Sponsored by: Keragon
Meet the AI agents built for healthcare.
Got manual work? Describe how you want it done. Keragon builds an agent to do it.
Intake, reminders, eligibility, no-shows, referrals, scheduling, billing, and more.
Just connect the EHR and tools you already use.
Live in 5 minutes.
Good morning —
Let’s dive right into the best health AI news roundup out there:
The “pacing AI” debate and what it means for healthcare
ARPA-H is testing autonomous clinical AI
How AI reshapes nursing
23 new tools/partnerships, 11 funding updates, new AI jobs & link-worthy content
Read time: 6 minutes
Our Picks ✨
Highlights if you’ve only got 2 minutes…
1/
The “pacing AI” debate and what it means for healthcare
Something clearly shifted in the AI debate this week. Anthropic’s Dario Amodei called on frontier labs to “pace the frontier,” arguing that capabilities are moving faster than safety work can keep up. Sam Altman agreed with the basic premise, Elon Musk backed more oversight, and Google DeepMind’s Demis Hassabis called the proposal a step in the right direction.
There is at least one concrete change coming out of all this. Anthropic committed to giving independent third-party evaluators permanent, employee-like access to its systems, and Altman said OpenAI will do the same. The idea is to let outsiders inspect models during development and verify safety practices rather than relying entirely on labs to grade themselves. Amodei has floated groups such as METR as possible evaluators.
The broader proposal goes much further, including common standards across frontier labs and international coordination. That has opened another debate. Critics worry stricter rules could conveniently strengthen the largest labs by raising the cost of competing, limiting open models, or slowing an expensive compute race. There is no clear evidence that recent warnings from former lab employees are coordinated, but the timing and stakes have made regulatory capture an obvious talking point.
Geopolitics makes all of this harder. Amodei argues that any pacing regime has to preserve the lead of the US and its allies. President Trump has rejected a broader slowdown, framed AI leadership as a strategic competition with China, and criticized efforts to restrict AI and data-center development.
Meanwhile, public sentiment is getting worse. One recent poll found 35% of Americans see AI as mostly bad versus 16% mostly good. Healthcare has historically been the exception. A Gallup survey found 61% had a positive view of AI’s impact on medical diagnosis and treatment, the only area tested where positive sentiment outweighed negative sentiment.
That puts health AI in an unusually important position. If we can show AI helping doctors, catching disease earlier, expanding access, reducing burnout, and making care cheaper, healthcare can be the shining light on the hill for the AI era.
The AI race is only getting more intense, and the arguments around it will probably get messier. That gives those of us working in health AI a pretty large responsibility. We need to earn trust with real outcomes, sensible safeguards, and products that genuinely make care better. If healthcare gets this right, it can show the public what all this technology is actually for. (link)(tweet)
2/
ARPA-H is testing autonomous clinical AI
The Advanced Research Projects Agency for Health (ARPA-H), a US federal funding agency that invests in high-risk, high-impact biomedical and health breakthroughs, just launched ADVOCATE, a $62.7M program to test autonomous AI agents for heart failure care.
The idea is pretty concrete. AI agents would monitor patients between visits, make certain care recommendations or medication adjustments, and escalate to clinicians when needed. Atman Health, Tempus AI, and UpDoc are building the patient-facing agents, while Stanford is developing a supervisory AI to watch for unsafe recommendations. Duke and Kaiser will handle real-world deployment and testing.
ARPA-H is also working with FDA and CMS on regulation and reimbursement, with a 39-month path from prototype to clinical use. Kaiser plans a randomized trial of about 2,500 patients across 21 medical centers.
This is worth watching because it turns the autonomous AI debate into something testable. Rather than asking whether AI should replace doctors, ADVOCATE is asking a narrower question: which parts of care can AI safely handle on its own, and when should it hand control back to a clinician? (link)
3/
How AI reshapes nursing
Anthropic’s Economics team built a model of how AI could reshape jobs, wages, and growth through 2030. They use nursing as a good example of what that transition could actually look like. Today, a nurse’s job is a bundle of very different tasks: triage patients, coordinate with doctors, draft discharge instructions, order supplies, monitor patients, and provide hands-on care.
