Sponsored by: Knack Health
Build and launch HIPAA-compliant healthcare apps with the AI tools you already use. Knack Health's MCP Server connects Claude, ChatGPT, and Lovable directly to a compliant backend, BAA included.
No DevOps team, just AI-built apps on a compliant data layer, ready to ship.
Good morning, readers —
Let’s get straight into it:
No signs of slowing in health
The AI billing bot war
How AI helps a shrinking surgical workforce
42 new tools/partnerships, 14 funding updates, new AI jobs & link-worthy content
Read time: 7 minutes
Our Picks ✨
Highlights if you’ve only got 2 minutes…
1/
No signs of slowing in health
The “pace the frontier” debate has dominated AI over the past few weeks. At the same time, reports that former Anthropic researcher Jacob Coxon worked with a PR firm tied to prominent AI-safety voices have raised questions about how organic some of the recent media frenzy really was.
Concurrently, the market is rapidly shifting away from a handful of closed frontier labs, which makes the timing of those same labs suddenly calling for slowdown and regulation look a little fishy. According to Vercel data, token usage flipped in roughly 12 weeks from ~20% open models / 80% closed to ~80% open / 20% closed. That is an extraordinary shift in a very short period.
And while everyone debates “slowing down”, the release calendar looks like the opposite (closed frontier models included): DeepSeek 4.1, Qwen 2.1 Flash, Nemo Pro, Opus 5.5, GPT-6 Astra, Sol and Luna, Grok 4.7, Meta Muse and more. Open or closed, the direction is mostly the same: stronger models, lower prices, faster iteration. Healthcare is moving with it via the bigger announcements from last week:
Anthropic Life Sciences launched a biology group and wet lab, where ~950 Claude agents spent 21 hours searching DNA and uncovered a previously uncharacterized enzyme system with CRISPR-like features. (link)
Anthropic x OpenEvidence is bringing free clinical decision support to doctors across ~100 low- and middle-income countries. (link)
OpenAI launched MentalHealthBench, built with 80+ psychiatrists and psychologists across 22 countries. (link)
Whatever happens with the pacing debate, the underlying trend looks hard to stop: more models, more competition, lower costs, and more intelligence flowing into healthcare. Onwards. (link)
2/
The AI billing bot war
One of the most discussed healthcare stories over the weekend was also one of the most American healthcare stories imaginable: hospitals are using AI to find more things to bill for, while insurers are increasingly using AI to scrutinize and deny those bills.
A new Blue Cross Blue Shield Association analysis estimates that AI-assisted hospital coding generated $942M in additional payments over two years. The share of inpatient cases coded as medically complex rose from 37% to 40%, including 55,000+ additional complex cases. BCBSA’s argument is that coding intensity increased without a matching rise in treatment, suggesting AI is finding more billable conditions without much evidence that patients became sicker.
Providers have a pretty reasonable counterargument. Medical billing is incredibly complex, and AI may simply be finding legitimate diagnoses that were previously missed. BCBSA’s analysis also relies primarily on claims rather than the underlying clinical records, so it cannot fully answer that question.
Meanwhile, payers are deploying their own automation for denials, downcoding, audits, and requests for more information. So we now have increasingly sophisticated AI on both sides fighting over the same healthcare dollar, with enormous administrative costs along the way.
There is another side to the math too. One estimate circulating over the weekend argued that BCBS plans could have saved roughly $2.2B over the same period by adopting AI more aggressively in their own operations, although that figure depends on a set of assumptions rather than the BCBS analysis itself.
And who ultimately pays for this arms race? Patients. Higher provider costs, higher insurer costs, and more administrative overhead eventually work their way into premiums, out-of-pocket spending, and the overall cost of care.
That is what makes the whole thing a little absurd. AI has the potential to be a major deflationary force in healthcare when it is used to expand clinical capacity, automate low-value admin work, help clinicians work more efficiently, and deliver more care at lower cost.
Instead, some of the biggest deployments are being pointed at figuring out who gets paid, how much, and whether that payment should be denied.
The bot war probably tells us more about the incentives of the US healthcare system than it does about AI. If we want AI to make healthcare cheaper, we should give it more opportunities to actually improve and deliver care or make the whole process simpler, cheaper, and less adversarial in the first place.
