Healthcare AI Guy Weekly | 10/7

 Subscribe | Advertise | AI Startup Database | Feature AI Jobs

Our Picks ✨

Highlights if you’ve only got 2 minutes…

1/

Radiology’s last exam

AI was supposed to make radiologists obsolete. Nearly a decade after Geoffrey Hinton’s famous prediction to “stop training radiologists,” the opposite has happened. Radiology residencies are at record highs, salaries have climbed above $500,000, and demand for specialists keeps growing. AI models like CheXNet and those from Annalise.ai or Aidoc can outperform humans on narrow imaging benchmarks, yet they still struggle in real-world hospital settings. That was reinforced by the latest radiology board exam results: expert radiologists scored 83%, trainees 45%, and GPT-5 just 30%. Even top-tier models misread CTs and X-rays, fell into reasoning traps, and slowed down under pressure.

Most systems are limited to narrow use cases, don’t generalize well across hospitals, and face legal and insurance barriers to autonomy. Even when they work, AI only tackles a small slice of a radiologist’s job, much of which involves communication, oversight, and patient care. That said, AI progress is accelerating fast and we expect that gap to close in the next years. The best path forward remains augmentation, not replacement. (link)(tweet)

2/

Trump to big healthcare: arm the rebels

The Trump administration is taking a harder line on healthcare AI oversight, breaking from the Biden-era approach that supported public-private collaboration. This week, HHS Deputy Secretary Jim O’Neill said the department does not endorse the Coalition for Health AI (CHAI) - a group backed by Microsoft, OpenAI, the Mayo Clinic, and Duke Health - calling it a potential “cartel” that could pressure startups to join. The coalition, which has 3,000 members and recently partnered with the Joint Commission on AI guidance, maintains that its work is voluntary and meant to inform policy, not regulate it.

The move signals a shift toward a more laissez-faire AI strategy focused on transparency and data sharing rather than private-sector standard setting. The FDA, meanwhile, is seeking public input on how to measure AI performance as regulators debate how, and who, should oversee medical AI. It seems Washington’s message to health AI giants is “arm the rebels.” (link)

3/

New Mercor benchmark shows gaps

Mercor, an AI labeling company that has grown from $1M to $500M in just 17 months, has released the AI Productivity Index (APEX), a new benchmark testing how AI performs real work across law, finance, consulting, and medicine. The General Practitioner (MD) benchmark was developed with experts from the University of Pennsylvania, Northwestern, Cornell, Brigham and Women’s, and Mount Sinai, and advised by Dr. Eric Topol. The medical benchmark focuses on whether AI can think like a clinician, not just pass exams.

GPT-5 ranked first, followed by Grok 4 and Gemini 2.5 Flash. While these models can ace the USMLE, most struggled on real patient cases, with scores averaging around ~50%. None meeting the standard for real-world autonomy. The results show that AI can help with diagnostics and documentation, but still falls short of independent clinical judgment. In line with other stories we’ve reported on in the last weeks, medicine, at least for now, still needs a human in the loop. (link)(tweet)

Tools & Partnerships 🔧

Latest on business, consumer, and clinical healthcare AI tools and partnerships…

TOOLS

• AI predicts 1,000+ diseases with GPT model: A Nature study introduced Delphi-2M, a GPT model trained on 400K UK patients to forecast diagnoses and time-to-disease across 1,000+ conditions. It matched or beat leading single-disease models, and outperformed biomarker-based models for heart attack and stroke. (link)

• AI scribes linked to drop in burnout, study finds: A JAMA study of 263 providers across six health systems found that using ambient AI scribes reduced burnout from 51.9% to 38.8% in just 30 days, while also improving after-hours workload, task load, and patient focus. (link)

• General Catalyst launches Percepta for enterprise AI transformation: Percepta is a new company embedding AI teams into enterprises to modernize operations. Backed by AWS and Anthropic, Percepta will focus on AI-driven transformation across healthcare, finance, and government, starting with clinical workflows and patient engagement. (link)

