Healthcare AI Guy Weekly | 8/19

Subscribe | Advertise | AI Startup Database | Feature AI Jobs

Our Picks ✨

Highlights if you’ve only got 2 minutes…

1/

AI scribes expand beyond the note

There has been a lot of movement in the AI scribe space with Epic and Doximity recently launching their own offerings, but leaders like Abridge are already moving beyond the note. Within a day of Epic’s scribe leak, Abridge announced real-time prior authorization at the point of conversation. Co-developed with Highmark Health, the workflow listens alongside the ambient note, surfaces payer criteria inline, fills forms, submits requests, and tracks status, with clinicians approving before anything is sent. The aim is to cut approvals from weeks to minutes and move decisions into the exam room. Strategically, Abridge is running a land and expand play: start with documentation, then automate the high-friction admin rails where its conversational context gives it an edge. That shifts Abridge from note maker to workflow platform, beginning with prior auth and extending to claims, care coordination, and decision support. Nabla is making a similar turn with an agentic assistant across ambient, dictation, coding, and EHR actions. The moat here is end-to-end workflow, not transcription. (link)(linkedin)

2/

OpenEvidence scores 100% on USMLE

OpenEvidence's latest model just scored a perfect 100% on the USMLE. It's packaging that skill with a free model that explains every answer step-by-step, designed to help med students prep with gold-standard sources like NEJM and JAMA. For context, physicians average roughly 70% without AI. Still, headline scores can mislead: current systems excel at pattern recognition and exam heuristics, not genuine medical reasoning, and founder Daniel Nadler notes a USMLE score isn’t the best clinical yardstick. Even so, the arc is clear, AI is getting very good at diagnosis, and the long-term promise is better, more cost-effective care. Explanations and evidence, not just accuracy, are the differentiator.

That said, there are hidden risks. In a Lancet trial, unassisted polyp detection fell from 28.4% to 22.4% after AI was withdrawn from endoscopy centers, and an MIT study found AI users recalled less and engaged their brains less when writing. The practical path forward is a human–AI co-pilot: keep clinicians in the loop, audit skills, and design workflows where model explanations augment judgment so outcomes improve without eroding the craft. (link)(tweet)

3/

AI slop in healthcare

As AI progresses and adoption grows, so does “AI slop.” That’s the confident, synthetic output that looks helpful but isn’t: hallucinated anatomy, made-up facts, and answers with shaky grounding. In healthcare, that stuff can creep into decision support, patient messaging, and even research, eroding trust and risking harm. Here are 3 recent stories that highlight this:

• Google Med-Gemini’s “basilar ganglia” error: A Google model conflated basal ganglia with the basilar artery. The invented anatomy passed through a paper and blog post before edits, a reminder that authoritative-looking slop can be missed. (link)

• Fake-term injection spikes errors: Adding just one fake term to a medical prompt caused six popular LLMs (ChatGPT, DeepSeek, etc) to produce convincing but false answers in 50–82% of 300 test vignettes. (link)

• Mount Sinai study: In fictional patient scenarios with made-up terms, popular chatbots repeated and expanded the fakes. Adding a one-sentence caution cut errors significantly, showing how minimal guardrails and prompt design can improve safety. (link)

This matters as health systems are adopting AI quickly, but lag on governance. Per a recent report, 88% use AI internally, yet only 17% report mature strategies. In healthcare, we should remember to treat AI as a co-pilot, not an autopilot: require human review for clinical use, ground answers in vetted sources, add caution prompts and retrieval, and pressure-test models with fabricated inputs before scaling.

Tools & Partnerships 🔧

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

TOOLS

• Athenahealth debuts AI-native upgrades for ambulatory care: Athenahealth is rolling out new AI features across its athenaOne platform, adding AI-powered document services, clinical summaries, interoperability, and revenue cycle tools for 160K+ providers. (link)

• Oracle Health unveils AI-powered EHR for ambulatory care: Oracle launched a cloud-native EHR with voice commands and AI agents for workflow automation and insights. Built separately from Cerner, it integrates via a semantic AI layer and rolls out to ambulatory sites first. (link)

• Google & NASA build AI “doctor” for space missions: The Crew Medical Officer Digital Assistant diagnoses and treats astronauts when Earth contact is delayed, with plans to expand into ultrasound and biometrics, and possible telemedicine use on Earth. (link)

• Oscar Health builds AI agent for care navigation: The insurer developed an agentic AI system to assist Care Guides with tasks like prior auth, benefits, and Rx renewals. Early pilots earned 82% satisfaction, and Oscar is testing GPT-5 to expand features like voice interaction. (link)

• Stedi debuts AI-native tools for clearinghouse workflows: The startup launched Stedi Agent and a new MCP server to automate processes like eligibility checks. Just 18 months old, Stedi says GenAI companies now make up a third of its customer base. (link)

• Epic AI helps detect early-stage lung cancer at Christ Hospital: A new Epic-integrated AI tool flagged early signs of lung cancer in a 64-year-old patient, enabling less invasive treatment. Clinicians say the tech is boosting early detection and survival odds. (link)

• Amazon bets on neurosymbolic AI to curb hallucinations: Amazon unveiled Automated Reasoning, a neurosymbolic AI feature combining neural networks with symbolic logic to cut hallucinations. Claimed 99% accurate, it could help authenticate AI outputs in healthcare, per the WSJ. (link)

• MIT uses AI to design new antibiotics for superbugs: Researchers developed two novel compounds, NG1 and DN1, that killed drug-resistant gonorrhea and MRSA in mice. The AI screened 36M molecules, pointing to a “second golden age” for antibiotic discovery. (link)

