Sponsored by: Keragon
What’s the first job you’d hand to an AI agent?
Healthcare teams are already building them on Keragon: intake, reminders, no-shows, insurance checks. Describe yours in plain English, connect your EHR, and the agent handles the rest.
Start with one: No-show recovery · Patient intake · Insurance verification
Hey all,
We got the hotly anticipated Anthropic S-1 this week and much more. Here we go:
Anthropic’s S-1: medicine front and center
AI x bio is on a tear
How much care can AI actually deliver?
48 tools, partnerships & deals, plus new research, policy updates, AI jobs & link-worthy content
Read time: 7 minutes
Our Picks ✨
Highlights if you’ve only got 2 minutes…
1/
Anthropic’s S-1: medicine front and center
Anthropic just filed its S-1 ahead of its IPO, offering a rare look at the scale of one of the fastest-growing AI companies in the world. The financials are pretty remarkable. Some of the biggest numbers:
$16.3B in revenue during H1 2026, up 14x year over year.
$693M operating loss, down from $2.1B a year earlier.
~$9B in compute and infrastructure spending over six months.
11,900+ customers spending at least $100K annually, including 1,900+ spending over $1M and roughly 250 spending over $10M.
~900% net revenue retention, highlighting how quickly existing customers are expanding their usage.
One of the most interesting details is how Anthropic opens its filing. The company puts biology and medicine FIRST among the areas where AI could have the greatest impact on humanity, pointing specifically to the possibility of curing cancer, Alzheimer's, heart disease, and rare genetic disorders. The filing also highlights several healthcare and life sciences applications already being explored:
Clinical trials: Reducing the time required to produce clinical trial reports from months to minutes.
Drug discovery: Working with pharmaceutical companies to accelerate basic research and clinical development.
Biotechnology: Exploring AI-assisted molecular machine discovery, with potential applications in developing new biological tools.
Global health: Using AI for malaria modeling and other public health initiatives in collaboration with the WHO.
A pretty remarkable look at both the scale of Anthropic's business and where it sees AI heading. If you’re building in healthcare AI, you’re in the right place. (link)(linkedin)

2/
AI x bio is on a tear
Models, protein design, scientific agents, robotic labs, pharma deployments, and even AI-designed biology reaching a patient. A lot moved at once this week:
Radical Numerics: Omnii-1 designed phages that were administered to a patient with an antibiotic-resistant infection under compassionate use. (link)
Anthropic: Claude generated 318K+ protein designs targeting scorpion toxins. 211 were tested in the lab and 17 met Anthropic’s criteria for a binder. (link)
OpenAI: acquired Orchestra Bio and is bringing its drug-development agents into Rosalind. (link)
OpenAI + PerturbAI + Scripps: thousands of agents analyzed data from millions of brain cells in 32 minutes and surfaced new leads around Alzheimer’s and Parkinson’s. (link)
NVIDIA + UW Institute for Protein Design: released RFD4-Proteina, an open model that jointly designs protein sequences and all-atom structures. (link)
Microsoft: released BioEmu-2, a 500M-parameter model that generates ensembles of 3D biomolecular structures to study how molecules move and interact. (link)
Rosetta Commons: launched Chaperone, a framework for connecting tools like RFdiffusion3, AlphaFold3, ProteinMPNN, and Boltz into protein-design workflows. (link)
Ligo Biosciences: launched Artios, an open enzyme-design system that posted state-of-the-art results on its reported benchmarks. (link)
Xaira: launched X-Scientist for agentic drug discovery, then X-Automate to connect its AI agents with robotic lab equipment. (link)
Sophont: launched Holon-S1, which predicts spatial gene and pathway activity directly from routine pathology slides. (link)
Boltz + Bristol Myers Squibb: bringing Boltz’s biomolecular models into BMS drug-discovery workflows and working together on custom models using BMS scientific data. (link)
Phylo + Bayer: deploying Biomni Lab across Bayer research teams for literature, omics, and drug-discovery workflows. (link)
Warp Sciences: former OpenAI science lead Kevin Weil launched the company with a $215M Series A to combine AI, robotics, and automated labs for drug discovery. (link)
Hugging Face: released Carbon-X, an open 14.8B DNA model that reads raw DNA alongside natural language. Its team reports an average score of 66.7% across five DNA benchmarks versus 67.5% for GPT-6.1 Sol, at roughly 1/1,800 of the inference cost per problem. (link)
This is happening across the whole stack: models, protein design, simulation, agents, robotic labs, pathology, pharma workflows, and actual patient use. Bio is slowly becoming computer. (linkedin)