• Ambience launches inpatient coding AI: Ambience rolled out a GPT-5-powered assistant to help hospitalists improve ICD-10 documentation and reduce billing errors. The CDI tool is live at 40+ health systems and designed for compliance and clinical nuance. (link)

• Qventus launches AI Solution Factory for health systems: Qventus unveiled an AI “Solution Factory” to help hospitals build their own agentic AI workforce. Core agent capabilities include chart mining, risk analysis, care gap coordination, document handling, and patient engagement. (link)

• Omada debuts Meal Map for AI-driven nutrition support: Omada launched Meal Map, a new AI-powered nutrition tool that offers real-time food feedback and guidance from care teams. The goal is to encourage sustainable eating habits without strict tracking. (link)

• Mayo Clinic launches Platform_Orchestrate: Mayo debuted a new platform offering biopharma and medtech firms access to de-identified data, clinical expertise, and its global network to speed discovery, trials, and deployment. (link)

• Cognixion pairs BCI with Apple Vision Pro for thought-powered speech: Startup Cognixion is integrating its EEG-based brain-computer interface with Apple’s Vision Pro to enable patients with paralysis to communicate using only their thoughts. The noninvasive system combines neural signals and personalized AI to support speech without physical movement. (link)

• Cancer AI Alliance launches secure research platform: Dana-Farber, MSK, Fred Hutch, and others formed the Cancer AI Alliance to train AI models on 1M+ patient records without moving data outside hospital firewalls. Backed by AWS, Google, Microsoft, and others, the platform supports collaborative research while maintaining strict privacy standards. (link)

• FDA seeks input on evaluating AI in real-world care: The FDA is requesting public comment on how to assess AI tools outside of clinical trials, citing gaps between study results and real-world outcomes. The issue is expected to be discussed at the upcoming Digital Health Advisory Committee meeting on Nov. 6. (link)

PARTNERSHIPS

• Memorial Sloan Kettering + ALIGNMT AI: The cancer center is piloting ALIGNMT AI’s governance platform to integrate real-time risk monitoring and lifecycle management for clinical AI, aiming to operationalize responsible AI deployment across oncology workflows. (link)

• Ascension + Oracle: Ascension is expanding its collaboration with Oracle to connect its EHR and ERP systems, using AI and analytics to reduce inefficiencies and lower operational costs across its 121 hospitals. (link)

• Duke Health + Trase Systems: Duke Health is collaborating with Trase Systems to develop agentic AI tools for administrative automation, patient scheduling, care coordination, and clinical research support. (link)

• Northeast Georgia Health System + Microsoft DAX: The health system deployed DAX Copilot to 480 clinicians, integrating AI-powered documentation and ambient orders into Epic to reduce screen time and ease administrative burden. (link)

• Seattle Children’s + Abridge: The pediatric health system is expanding Abridge’s ambient AI platform to 18 specialties after a successful pilot that cut documentation effort by 79% and supported 72% of eligible encounters. The rollout may extend to Abridge’s diagnosis intelligence and inpatient tools. (link)

• Hackensack Meridian Health + SpotitEarly: The NJ-based health system is partnering with the startup to validate its AI-powered breath test for early cancer detection, which combines trained dogs’ scent recognition with machine learning. (link)

• Bethany Children’s Health Center + Commure: The Oklahoma pediatric specialty hospital adopted Commure’s ambient documentation solution to automate note-taking and streamline clinician workflows. (link)

• Vanderbilt University Medical Center + Latent: The academic medical center selected the AI-powered prior authorization platform to accelerate medication approvals and improve patient affordability workflows. (link)

• Allina Health + AssistIQ: The Minnesota health system is deploying AI-driven supply capture across 100+ OR and procedural rooms to improve inventory accuracy and reduce waste. (link)