• Boston Children’s launches AI lab for pediatric heart care: The new Congenital Heart AI Lab will develop tools like AI-powered ECGs, using six decades of heart data to improve diagnosis and management of congenital heart disease. (link)

• Xsolis launches Dragonfly Navigate for discharge planning: The new AI tool integrates length-of-stay analytics into clinical workflows, helping hospitals optimize bed capacity. West Tennessee Healthcare is the first pilot site. (link)

• South Korean researchers build AI to predict malpractice risk: Asan Medical Center developed Dr. Lawn, an algorithm trained on 15k+ legal rulings that summarizes medico-legal risks tied to symptoms, diagnoses, and procedures, offering clinicians precedent-based guidance. (link)

• Cognizant launches TriZetto AI Gateway for payers: The company introduced a governed AI layer across its TriZetto portfolio, embedding AI into claims, care management, and customer service to streamline payer workflows. (link)

• Korean researchers develop AI for cancer drug design: KAIST’s BInD model creates targeted cancer drugs from scratch without prior molecular data, optimizing for safety, stability, and manufacturability in a single step. (link)

• Mount Sinai finds AI improves ED triage efficiency: Researchers trained an AI model on 1M+ prior visits and found it predicted hospital admissions more accurately and earlier than nurses across 50k patient encounters. The system could help reduce overcrowding and improve ED flow. (link)

PARTNERSHIPS

• Stanford Health Care + Qualtrics: Stanford is expanding its partnership with Qualtrics to build AI agents that improve patient experience by predicting missed appointments, resolving care barriers, and easing coordination for care teams. (link)

• Sutter Health + GE HealthCare: Sutter Health is rolling out AI-powered imaging upgrades across 300+ facilities under a seven-year partnership with GE, modernizing its fleet, shortening scan times, and expanding patient access. (link)

• Cleveland Clinic + Piramidal: Cleveland Clinic is piloting Piramidal’s neural network model that mimics brain signals to monitor patients with critical neurological injuries, bringing AI into real-time ICU care. (link)

• HealthEx + CLEAR + athenahealth + MedAllies + CommonWell: HealthEx launched a platform for real-time patient access to complete health records, using CLEAR for ID verification and connecting through athenahealth, MedAllies, and CommonWell under TEFCA to enable secure, patient-directed data sharing. (link)

• Epic + Ambience Healthcare: Ambience was added to Epic’s Toolbox program for ambient AI, expanding access to its documentation, coding, and workflow platform. The integration brings Ambience’s tools natively into Epic’s Haiku app and Ambient Module. (link)

• Meditech + AvoMD: Meditech is embedding AvoMD’s AI scribe into its Expanse EHR, bringing ambient documentation, coding suggestions, and clinical decision support to rural and community hospitals. (link)

• iRhythm + Lucem Health: iRhythm is partnering with Lucem Health to integrate AI-driven arrhythmia risk detection into population health and value-based care programs, routing flagged patients to Zio monitoring for intervention. (link)

Deal Desk 💸 

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

FUNDING

• Merge Labs, a brain-computer interface startup, is reportedly raising at an $850M valuation, with Sam Altman co-founding and OpenAI investing. (link)

• Stedi, an AI-enabled clearing house, just raised a $70M Series B fundraise, co-led by Stripe and Addition, with participation from USV, First Round, Bloomberg Beta, BoxGroup, Ribbit Capital, and others. (link)

• Citizen Health, a rare disease data advocacy startup, raised $30M in Series A funding. 8VC led, joined by Headline and Transformation Capital Partners. (link)

• Tahoe Therapeutics, an SF-based developer of a foundational dataset for training "virtual cell models," raised $30M. Amplify Partners led, joined by Databricks Ventures, Wing Venture Capital, General Catalyst, Civilization Ventures, Conviction, Mubadala Capital Ventures, and AIX Ventures. (link)

• Protege, an NYC-based developer of data sets designed for AI training, raised $25M in Series A funding. Footwork led the round and was joined by CRV, Bloomberg Beta, Flex Capital, Liquid 2 Ventures, Shaper Capital, and others. (link)

• Arintra, an Austin, Texas-based AI-powered medical coding automation platform, raised $21M in Series A funding. Peak XV Partners led the round and was joined by Endeavor Health Ventures, Y Combinator, and others. (link)

• Fountain Life, a longevity clinic chain co-founded by Tony Robbins and Peter Diamandis, just raised a $18M Series B. (link)

• Isaac Health, a NYC-based health technology company for brain health and dementia care, raised $10.5M in Series A funding. Flare Capital led the round and was joined by others. (link)

• Better Medicine, a Tartu, Estonia-based developer of AI software designed to improve kidney cancer scans, raised $1.16M in pre-seed funding. (link)

MERGERS & ACQUISTIONS

• Abridge: With 80% of its cash reserved for doubling down on its tech, Abridge is earmarking the remaining 20% for potential acquisitions. (link)

• Sword Health: Sword announced they are actively acquiring healthcare teams & companies across the US, UK, and Europe. (link)

Other Relevant News 🔍

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

• Bessemer: the state of AI 2025 (link)

• Inside a 69-hospital system’s AI strategy (link)

• Whole-body MRI AI shows mixed disease prediction results (link)

• Rural hospitals lag in AI adoption, study finds (link)

• AI companion apps on track to pull in $120M in 2025 (link)

AI Job Opportunities 💼 

Contact us to feature roles in our newsletter…

• AI Research, Health, Clinical Specialist at Google, a large tech company
  $146 - 214K | Mountain View, CA (link)

• Senior Product Engineer at Slingshot AI, an AI therapy provider

  $90 - $200K | NYC (link)

• B2B Growth Marketing Manager at Keragon, a healthcare automation platform

  $NA | Remote (link)

Visuals of the Week 📸

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