3/
How much care can AI actually deliver?
McKinsey estimates AI could already perform work represented in 16% to 22% of US outpatient claims, or roughly 2B to 3B claims a year and 13% to 19% of outpatient spending.
Most of that is “discursive care,” the reasoning and conversation around medicine rather than hands-on treatment. Think intake, triage, low-complexity visits, referrals, follow-up, and interpreting diagnostic results. McKinsey also estimates AI could perform a meaningful share of core tasks for 11M of the 19M US healthcare workers.
The access implications are meaningful. Around 92M Americans live in primary care shortage areas, with average appointment waits around 31 days. More AI-delivered care could help stretch clinical capacity without waiting years to train more physicians.
Health system leaders are more cautious on the 22% number. Several called it plausible; others think the real figure is much lower, and almost everyone came back to the same issues: validation, escalation, liability, and who is accountable when AI gets something wrong.
So the debate is increasingly about where to draw the line and how quickly to move it as the evidence gets better. We should be open to pushing that boundary where the evidence supports it, so long as the priorities stay simple: make healthcare cheaper and better, put the patient first, and make technology serve the people delivering care rather than burden them. (link)

Healthcare AI Rundown 🔍
Latest on tools, partnerships, research and policy…
Tools & Product Updates 🔧
Microsoft’s diagnostic AI scores 83% on complex medical cases: MAI-DxO beat physicians using references, who scored 36%, across 304 simulated NEJM cases; prospective testing is planned at Beth Israel Deaconess. (link)
Speechmatics launches a multilingual medical speech model: Oak transcribes 55+ languages through one API and can follow clinicians and patients as they switch languages within the same conversation. (link)
Epic launches Connectors to link its AI agents with third-party apps: Built on MCP, Connectors let Epic’s Emmie, Art and Penny assistants access information and perform tasks across external applications. Developer-built Connectors will be available through Epic Showroom starting March 2027. (link)
careCycle launches a free Medicare plan comparison tool: Match Your Medicare compares up to 3 plans side by side, including estimated costs and in-network doctors. (link)
Inovia Bio launches AI workflows for drug development: Falcon drafts study protocols, clinical reports and forecasts in hours, the company says, with human review and sign-off. (link)
Talkiatry takes virtual therapy nationwide: The expansion broadens its behavioral health platform, supporting patients with needs ranging from routine therapy to more intensive care. (link)
Protege launches independent evaluations for healthcare AI: The program uses double-blind testing on real-world patient data, starting with a small cohort of AI developers. (link)
micro1 launches a Medical Index for AI models: The live leaderboard combines 3 expert-built evaluations covering pathology, FDA drug documentation and adolescent health safety. (link)
Eko launches FDA-cleared stethoscope with analog and digital listening: CORE 300 combines traditional listening with digital amplification and supports AI-assisted structural murmur detection through the Eko app. (link)
athenahealth connects clinics and surgery centers in athenaOne: The new ambulatory surgery center offering unifies patient records, scheduling, documentation and billing across clinic and surgical workflows. (link)
Vene launches an AI advocate for navigating healthcare: The company says 5,000+ families use its care coordination service, with rollout planned for 50,000 additional patients across provider networks. (link)
Concurrence launches Superintelligent Care research program: The initiative aims to combine fragmented patient records into models that forecast individual health and transfer insights across patients. (link)
Notable gives operators more control over their AI workforce: Flow Compass lets teams query automation performance, generate test cases, trace failures and track ROI through plain-language requests. (link)
Partnerships & Deployments 🤝
Memorial Sloan Kettering + OpenEvidence: OncoKB brings MSK’s expert interpretation of tumor genetics into OpenEvidence for clinicians nationwide; over 85% of MSK oncologists already use the platform. (link)
UCSF + Conceivable Life Sciences: The planned UCSF Fertility Excellence Center will study AI-assisted IVF laboratory automation using Conceivable’s AURA platform, with embryologists supervising the technology. (link)
Sentara Health + Regard: The health system deployed a co-developed AI case management tool that lets case managers query patient charts in plain language to support care transitions. (link)
Microsoft + Opella: The partners are building Lumen, a platform with 14 AI agents that check health content claims, sources and compliance, with human experts retaining final approval. (link)
K Health + Wix: The partners will integrate PatientGPT’s record-grounded clinical AI into adaptive websites that personalize health system navigation around each patient’s history and needs. (link)
Doximity + Wolters Kluwer: The partnership will bring content from UpToDate’s 13,000 topics into Doximity Ask alongside AI-generated answers, with rollout expected in the coming weeks. (link)
Epic + Amazon Pharmacy + Cost Plus Drugs + Costco + LillyDirect: A planned prescription marketplace will let clinicians and patients compare insurance and cash drug prices before prescribing. (link)
Oracle Health + U.S. Department of Veterans Affairs: The VA’s EHR pharmacy module now supports AI-drafted pharmacist notes, with a Pharmacy Forward upgrade planned for 2027. (link)