AI affects each one differently:
Unchanged: physical care like bathing a patient
Augmented: drafting discharge instructions, monitoring patients remotely, planning shifts
Automated: charting vitals or ordering supplies
New: checking how well AI triaged a patient or reviewing an AI-generated care plan
The result isn’t necessarily “AI replaces nurses.” Anthropic’s model suggests nurses could oversee more work while spending more time with patients, as AI handles more of the administrative and cognitive load. And there’s a bigger labor-market implication: in more aggressive AI scenarios, displaced knowledge workers may actually move into less AI-exposed jobs like nursing.
So one possible AI future looks less like healthcare jobs disappearing, and more like the job itself changing: fewer repetitive tasks, more AI supervision, and more time on the parts of care that still require humans. (link)(linkedin)
Tools & Partnerships 🔧
Latest on business, consumer, and clinical healthcare AI tools and partnerships…
TOOLS
Google DeepMind maps every possible human DNA mutation: AlphaGenome Atlas is a free, searchable map predicting the regulatory effects of all 9 billion possible single-letter DNA changes, helping researchers prioritize variants tied to disease. (link)
Apple brings more AI into everyday health tracking: New Apple Watch and iPhone Health features add higher-frequency heart rate and HRV sensing, an AI Readiness Score, Health Age, and AI-powered movement evaluations. (link)
AMA adds 10 AI-specific codes to CPT 2027: The new code set includes 299 additions overall and brings the total number of AI-related CPT codes to 43, expanding reimbursement infrastructure around AI-enabled clinical services. (link)
Withings turns BeamO into an AI stethoscope: StethO Sense adds on-demand AI analysis of heart and lung recordings to the FDA-cleared three-in-one device, using eMurmur to flag abnormal murmurs and StethoMe to analyze breathing sounds, with results delivered in seconds through Withings+. (link)
Abridge pushes into revenue cycle with AI pre-bill review: The new tool compares inpatient diagnoses and DRGs against clinical documentation before a claim is submitted, surfacing discrepancies and supporting evidence for CDI and coding teams while keeping final decisions in human hands. (link)
Insilico’s AI-designed drug shows an early anti-aging signal: In follow-up analysis from a Phase 2a trial, all six biological aging clocks estimated rentosertib-treated patients as younger, with one model showing reductions of 2.7 to 3.5 years. (link)
NVIDIA speeds protein structure prediction nearly 3x: BioNeMo Inference Runtime delivered 2.9x higher throughput in a 1,000-target benchmark, helping researchers run structure-prediction models across much larger biomolecular workloads. (link)
FDA approves daraxonrasib for pancreatic cancer: Revolution Medicines’ KRAS-targeted drug won FDA approval for advanced pancreatic cancer, with the source reporting substantially improved survival and lower mortality risk. (link)
Penguin AI brings autonomous medical coding to AWS: Its new coding and RCM tools automate code validation, claims review, denial workflows, and chart summarization while keeping healthcare data inside providers’ own AWS environments. (link)
Xsolis brings GenAI to payer denial reviews: Its Peer-to-Peer Clinical Synopsis automatically turns medical records into concise clinical summaries, helping physician advisors prepare faster for time-sensitive reviews following payer denials. (link)
Avon launches rules engine for touchless RCM: Practices can create payer-specific rules that automatically review and fix claims before submission, reducing manual billing work as Avon pushes toward more automated revenue cycle workflows. (link)
FDA-cleared AI devices show a major clinical evidence gap: A study found just 3 of 1,357 FDA-cleared AI medical devices were evaluated for patient outcomes, while only 2.5% were linked to prospective trials and 0.9% underwent peer review. (link)
LUMA Vision wins FDA clearance for 360° cardiac navigation: The expanded VERAFEYE platform combines real-time imaging, 3D anatomy, catheter tracking, and procedural guidance to give physicians a more complete view during minimally invasive cardiac procedures. (link)
AI predicts which ICU patients are at risk of underfeeding: The model flags mechanically ventilated patients likely to receive less than 70% of their daily calorie needs, giving clinicians an opportunity to intervene before nutritional deficits worsen. (link)