3/
How AI helps a shrinking surgical workforce
A new American College of Surgeons analysis puts the care access problem in pretty stark terms: 15M+ Americans live in a county with no general surgeon, while ~10M live somewhere with no specialty surgeon. General surgeons are absent from 38% of US counties, and 434 critical access hospitals sit in rural counties with no recorded full-time surgeon.
That matters even more as the population ages. Researchers note the average American may undergo as many as 20 diagnostic or therapeutic procedures over a lifetime, from endoscopies to major surgery.
AI isn’t built yet to put a surgeon in every rural hospital, but it could help stretch the workforce that exists. A separate Covista analysis estimates AI could close 37% of the projected healthcare workforce gap by 2040 without replacing a single worker, largely through documentation, referrals, medication lists, billing, and other administrative work.
The interesting question is how much further that goes as AI moves into diagnostics, care coordination, and eventually surgical assistance. More capacity will matter, because demand is ticking up.
Tools & Partnerships 🔧
Latest on business, consumer, and clinical healthcare AI tools and partnerships…
TOOLS
VA awards $775.72M ambient scribe contract to Knowtex and Carahsoft (+Abridge): The five-year IDIQ names Knowtex and Carahsoft as the two prime vendors under a shared ceiling. Abridge is participating through Carahsoft, while additional subcontractors can compete for task orders across the VA system. (link)
AI catches esophageal cancer on routine chest CT scans: Researchers unveiled EAGLE, an AI model validated on 80,000+ patients across 12 centers in three countries that detected 90% of esophageal cancers with 98.5% specificity. (link)
NVIDIA and partners open up protein structures from 2,800+ viruses: The new AlphaFold Database dataset gives researchers predicted viral protein complexes to study potential outbreak threats, alongside an open BioNeMo pipeline for generating additional structures. (link)
Neuralink helps a man with ALS speak using his thoughts: A participant in the experimental VOICE study used the brain-computer interface to translate speech-related neural activity into words spoken in a synthetic version of his own voice. (link)
Google’s MedGemma passes 10M downloads across global healthcare: The open medical AI models are being used for triage, screening, and public-health programs across hospitals, remote clinics, and other healthcare settings worldwide. (link)
Jefferson Health says AI returned 1M+ hours to clinicians in one year: The health system is expanding automation into nursing as it works toward a goal of saving 10M clinician hours by 2028. (link)
OpenAI agent accessed non-public files on an Australian Medicare portal: The incident occurred during testing and triggered an investigation, illustrating the security risks that can emerge when autonomous agents receive broad access to government systems. (link)
Novo Nordisk says AI is accelerating drug discovery at scale: The drugmaker reports a 3x improvement in target selection and 10x faster clinical-study report generation, supported by 30+ AI partnerships and hundreds of AI and digital employees. (link)
Ambitious Bio launches the world’s largest healthy human protein atlas: Corpus maps proteins across healthy tissues to give biological AI a more complete view of the human body, increasing detected protein inventory by as much as 78% in head-to-head tissue comparisons. (link)
Merge brings AI across enterprise imaging workflows: MergeIQ creates a common AI layer across radiology, cardiology, pathology, and other imaging specialties, initially surfacing relevant prior reports and clinical context directly inside diagnostic workflows. (link)
Neko Health opens its first US center in New York: The preventive-health startup opened its SoHo clinic after more than 25,000 New Yorkers joined the waitlist, marking the start of a broader U.S. expansion following its growth in Sweden and the U.K. (link)
R1 moves hospital utilization review from retrospective to real time: Phare UM continuously reviews patient encounters for payer and denial risk, with an early assessment finding roughly 65% of evaluated observation cases met inpatient billing criteria. (link)
athenahealth adds AI-native tools for value-based care: Its Clinically Inferred Diagnosis Gaps tool produced 2x higher closure rates and 3.5x lower dismissal rates than traditional risk-adjustment sources in beta testing. (link)
Doximity opens up its clinical AI safety benchmark: Bedside Bench tests models across 500 physician-validated cases covering clinical safety, calculations, hallucinations, drug safety, and health equity, with the grading rubrics released publicly. (link)
