the bottom.
🔬 The Big Thing
The scribe didn’t take the work away. Did it just hand you a different job?
Dr. Ambreén Zaman Riaz builds clinical AI for a living. On Nov. 5 she wrote that AI documentation “did not reduce clinician burden. They moved it.”
Her version: “The interface has shifted from typing to vigilance.”
After the draft appears, you still check that it’s accurate, that your intent survived, and that it supports the billing and the quality measures. Then you imagine how an auditor will read it.
Minutes saved typing is a clean number. Minutes spent checking is real work, and almost nobody measures it.
📜 History rhymes: HITECH-era EHRs were sold as efficiency too. Years later, time-motion studies found much of the typing had moved to evenings at home. The clinicians who measured that early shaped what came next. The ones who didn’t got “pajama time.”
😤 “My scribe saves me an hour a day. Don’t take it away.” Nobody’s taking it. The question is whether the hour shows up in your evening or just moves to the morning sign-off queue.
😤 “This is a founder selling her own approach.” Partly. She builds decision-support AI, not scribes, so read it as a pitch for her side of the market. The questions she asks still hold up.
Ten beds, one cart.
Mon Marion Neighborhood Hospital sits between a DMV and a Tractor Supply. Operator Emerus runs or owns 52 micro-hospitals and has 24 more in development. Specialists arrive by video.
Ten beds and a cart on wheels. For a patient 40 minutes from anywhere, that’s the whole hospital.
😤 “Micro-hospitals are just freestanding EDs with beds.” Mostly. The beds are what let you keep a patient overnight instead of putting them in an ambulance.
The doctor whose job is selling
Navina’s new CMO, Dr. Joshua Newman, spent nearly 14 years at Salesforce. He’ll “partner with commercial go-to-market teams,” per the release.
When a clinical AI company’s CMO hire is about enterprise sales, the bottleneck isn’t clinical credibility. It’s getting a yes from the health system.
🔮 Where this lands: the clinician seat that grows fastest at vendors is the one that translates for buyers, not the one that signs off on safety.
athenahealth confirmed to Becker’s that it’s collaborating with Epic on data exchange for ambulatory practices. Paul Brient announced it on stage at THRIVE26, and Colin Hung reports the goal is for athena practices to appear as an Epic system inside Epic. No timeline, no named capabilities, and no comment from Epic yet. Wait for a spec before you plan around it.
Trick-or-Treat: adversarial evaluation of multimodal medical reasoning
Researchers paired real chest X-rays with synthetic clinical photos that either matched or contradicted them. Three frontier models hallucinated significant photo findings in 45–57.5% of mismatched pairs, versus 25–37.5% for the photo alone. One read acne as chickenpox crusts to keep the diagnosis tidy.
Why it matters: Conflicting evidence makes models invent findings to fit the story. Single-image accuracy understates that risk.
😤 “Residents do that too.” Yup.
💡 BTW: Joshua Newman, Navina’s new CMO, designed and coded his own health apps during residency and fellowship, for residency administration, billing and EHR adoption. He was building before it was a job title.
🩻 Real or AI answer: A is real, B is AI.
The tell: Hard one. The fake’s palate petechiae are a perfectly symmetric spray, and its red tongue dots are evenly spaced. Real papillae sit irregularly and are densest at the tip and edges.
Real image (as far as we can tell): https://commons.wikimedia.org/wiki/File:Strep_throat_with_white_strawberry_tongue.jpg CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0), via Wikimedia Commons
How we know B is fake: we made it with Nano Banana 2.1 (Google) on 2026-11-05. Google signs the original file with Content Credentials (C2PA: “Created by Google Generative AI”) and an invisible SynthID watermark.
The future is here but just not evenly distributed, let me know your thoughts about the future at kevin@clinicians.build.
— Kevin & AI
(Please verify content for yourself, partially AI generated and may contain errors.)




