|
Medical Superintelligence, Anthropic IPO, and 2027 Predictions
By Jason Barry and Virginia Hunt
October 29, 2026
|
|
|
|
|
Together with
|
|
|
|
“In healthcare, a guess isn’t good enough. If agents send a patient to a doctor who doesn’t exist, an appointment that isn’t available, or a provider who’s out of network, the consequences are real. AI may never know it got it wrong, but the patient will pay the price. The front doors to care are increasing exponentially and the industry needs reliable infrastructure to manage it all.”
|
|
Zocdoc CEO Oliver Kharraz, MD
|
|
|
What better way to head into Halloween weekend than confronting one of the scariest parts of bringing a new product to market? Clinical trials can be frighteningly difficult, which is why our very own Jack Troy is teaming up with KoHealth Labs for The Bio Wire’s debut webinar this Friday at 12pm ET to discuss how to streamline patient matching, diagnostics, staffing, and beyond. Grab your spot here.
– Jason Barry, Managing Editor
|
|
|
|
Microsoft just wrapped up a huge week in medical AI. A new paper in Science showed MAI-DxO is more accurate than doctors, Microsoft AI charted its course to medical superintelligence, and the same team published a framework for what that actually looks like.
There’s a long road ahead. The Science paper formalizes SDBench, which turns 304 NEJM cases and 325 MIMIC-IV emergency admissions into step-by-step encounters where the diagnostician asks questions, orders priced tests, and decides when to commit.
- The setup is meant to emulate the day-to-day realities of how doctors operate, which means gradually gathering information by asking a whole lot of questions.
MAI-DxO plays a panel of virtual physicians that debate each case.
- On complex NEJM cases, MAI-DxO scored 83% on diagnostic accuracy, versus 36% for physicians using references (and 22% without).
- On ED cases, MAI-DxO notched 80%, again outshining the physicians at 47%.
- The AI asked 25.5 questions before its first test. Physicians asked 5.6.
That might sound like medical superintelligence. Microsoft AI’s companion post frames the paper as a step toward that AI holy grail, but confirms there’s still a long road ahead.
MAI CEO Mustafa Suleyman predicted medical superintelligence was two to three years away last summer, and is now taking steps to get us there even faster:
- Microsoft is open-sourcing SDBench for select NEJM cases, with the Gatekeeper, Judge, and cost code included so others can test their models against it.
- Microsoft is running a study at Beth Israel Deaconess Medical Center to have patients interact with MAI-DxO directly. That’s the same BIDMC working with Google’s AMIE team.
- Microsoft is folding its health scientists and the Dragon Copilot and PowerScribe teams into Microsoft AI, alongside its Mayo Clinic frontier model collaboration.
Now the fine print, also written by Microsoft. A new BMJ Digital Health & AI paper defines medical superintelligence as surpassing the strongest human-led health systems across the full arc of care, then lists six dimensions that would entail:
- medical knowledge, diagnostic reasoning and triage, therapeutic decisions, longitudinal context, clinical coordination, and alignment with human values.
MAI-DxO only covers about two of the six. The other four involve sensing what patients actually want, treating them over years, and coordinating humans to make it happen
The Takeaway
Microsoft definitely gets credit for publishing its to-do list alongside its victory lap. Diagnosing NEJM cases is a nice box to check, but the remaining boxes are the ones without answer keys.
|
|
Turning Ideas Into Action
Healthcare has never been short on great ideas. What is new is the ability to turn them into reality at the pace patients need. Discover how Bunkerhill Health is closing the distance between knowing and doing, so providers can get back the two things they never have enough of: the time and the clarity to care for people.
|
|
How Clinicians Are Using AI to Find Relief
Clinicians are facing a documentation burden that’s outpaced every attempt to contain it, and they’re adopting AI to tackle the crisis faster than health systems can keep pace. Download Heidi’s new report to see how over 1,800 clinicians around the world are already using AI to find relief, and what that means for the institutions still catching up.
