Magicare.AI

@ACNAPPresident Strong findings! The predictors they surfaced (COPD, valve issues, iron deficiency) underscore how upstream clinical clarity during transitions matters enormously. Better signal at intake and handoff helps teams focus resources where they prevent the bounce-backs that hurt patients and margins. Integrated pathways like this are exactly where post-acute AI can add the most leverage when it surfaces the right priorities.

Big congrats to the Prosper AI team, healthcare ops automation is long overdue. The full picture also includes the upstream clinical intake layer. Turning messy, multi-portal referrals into structured, clinically reliable decisions in seconds is another massive lever for reducing burnout and speeding up care. Excited to see more agentic systems tackling these fragmented workflows end-to-end.@Minh_Q_Tran

Prosper AI lands $30M funding led by Andreessen Horowitz to accelerate end-to-end automation of healthcare operations. Read more: buff.ly/9nUOafg #healthtech

At Magicare we see the same principle on the admissions side: surface the clearest clinical signal from messy referrals so teams can direct attention where it matters most, without adding extra work or cognitive load. RESPECT you building a SNOS from real floor experience, Adina. The post-acute space needs more of this. Excited to follow the journey. 🩺"

Building SNOS (Space Nursing Operations System) in public with @grok. An AI-assisted clinical decision support platform built from real nursing experience. Core philosophy: AI proposes. Nurses decide. Starting with skilled nursing, rehab, and memory care. Long-term vision: healthcare on Earth, Mars, and beyond.

Couldn’t have said it better ourselves ❤️ Patient data still lives behind dozens of pre-API portals. So we built agents that don’t just fetch data, they reason. Hundreds of browser agents run per case on @browserbase, turning 60-minute manual reviews into 60-second clinical decisions: 📈 200,000+ referrals 📄 2M+ documents 💬 500,000+ messages All so admissions teams can focus on patients, not paperwork. Thank you for helping us make this possible! @browserbase

Exactly. @TheBigUnlock AI should protect the clinician-patient relationship, not get in the way of it.

Spot on. The scariest part of clinical LLMs is how confidently they echo bad or incomplete input. We designed our system to do the reverse: dig through the actual referral PDFs in legacy portals, hunt for contradictions, and give admissions teams grounded signal instead of amplified noise. Curious how you’re thinking about guardrails here @RoupenMD?

A new study found that the same input that makes a medical AI smarter, a clinician talking to it, is also what makes it dangerous. Across 8 frontier models on real differential-diagnosis tasks: when an expert clinician added context, agreement on the top diagnoses jumped from 66% to 94%. The tool genuinely sharpens when a good doctor engages it. Then they fed the models a confident but wrong clinician steer. 14 of them degraded. They didn't resist. They rationalized the bad idea and echoed it back, at times to recommendations the authors graded at the "death" harm tier. The uncomfortable truth for anyone using these in clinic: the model can't tell my sharp day from my anchored one. It mirrors confidence, not correctness. On the morning I've locked onto the wrong diagnosis, the AI most eager to "agree with the physician" is the one most likely to hurt the patient. We keep grading these models on solo accuracy. But that's not how they're used. They're used mid-conversation, by a tired human who already has a hunch. The benchmark that matters is how well a model pushes back when I'm wrong, and almost no one is testing for that. Should a clinical AI's most important score be its accuracy, or its willingness to disagree with me?

Vertical AI at its best removes the soul-crushing drudgery so clinicians can actually move faster. We built precisely this for Skilled Nursing: inference across legacy portals + unstructured PDFs → structured, contradiction-flagged output. No more 12 logins and hundreds of pages just to decide on one admission. Strong points. @mauricebeckand

Precisely. Most clinical “hallucinations” are just noisy upstream data wearing a confident mask. Magicare’s inference engine doesn’t generate from vibes, it extracts 500+ validated points from the actual messy PDFs in legacy portals, surfaces contradictions in real time, and only then supports a decision. Garbage in, structured clinical truth out.

AI doesn't hallucinate randomly, it hallucinates when the underlying clinical data is wrong. @medicompdoc breaks down the hidden cost of bad data: wrong diagnoses, denied claims & what health systems must get right before trusting AI. 🎙️@AkrosAdvisory: akrosadvisory.com/podcast/episod…

Spot on, Rachel. New portals just add another login and a PDF graveyard. Magicare attacks the actual bottleneck: we run clinical inference across messy Epic/WellSky docs, pull 500+ structured points, and give post-acute teams a 60-second decision instead of a 60-min manual slog.

