Hi HAIFer - before you dive into this week’s market intelligence, a quick observation I noticed this week.
I shared my view on what OpenAI is likely building toward back in October, and again four days ago with the full system map. The core thesis is that OpenAI would eventually connect patients and providers through a single model infrastructure, positioning itself as a potential gateway to care. That framing is now showing up much more widely across LinkedIn posts and conversations now. One post even repeated it word-for-word without attribution.
It’s not an isolated incident either. I’ve noticed this pattern a few times.
That won’t change how I publish, though.
As a subscriber, just know this: you’ll hear it here weeks and months before it becomes mainstream. It’s why I called this Health AI Foresight.
Your shortcut for this week:
Read on for the foresight of the week and the full breakdown of what shaped healthcare AI 👇
Solutions and Launches
AstroDoc (US): Launched ASTRID to solve the "Last Mile" problem. Defined as "AI that connects directly to actual care delivery," it operates as a HIPAA-compliant covered entity to provide direct physician access, specialist connections, and lab orders. [Link]
Governance, Policy, and Ethics
FDA (US) & EMA (EU): Issued ten joint principles for "Good AI Practice" to support evidence generation in medicine development. The publication guides AI use across the lifecycle and marks renewed EU-US cooperation. [Link]
Royal College of Physicians (UK): Warned that 68% of doctors feel the NHS lacks the IT infrastructure to safely deploy AI, despite high support for the technology. [Link]
NHS England (UK): Launched the "AVT Supplier Registry" to standardize AI scribe adoption, backed by data showing a 23.5% increase in patient interaction time and 13.4% more patients seen per A&E shift. [Link]
World Economic Forum (Global): Released "A New Era for Digital Health," arguing digitization is insufficient. It profiles Abu Dhabi as a blueprint for "intelligence-driven" systems using real-time AI to deliver personalized care at scale. [Link]
Research and AI Advancements
Aignostics (Germany/US): Released Atlas 2, a 2-billion parameter pathology foundation model trained on 5 million slides. It achieved state-of-the-art performance across 80 public benchmarks and includes clinical-grade regulatory documentation, enabling deployment for tasks like tumour microenvironment profiling. [Link]
ARISE (Stanford/Harvard): Released the "State of Clinical AI Report 2026," identifying a "jagged frontier" where models excel at reasoning but struggle with uncertainty. [Link]
Google Research (US): Released MedGemma 1.5 and MedASR, open models updated to interpret 3D scans (CT/MRI) and clinical dictation. [Link]
Google Research (US): Validated that consumer smartwatches can estimate spatio-temporal gait metrics (e.g., speed, step length) with accuracy comparable to lab equipment. This suggests neurological tracking can move from expensive gait labs to continuous background monitoring. [Link]
Ambient Artificial Intelligence Scribes and Physician Financial Productivity: A UCSF study found that AI scribes increased physician productivity by 1.81 RVUs (a standard metric used to calculate physician fees) per week. [Link]
Detecting sleep stages after stroke: Training AI on mixed datasets improved sleep staging for stroke patients, yet accuracy remained low. [Link]
Machine learning reveals country-specific drivers of global cancer outcomes: Analysis of 185 countries identified Universal Health Coverage as the primary survival driver in Brazil, whereas GDP per capita drives outcomes in the US. This suggests policy structure impacts survival as much as economic output. [Link]
Partnerships & Adoption
BillionToOne & Epic (US): Partnered to integrate prenatal and oncology testing directly into Epic’s Aura suite. This allows clinicians to order tests within the EHR, with results flowing back for immediate decision support. [Link]
OpenAI & Cerebras (US): Partnered to integrate 750MW of ultra-low-latency compute into OpenAI's inference stack. [Link]
Ultrahuman & Click Therapeutics (Global): Partnered to launch the "Migraine PowerPlug," embedding FDA-authorized digital therapeutic logic directly into a consumer wearable platform. [Link]
Konica Minolta & deepc (Global): Integrated the deepcOS AI platform into Konica’s radiology workflow. [Link]
Bets on the Next Health System
Investments:
Merge Labs (US): Raised $252M (Seed), OpenAI also invested, to develop non-invasive brain-computer interfaces. [Link]
Proxima (US): Secured $80M to design "molecular glues" for undruggable targets. [Link]
Vista AI (US): Raised $29.5M to automate MRI scanning with backing from major health systems. [Link]
M&A:
AstraZeneca (UK): Acquired Modella AI to embed agentic AI into its oncology pipeline. [Link]
That’s a wrap for Edition #15 of Health AI Foresight.
My goal with this newsletter is simple - to connect the present, the emerging, and the future of healthcare AI.
See you next week.
- Dr. Aboufandi
📩 Something stand out? Or have a suggestion? Hit reply - I read every message.