AutoMedBench puts medical AI agents through full research workflows: NVIDIA and UC Santa Cruz’s open benchmark covers 6,300+ runs, 96 task combinations, and seven tracks, with validation and submission errors emerging as major sources of agent failure. (link)
Hims & Hers expands its AI-native care model beyond chatbots: The system maintains persistent patient context, learns from outcomes, uses human-in-the-loop safeguards, and brings clinicians into the conversation when medical judgment or treatment is needed. (link)
Hale launches a platform for catching heart disease earlier: The preventive-health platform combines coronary calcium CT scans, advanced blood testing, physician interpretation, and ongoing monitoring to help people understand cardiovascular risk before a heart attack. (link)
Oracle expands AI across healthcare and life sciences: Oracle unveiled a new AI-powered oncology EHR, expanded AI across revenue-cycle workflows including prior auth, coding, and appeals, and added domain-trained AI agents and natural-language analytics to its life sciences data platform. (link)
Evidently uses AI to understand the entire patient record: Omni creates a longitudinal view across EHR data, scanned documents, and external systems, helping clinical and revenue teams surface relevant information from fragmented records. (link)
Americans want humans when AI conversations carry real stakes: A national survey found 67% prefer a real person for consequential calls, with healthcare showing particularly strong demand for human service and 57% preferring a human during a 911 medical emergency. (link)
Epic ramps up cybersecurity without slowing its AI roadmap: The company is putting more resources into identifying vulnerabilities while continuing development of Agent Factory, EpicOps, interoperability tools, and other AI capabilities. (link)
Payerset puts 20T+ healthcare pricing records behind an AI interface: Research Assistant lets health systems query payer rates, claims, policies, and benchmarks in plain language to support reimbursement analysis, contracting, and revenue planning. (link)
WellSky brings agentic AI to skilled nursing: New Navigator and RevenueQueue workflows surface changes in patient condition, documentation gaps, and revenue priorities so skilled-nursing teams can act sooner. (link)
Veeva says its AI agent can build clinical studies in one day: Study Builder Agent turns protocols into EDC forms, visits, edit checks, and data-quality listings while automating testing and reusing organizational standards. (link)
Real-world evidence boosts clinical LLM performance more than 300%: Atropos Health’s Precision Evidence Bench found models produced substantially better patient-specific answers when given high-quality real-world evidence alongside clinical context. (link)
Paper2Agent turns scientific papers into interactive AI agents: The framework successfully converted 74 of 100 computational-biology papers into agents that can answer questions, run analyses, and apply the papers’ methods to new data. (link)
Epic’s end-of-life AI model overpredicts mortality across 39 hospitals: A large external evaluation found the model maintained moderate-to-high discrimination but was poorly calibrated, reinforcing the need for ongoing local validation after clinical AI deployment. (link)
CentralReach launches AI to automate ABA reporting: ReportDraftAI pulls authorization, clinical, and progress data already in the platform into structured, review-ready drafts, reducing manual assembly work for BCBAs. (link)
Radiology groups are becoming AI developers themselves: Some outpatient and teleradiology practices are building, acquiring, and deploying their own AI tools, blurring the traditional boundary between technology development and clinical practice. (link)
Xsolis uses GenAI to prepare physicians for payer denial reviews: Its Peer-to-Peer Clinical Synopsis turns patient records into structured clinical summaries, reducing the manual chart review required before time-sensitive conversations with payers. (link)
PARTNERSHIPS
Sword Health + Brazil’s Unified Health System: Sword partnered with Brazil’s national health system to explore AI across clinical triage, waitlist prioritization, scheduling, clinical decision support, and prevention for a system serving 210M+ people. (link)
Concurrence + Databricks: Concurrence partnered with Databricks to bring its Health World Model and AI agents onto the Databricks platform, creating a unified foundation for healthcare data, AI, and care delivery. (link)