Forward-deployed engineers spread across healthcare: FDEs work directly alongside frontline users to identify workflow problems, build tailored software, and rapidly adapt solutions to each health system’s local needs. (link)
Mercy brings ambient AI for nurses to Android: Mercy became the first health system to deploy the nursing documentation tool on Android after pilot units cut flowsheet documentation time by 22%. (link)
Mayo and Microsoft target “whole patient” AI: Their new frontier model combines longitudinal clinical data to support earlier diagnosis, personalized treatment, and a more holistic view of patients over time. (link)
PARTNERSHIPS
Cleveland Clinic + Luminai: Cleveland Clinic partnered with Luminai to automate complex administrative workflows, beginning with referral fax transcription and expanding to additional enterprise use cases. (link)
OpenEvidence + MINC#NIMC: OpenEvidence partnered with Canada’s Medical Identification Number Corporation to streamline physician verification and maintain free access to its clinical AI platform for verified Canadian physicians. (link)
HealthEx + Meta: HealthEx partnered with Meta to let users connect their medical records to Muse, enabling the AI agent to create visit summaries, draft physician questions, and flag medication refills. (link)
Heidi + HealthPathways: Heidi partnered with HealthPathways to integrate local Canadian clinical guidance and referral pathways directly into Heidi Evidence at the point of care. (link)
Anomaly Insights + Inova Health: Anomaly Insights partnered with Inova Health to deploy AI-powered payer intelligence and analytics aimed at identifying payer behavior and improving revenue cycle performance. (link)
Qualderm + Parakeet Health: Qualderm partnered with Parakeet Health to deploy agentic AI across nearly 160 practices, automating patient access workflows and helping fill canceled or unused appointments. (link)
Deal Desk 💰
Spotlight on latest capital raises, M&A, and investments…
FUNDING
Forus (fka Tandem), an AI platform for doctors' offices, raised $150M in Series C funding at a $3B valuation. Bain Capital Ventures led, joined by insiders Thrive Capital, General Catalyst, Accel, Redpoint, BoxGroup, and Pear VC. (link)
Tandem Health, a Swedish medical scribe startup, raised $100M in Series B funding. Scaleup Europe Fund led, joined by Kinnevik, Northzone, Amino Collective, and Visionaries. (link)
Archy, a San Jose, Calif.-based provider of practice-management software to dental offices, raised $50M in Series C funding. JMI Equity led, joined by TCV, Entrée Capital, Bessemer Venture Partners, CRV, and Alven. (link)
Implicity, an AI-driven cardiac remote monitoring platform, raised $40M in growth equity funding led by IRIS, with Five Arrows also participating. (link)
EZ Health, a senior-care startup, raised $26M in pre-seed and seed funding from backers like Primary Ventures and Construct Capital. (link)
Epsilon Health, an SF-based provider of AI-enabled radiology services, raised $20M in Series A funding. AlleyCorp led, joined by Uncork Capital, Renegade Partners, and SemperVirens. (link)
GenHealth.ai, a healthcare AI company building back-office agents for providers, raised $16.5M in Series A funding led by Flare Capital Partners. (link)
Graph AI, the Pleasanton, Calif.-based developer of AI software for drug-safety reporting, raised $13.3M in Series A funding. Insight Partners led the round and was joined by existing investor Bessemer Venture Partners. (link)
Nara Health, a third-party administrator, raised $10M in seed funding led by Khosla Ventures. (link)
Verily, a precision health AI company, secured an undisclosed investment from Nvidia to extend the $300M round it announced in March. (link)
M&A
H1 + Defacto Health: H1, an AI-powered healthcare data platform, acquired Defacto Health to expand its provider network intelligence offerings for health plans. (link)
as of 9/13/26
Other Relevant News 🔍
News, podcasts, blogs, tweets, resources, etc…
AI Job Opportunities 💼
Explore our AI Job Board or contact us to feature roles in our newsletter…
Visuals of the Week 📸
Funny memes, cool pics, and interesting data from around the web…
That’s it for this week friends! Back to reading — I’ll see you next week.
Stay classy,
PS. I write this newsletter for you. So if you have any suggestions or questions, feel free to reply to this email and let me know
How was this week's newsletter? Tap your choice below👇