|
|
- Anthropic Files to Go Public: Anthropic filed its S-1 to list in the Nasdaq under ticker symbol ANT, with healthcare and life sciences both flagged as key growth markets. The frontier AI lab already counts eight of the 10 largest healthcare companies and nine of the 10 largest pharmas among its prized customers, and a long list of marquee health systems like Banner, CommonSpirit, and Tampa General are all running Claude. Revenue hit $16.3B in the first half of 2026 (up from $1.2B), while net losses hit an eye-watering $80.7B (not as bad as it looks, mostly a one-off accounting item). The risk factors concede that Claude can still produce inaccurate or harmful outputs, and also included the gem that it “may pose catastrophic or existential risks to humanity.”
- Forrester’s 2027 Crystal Ball: Forrester just delivered our first batch of 2027 healthcare predictions, and they all point one direction: patients are taking care into their own hands. More than half of U.S. consumers will turn to AI tools first, AI agents will handle 20% of routine check-ins like refills and chronic monitoring, one in five Americans will be on a GLP-1, the DTC model will spread to new areas (women’s health, derm, longevity), and 25% of patients will run at-home diagnostics before seeing a provider. On the payer side, Forrester predicts five of the 10 largest payers will roll out deductible-free alternative plans as premiums continue climbing.
- Walmart Opens Healthcare Store: Walmart expanded Better Care Services with a shoppable care catalog powered by General Medicine, the recently very-well-funded healthcare store from PillPack’s founders. Customers can browse urgent care, pediatrics, mental health, nutrition, and weight management from 10+ third-party providers, see estimated out-of-pocket costs, and book visits directly through the platform. General Medicine runs the catalog and pricing, while reportedly also keeping health data walled off from Walmart’s retail business. Walmart has 280M weekly customers, making it one of the best partners you can have to get foot traffic if you just raised $120M to build a healthcare store.
- Glooko Insulin Data: Glooko presented real-world results for EndoTool IV at the Diabetes Technology Society meeting, highlighting the potential of patient-specific IV insulin dosing during pregnancy, The analysis of 67 patients with Type 1, Type 2, or gestational diabetes during labor and delivery across 43 sites found no severe hypoglycemia (under 40 mg/dL) in any patient group, blood glucose under 60 mg/dL in just 0.63% of Type 1 readings and 0.16% of the Type 2 group, and readings above 180 mg/dL stayed under 10% in both. The analysis fits into broader efforts with GestationalCare, Glooko’s new outpatient module for tracking glucose through pregnancy and postpartum.
- Shadow AI Bites Hopkins: Johns Hopkins Bayview is notifying 3,182 patients after an employee uploaded a spreadsheet of patient data to an unapproved public AI tool. The file covered May-to-September admissions and treatment, including names, MRNs, account numbers, medications, and provider names, but no SSNs or financial data. Hopkins had the prompt and file deleted, asked the unnamed AI company to purge it, and is offering 12 months of credit monitoring. This’ll definitely be a good slide for the AI training decks at more systems than just Johns Hopkins.
- Talkiatry Goes 50 for 50: Talkiatry expanded its virtual therapy to all 50 states, pairing 400+ employed therapists with its 800+ full-time psychiatrists on one platform. The in-network model spans 100+ payers and 170M covered lives, has delivered 4M+ visits, and refers for 70+ health systems. With the average behavioral health wait at 48 days and multistate licensure still in a rough spot, Talkiatry is looking to build a coordinated door from routine therapy to higher-acuity care, funded by this year’s $210M Series D.
- Small Model, Big Results: MGB researchers released Clinical System One, a 4B-parameter model that answers questions from patient records. Trained on 12.4M Q&As from 3.6M patients, CS1 beat open models up to 27B parameters on 13 of 14 tasks across 963k held-out patients. The preprint paired CS1 with a generalist LLM to screen patients against 17 clinical trials, surfacing 4x more eligible candidates in its top 20 than keyword search, while using 690x fewer tokens than reading every chart with the big model.