Fortuna + R1 RCM: Fortuna will bring digital Medicaid enrollment into R1’s financial clearance workflows, connecting coverage enrollment with providers’ existing revenue cycle processes. (link)
Oura + Counsel Health: Eligible members can access medical AI and board-certified doctors inside the Oura app, with consent-based sharing of biometric data to inform their care. (link)
St. Luke’s Health System + hellocare.ai: The Idaho health system is deploying AI-assisted virtual nursing, telehealth and virtual patient observation across inpatient settings in Southern Idaho. (link)
Google Health + Flo Health: Consent-based app syncing will connect Flo’s cycle insights with wearable signals and Google Health Coach, with rollout beginning in early 2027. (link)
NVIDIA + Siemens Healthineers: The partners are exploring AI-assisted radiotherapy planning, patient-specific simulations and real-time computing to help clinicians adapt cancer treatment to individual patients. (link)
NVIDIA + Danaher + Benchling + Automata: Scientific agents will connect protein design with laboratory workflows, including Genedata’s planned use of RFD4-Proteina in Danaher’s autonomous lab. (link)
Research & Evidence 🔬
AI recommends care costing 3.2x as much as physicians: Across 200 cases, 20 general-purpose models recommended care averaging $314 per visit versus $97, using 2026 Medicare rates. (link)
Mass General Brigham AI estimates cancer survival from face photos: Researchers used FAHR-Face, trained on 40M+ images, to develop tools estimating biological age and survival risk. (link)
Mass General Brigham’s smaller AI beats larger clinical models: Trained on records from 3.6M patients, Clinical System One outperformed open models up to 27B parameters on 13 of 14 tasks in a preprint. (link)
Lirio reports 20%+ increase in vaccine uptake with AI outreach: The program reached millions of patients, with 40% of message recipients getting vaccinated, according to the company. (link)
63% would wait for human review of AI medical messages: In a 2024 survey of 1,185 adults, most preferred a 48-hour wait for clinician-reviewed medical replies over instant AI answers. (link)
Singapore researchers use casual speech to flag psychosis risk: NTU and IMH’s AI framework identified up to 93% of at-risk youths by analyzing emotional patterns in short speech samples. (link)
Policy & Regulation 🏛️
FDA expects draft generative AI device guidance within months: Its digital health director outlined the timeline after the agency received 300+ comments on its regulatory discussion paper. (link)
OpenAI discloses model bypassing security to retrieve hospital statistics: During internal training, the model retrieved server secrets and bypassed a government website’s authorization checks while researching hospital-bed counts. (link)
Iowa sues OpenAI over ChatGPT’s medical advice and safety: The attorney general alleges misleading safety claims and risks to vulnerable users, including dangerous medical advice; OpenAI disputes the allegations. (link)
Bob Wachter argues healthcare must deliberately encode AI’s values: The UCSF medicine chair says systems should make treatment tradeoffs visible and balance patient safety with appropriate testing and spending. (link)
Deal Desk 💰
Spotlight on latest capital raises, M&A, and investments…
FUNDING
OpenAI, is nearing a funding round of up to $30B at a $1.25T valuation, led by UAE investors. (link)
Warp Sciences, an AI drug discovery startup led by ex-OpenAI science head Kevin Weil, raised $215M in early-stage funding. Lightspeed Venture Partners led, joined by IVP, Lux Capital, and Premji Invest. (link)
Baba, an AI-powered patient advocacy startup for Medicare and Medicare Advantage, raised $40M in Series A funding led by FirstMark Capital, with CVS Health Ventures and Valtruis joining. The total funding is now $47M. (link)
Sunlight AI, a New York-based AI company that runs the business side of independent medical practices, launched with $17.4M in seed funding. Lan Xuexhao led, joined by Social Leverage and others. (link)
Kivo Health, a San Francisco-based telehealth provider for lung disease, raised $10M in Series A funding. Allumia Ventures led. (link)
BioStack, a San Francisco-based healthcare data network for training medical AI, raised $9.5M in seed funding. Peak XV Partners, Afore Capital, DCVC, Emerson Collective, and SV Angel participated. (link)
XRlabs, a London and Palo Alto-based surgical intelligence company, raised $8.4M in seed funding. Playground Global led, joined by Eka Ventures. (link)
Moodr Health, a West Virginia-based patient engagement platform for health plans and health systems, raised $3M in seed funding. (link)
M&A
OpenAI + Orchestra Bio: OpenAI will integrate the startup’s AI drug development platform with Rosalind to help biopharma teams coordinate research and development workflows. (link)
Redesign Health + ARC Global: The two are merging into an end-to-end platform that helps health systems turn clinician ideas into technology products and standalone companies. (link)
Carenet Health + Invene: Carenet acquired the healthcare AI and data engineering consulting firm, adding AI and data consulting to its engagement suite. (link)
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…
Staff Growth Marketer at Abridge, an AI clinical documentation platform
$200K - $250K | San Francisco, New York City (link)
Technical Product Manager, EHR at Nabla, an ambient clinical AI platform
$140K - $180K | New York City (link)
Lead Product Data Analyst at Hippocratic AI, AI clinical agents for healthcare
N/A | Menlo Park, CA (link)
Visuals of the Week 📸
Funny memes, cool pics, and interesting data from around the web…


life of man

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👇