UTMB + OpenEvidence: UTMB integrated OpenEvidence directly into its EHR, giving physicians secure access to cited medical evidence within their existing clinical workflows. (link)
Tempus AI + Recursion: Tempus extended its data partnership with Recursion through 2029 and licensed Recursion’s RNA Foundation Model to support AI-driven precision medicine and drug discovery. (link)
Cigna Group + OpenAI: Cigna partnered with OpenAI to develop personalized AI support for patients with complex conditions, beginning with cancer care and care coordination. (link)
Vitality + Google: Built on Google Cloud and Gemini, Vitality AI uses 2,800+ health and behavioral dimensions to personalize recommendations, coaching, and incentives for employers and health plans. (link)
Business Group on Health + BioRadar: Business Group on Health partnered with BioRadar to give employers AI-powered biological threat intelligence for identifying emerging risks and improving workforce preparedness. (link)
Iambic + AbbVie: Iambic partnered with AbbVie on a multi-year collaboration using AI and automated experimentation to accelerate small-molecule drug development across neuroscience, oncology, and immunology. (link)
Knowtion Health + Ghamut: Knowtion Health partnered with Ghamut to apply AI to complex revenue cycle work, beginning with claims appeals requiring detailed clinical, payer, and regulatory review. (link)
SUNY Downstate + Altera Digital Health: University Hospital at Downstate became the first site to pilot Altera’s Sunrise Thread AI ambient scribe, with 50 providers using it across more than 900 encounters. (link)
GE HealthCare + Mass General Brigham: GE HealthCare and Mass General Brigham expanded their research collaboration to explore generative AI for helping radiation oncology teams access and interpret fragmented patient information. (link)
Xella Health + Oura: Xella Health integrated Oura data into its women’s care platform, combining wearable trends with diagnostic testing and physician-guided treatment to personalize care. (link)
Deal Desk 💰
Spotlight on latest capital raises, M&A, and investments…
FUNDING
Oura, the smart ring maker, is seeking up to $2.2B in its IPO, offering 50M shares at $40–44 each, with nearly two-thirds sold by existing shareholders. (link)
Heidi, an AI care partner for clinicians, raised $340M, a $100M Series C led by Blackbird alongside a $240M growth investment from General Catalyst, valuing the company at $900M. (link)
Enveda, a Boulder, Colo.-based AI drug-discovery company, raised $311M in Series E funding. Catalio Capital Management led the round and was joined by Durable Capital Partners, ICONIQ, Lightspeed, Surveyor Capital, and others. (link)
OpenEvidence, an AI search engine for doctors, raised $250M at a $15B valuation led by a16z, per Business Insider. (link)
Precision Neuroscience, an NYC-based brain-computer interface startup, raised $250M at around a $1B valuation. The Ackman Oxman Institute and Pershing Square led, alongside an undisclosed life sciences fund. (link)
Rightway, an NYC-based pharmacy benefit management and care navigation startup, raised $155M in Series E funding. Francisco Partners led, joined by Thrive Capital and Khosla Ventures. (link)
Biossil, a Toronto-based AI drug-revival startup, raised a $153M financing round led by OpenAI. (link)
Basecamp Research, a Cambridge, Mass.-based developer of AI models for therapeutic designs, raised $140M in Series C funding. S32 led, joined by Anthropic's Anthology Fund, and more. (link)
Biolevate, an AI automation platform for life sciences, raised €30M in Series A funding co-led by RAISE France and Orange Ventures, with participation from MSD Global Health Innovation Fund, Station F and EQT Ventures. (link)
Ramona Optics, an AI-powered microscopy company for biomedical research, raised $25M in Series A funding led by ARCH Venture Partners, with participation from Stealthpoint Ventures, ND Capital, Fall Line Capital and others. (link)
UltraSight, a Boston, Mass.-based developer of AI-guided cardiac imaging software, raised $24M in Series B2 financing. Alive HealthTech Growth Fund led the round. (link)
Basalt Health, a Nashville-based provider of post-acute care software, raised $20M in Series A funding. NEA led, joined by insiders Frist Cressey Ventures and 25m Health. (link)
Iambic Therapeutics, an AI drug discovery company, filed for an IPO and plans to list on Nasdaq under IAM; the offering size has not yet been disclosed. (link)
M&A
DexCare + Mila Health: DexCare acquired Mila Health, extending its AI agents from patient scheduling into care coordination. (link)
Other Relevant News 🔍
News, podcasts, blogs, tweets, resources, etc…
NYC friends, come hang! Register below 👇
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👇