- Sunlight’s $17.4M Seed: The days may be getting darker with daylight savings, but Sunlight AI could brighten things up for independent practices thanks to a $17.4M seed round. The AI-enabled operating platform aims to help physicians prosper in their own practices by automating backend business tasks across payroll and operations, RCM, and tax planning. The solution comes at a pivotal time considering only 18% of physicians remain independent as the pressure to fold-in to larger enterprises with deep pockets grows.
- Athenahealth Assists ASCs: On the topic of assisting independent practices, athenahealth is doubling down on their investment in ambulatory surgery centers with the launch of athenaOne. The new platform connects independent providers to adjacent ASCs in one place with a single patient record, unified RCM, specialized surgery reports, and streamlined coordination via shared documentation. ASCs offer same-day care at significantly lower costs than outpatient hospital departments, and athenahealth is looking to support that mission with plans to launch more ASC capabilities over time.
- Pilot Purgatory, Quantified: A TIAA Institute synthesis of industry surveys found that 52% of health systems had implemented generative AI by late 2025 (up from 25% two years earlier), yet only 35% of AI proofs-of-concept make it to full production. Of 36 use cases studied, radiology is nearly universal at 90% deployed while most remain in pilot. The prescription for escaping pilot purgatory is readiness on four fronts (strategic, technical, workforce, mission) rather than better pilot results.
- Experian Acquires eInsights: Sometimes health coverage only becomes active or discoverable after a patient has received a procedure, so Experian is trying to help identify missed coverage with the acquisition of eInsights. The newly added company already assists in coverage gap detection for 200+ healthcare orgs, and also offers secondary coverage identification, denial prevention, and benefits optimization. Experian put the move in the context of impending Medicaid eligibility changes as Trump’s One Big Beautiful Bill Act goes into effect in 2027.
|
|
Making Healthcare Feel More Human
Abridge has always believed that the best healthcare technology should make healthcare feel more human. That means focusing on the most human signal we have: the conversation. Head over to CEO Shiv Rao’s latest blog to see Abridge’s vision for bringing the broader healthcare ecosystem into the picture, so clinicians can focus more on the practice of medicine and less on the process of medicine.
|
|
Clinician-First Copilot for Value-Based Success
Navina’s AI copilot brings clinical intelligence directly to care teams, turning fragmented data into actionable insights that transform value-based workflows from the back office to the point-of-care. Designed for and loved by physicians, Navina’s Best in KLAS AI reduces missed diagnoses while improving quality metrics and risk adjustment accuracy. Discover how practices are leveraging Navina to enhance VBC performance and improve the clinician experience.
|
|
State of Payor Enrollment and Credentialing
AI is changing the way that healthcare leaders tackle provider network management. Medallion’s latest report breaks down the biggest challenges, emerging trends, and how automation is transforming the landscape. Get the insights you need – read the full report today.
|
|
- A New Path for ACCESS Model Participants: Building your CMS ACCESS program? Withings was accepted into the inaugural cohort as both a participant and a device partner via the ACCESS Tools Directory. Bring the same FDA-cleared, cellular-enabled devices into your program, including BPM Pro 2, capturing clinical measurements and patient-reported outcomes at the point of reading, purpose-built for value-based cardio-kidney-metabolic care.
- Solving Patient Leakage with Virtual-First Care: Patients expect more from their healthcare – easy self-scheduling, ample availability, and appointments that allow them to get high-quality care wherever and whenever they need it. If your health system isn’t offering that, someone else is. Download the new white paper from Ovatient to see how health systems can look beyond traditional telehealth solutions to deploy an EHR-integrated, virtual-first care model that expands patient access and choice, builds long-lasting care relationships, and reduces leakage.
|
|
|
|
|