Deal Desk 💸 

Spotlight on latest capital raises, M&A, and investments…

FUNDING

• Periodic Labs, an "AI scientist" startup, raised $300M in seed funding. A16z led, joined by a16z, DST, Nvidia, Accel, Elad Gil, Jeff Dean, Eric Schmidt, and Jeff Bezos. (link)

• Assort Health, a patient communications automation startup, raised $76M in Series B funding. Lightspeed led, joined by First Round, Chemistry, Felicis, A*, Liquid 2 Ventures, and Quiet Capital. (link)

• Heidi Health, a New York-based medical scribe, raised $65M in Series B funding. Point72 led, joined by Goodwater Capital, Headline, Blackbird VC, LG Technology Ventures, and Alumni Ventures. (link)

• Latent Health, a startup automating pharmacy workflows, is raising a $50M Series B. (link)

• Midi Health, a women’s health virtual care platform, has raised $50M in a Series C round led by Advance Venture Partners. The funds will be used to bet on longevity and AI. (link)

• Oath Surgical, AI-based surgical centers, raised $24M in Series A financing led by FPV Ventures. The company also received a new investment from McKesson Ventures. (link)

• Confido Health, a developer of healthcare scheduling agents, raised $10M in Series A funding. Blume Ventures led, joined by Schema Ventures, Vicus Ventures, Together Fund, DeVC, and Medmountain Ventures. (link)

• BeSound, a breast cancer imaging startup, raised $6.8M in seed funding from Overwater Ventures, Kindred Ventures, Muse Capital, Lux Capital, Wisdom.VC, Retron VC, Mana Ventures, Pave Health Ventures, and SSC Venture Partners. (link)

• Praxipal, a German AI receptionist for health care, raised $6.7M in seed funding. HV Capital led, joined by Nebular, Anamcara Capital, HPI Ventures and Angel Invest. (link)

• DataJoint, a Houston-based computational database for life sciences, raised $4.9M in seed funding led by Nina Capital, Inoca Capital Partners, and Capital Factory. (link)

MERGERS & ACQUISITIONS

• Smarter Technologies + Pieces: Smarter adds Pieces’ ambient scribing to launch SmarterNotes, embedding RCM intelligence into notes. The move follows its merger of SmarterDx, Thoughtful.ai, and Access Healthcare. (link)

• Qualtrics + Press Ganey: Qualtrics, an AI-powered customer engagement company, agrees to acquire healthcare market research firm Press Ganey for $6.75B. The deal gives Qualtrics a major foothold in healthcare, adding Press Ganey’s network of 41,000+ provider clients and expanding its AI capabilities into patient experience analytics. (link)

• Talkspace + Wisdo Health: Talkspace adds Wisdo, an AI-powered platform for coaching and peer-to-peer support, to expand its mental health offerings. The acquisition adds 24/7 peer support and group coaching to Talkspace’s telehealth services. (link)

• Switchboard + Conduce: Specialty care marketplace Switchboard acquired Conduce, an AI-driven tool for patient-provider matching, to enhance precision and efficiency in connecting patients with the right specialists. (link)

• Wysa + Kins: Wysa acquires Kins, a virtual and at-home physical therapy provider, expanding its human-centric care model beyond mental health. (link)

GRANTS

• White House pledges $100M to pediatric cancer data initiative: The AI-powered program, launched in Trump’s first term, aims to collect and analyze childhood cancer data to drive better diagnostics and treatments. (link)

Other Relevant News 🔍

News, podcasts, blogs, tweets, resources, etc…

• Experian’s 2025 claims report found providers’ claim denials up 10% (link)

• California enacts 1st US law requiring AI safety disclosures (link)

• Fortune's AI readiness list favors insurers, overlooks hospitals (link)

• Health AI leaders split on utility of AI regulatory sandboxes (link)

• Bezos says the AI boom is a bubble—and that’s fine (link)

Visuals of the Week 📸

Funny memes, cool pics, and interesting data from around the web…