The practice I left probably thought they modernized when they migrated to a new portal. They added a channel. They didn't fix access. Adding tech without removing broken steps adds cost. Process fixes are the foundation. hfi.consulting/articles/when-… #HealthcareAI #HealthcareFinance

If an AI cannot explain its reasoning, it has no place in clinical decision-making. We must move from blind trust to verifiable transparency. Read the full breakdown by @adammoisa in the link above. 👇

In healthcare, "black box" AI outputs are not just unhelpful, they are dangerous. True clinical AI requires reasoning that is traceable, monitorable, and open to continuous reflection. Our CTO @adammoisa unpacks the exact blueprints for building this level of clinical trust here: magicare.ai/articles/what-… #HealthTech #ClinicalAI #HealthcareIT

Thanks for sharing! Positive step from CMS on tighter budget neutrality for Medicaid Section 1115 waivers- good for long-term sustainability. Stronger fiscal guardrails help sustain community-based models while ensuring SNFs can focus on the higher-acuity patients who need post-acute intensity. In the post-acute world, this reinforces the value of smart tools for seamless transitions and cost-effective outcomes.

New federal guidance aimed at placing tighter #budgetneutrality requirements on #Medicaid Section 1115 waiver programs is being called a “constructive step” toward strengthening the program’s sustainability by the #seniorliving industry. tinyurl.com/4y7r5s66

Strongly agree @ahcancal these OIG findings are deeply concerning. This delays care, prolongs hospital stays, and creates massive admin burdens for providers. Agentic AI that strengthens real-time clinical + financial decisioning at intake can help build stronger, defensible cases upfront, reducing inappropriate denials/delays while ensuring the right patients get timely access. Patients and providers both deserve better systems.

#ICYMI: A new OIG report found that the largest Medicare Advantage Organizations are denying skilled nursing and other post-acute care requests at alarmingly high rates. Our nation’s seniors deserve better. Congress must act. Read more: brnw.ch/21x3mL9

@NedOLutz We see this every day from the admissions side. When MA delays or denies appropriate post‑acute care, hospitals keep patients longer and SNFs juggle last‑minute referrals under pressure.

Medicare Advantage denied 4.1M prior-auth requests in 2024. 80.7% of the ones patients appealed got reversed. Only 11.5% got appealed. Read that again. The denial works because you won't fight it. The Machine prices in your exhaustion. (KFF, 2024)

The best technology doesn’t add steps. It removes them.

@HealthcareAIGuy @nvidia @NVIDIAHealth @AbridgeHQ Glad to see AI‑native platforms move from notes into CDS, workflow and payment. The next frontier: connecting that visit intelligence to the rest of the journey, especially post‑acute.

NEW: Abridge unveiled an AI-native clinician intelligence platform and announced a collaboration with NVIDIA to build a foundation model for clinical conversations The platform expands beyond documentation into CDS, workflow automation, coding, payment, & care coordination

@ahcancal @OIGatHHS MA delays in medically necessary post-acute placements are not just ‘administrative issues’, they’re care issues. Seniors deserve timely, appropriate post-acute care. Hospitals and SNFs deserve workflows that make approvals clearer, faster, and more consistent.

🚨 New @OIGatHHS confirms what skilled nursing has been saying for years: #MedicareAdvantage plans too often deny and delay access to post-acute care. Seniors deserve better. Congress must hold MA orgs accountable. Read the report: brnw.ch/21x3hkD

What does Magicare actually do for a post-acute team? For one org: referral review time dropped from 38 min to 3 min in less than 2 months. SPEED That's real time back for the decisions that matter most. 🏥

@nikillinit Respect!

full post is here - we can solve this problem! outofpocket.health/p/fake-doctors…

Fake doctors online are a growing problem and I think I know how to solve it In the last couple of years, AI has made it easier than ever to fake being a doctor and social media amplifies it This stems from the fact that we don't have a digital identity layer than can actually verify if a person that says they're a doctor has a credential In today's post, we walk through - Some specific scams that are using fake doctors - Suggestions on a system that can fix it - Startup ideas that could exist only if you had a digital ID layer full post in the next tweet

38 minutes to 3 minutes. 13X faster! That's how fast one organization cut referral review time after going live with Magicare. Same team. Same volume. Smarter clinical inference.