The Truth About Lying to Your Doctor
with Stephen Rouse, Savva
The Truth About Lying to Your Doctor
Show Notes
Stephen Rouse didn't set out to pick a fight with Google, OpenAI, Amazon, and Microsoft. He just noticed something broken. Every founder in his orbit was tracking their body through a circus of apps that refused to speak to each other. A Whoop on the wrist. A Garmin for skiing. MyFitnessPal for food. Epic MyChart for labs. Strava for runs. Six logins, zero clarity.
Meanwhile, the 21st Century Cures Act had quietly opened the door: third parties could now legally pull patient medical records directly from hospital EHRs. Stephen and his co-founder Amit Shah had already spent years building exactly that infrastructure at Protocol First, which was acquired by Roche Pharma via Flatiron Health after becoming the first FHIR app to extract patient health data from Epic hospitals for FDA clinical trial submissions.
So they built Savva. A unified health intelligence layer that pulls in your medical records, wearables, labs, and meds, then lets you run them through Claude, GPT, Gemini, Grok, Llama, and more like a round table of second opinions. For ten dollars a year. Stored locally on your device. Not sold to insurers. Not uploaded to a cloud that gets monetized in a bad quarter.
The philosophical core of the episode is trust. Stephen argues that people lie to their doctors because the incentives are broken. Admit you smoke a cigar on the golf course and your life insurance premium jumps $300 a month. Admit you had seven vodka sodas last night and it lives on a clipboard forever. But you'll tell the AI. Because the AI already has the data, doesn't judge you, and isn't reporting back to your payer.
Frameworks from This Episode
The Unidentified Data Principle
Most apps say encrypted, in transit, at rest, de-identified. Stephen goes one step further with an architecture where the business model physically cannot pivot into data harvesting.
- ›No accounts. Nothing tied to a person.
- ›Local device storage, not cloud storage.
- ›The app grows on your phone as records accumulate, not on their servers.
- ›If acquired tomorrow, there is no data sitting there to monetize.
The Round Table of Second Opinions
Instead of marrying one model, poll all of them. Each model has different training data, different personality, different blind spots.
- ›Ask the same health question to Claude, GPT, Gemini, and Grok in sequence.
- ›Cost is distributed: roughly 12,000 questions a year across all models for ten dollars.
- ›Replaces the "I don't trust that doctor, I want a second opinion" loop with a two-second model switch.
- ›Founder insight: the disagreements between models are often more valuable than any single answer.
The Blue Collar Infrastructure Play
How Savva reached 314,000 connected healthcare institutions without venture capital, by cutting every middleman out of the stack.
- ›Direct EHR integrations instead of Health Information Exchanges, which eliminates middleman API fees.
- ›Wearables pulled through Apple HealthKit instead of direct device APIs, which means one integration covers all devices.
- ›Free ingestion on both sides is what makes a ten-dollar price point survive.
- ›The lesson: infrastructure commoditizes slowly, but unit economics are destroyed instantly by middlemen.
The Global Footprint Thesis
The reason Savva costs ten dollars a year is not marketing. It is a deliberate design decision for a billion-person market that most health tech ignores entirely.
- ›100 million people in the West have access to modern EHRs. A billion people in underserved regions do not, and will not in our lifetimes.
- ›An EHR build costs hundreds of millions of dollars and takes a decade.
- ›Savva works without an EHR: upload a document, it treats it as a visit, and chronological history emerges.
- ›The ten-dollar price is designed to be affordable in Dar es Salaam, not just Denver.
Founder Experiment: Build Your Own Round Table Agent
Build a personal round table agent before your next doctor visit using Cursor or Replit.
- 1Take one health input: a lab result PDF, a week of sleep data, or a medication list.
- 2Write a script that sends the same question through three different model APIs: Claude, GPT, and Gemini.
- 3Print the three answers side by side. Notice which model is cautious, which is confident, which sees a pattern the others missed.
- 4Run it for a week on your own data. You will develop an intuition for which model to trust for which kind of question.
- 5Bonus: wire in local model execution through Ollama so nothing leaves your laptop. You will understand Savva's architecture by feeling it.
The insight: The disagreements between models are often more valuable than any single answer. That tension is exactly the muscle Savva is productizing at scale.
Key Terms
These terms have been added to the AI for Founders Glossary. Search by Stephen Rouse to filter them.
Q&A
What is Savva?
Savva is a unified health intelligence app that connects medical records from 314,000+ U.S. healthcare institutions with wearable data from devices like Whoop, Garmin, Oura, and Apple Watch, then lets users query their combined health history through multiple AI models including Claude, GPT, Gemini, and Grok. It costs $10 per year and stores all data locally on the user's device.
Who founded Savva?
Savva was co-founded by Stephen Rouse and Amit Shah. The two previously worked together at Protocol First, a healthcare data interoperability company acquired by Roche Pharma via Flatiron Health.
How is Savva different from Apple Health, Epic MyChart, or Google Health?
Savva combines three things no other app does together: direct EHR medical record integration, wearable data unification, and multi-model AI querying. It stores data locally rather than in the cloud, has no user accounts, and does not monetize data.
Why does Savva cost only $10 per year?
The pricing is deliberate. Savva built direct EHR and Apple HealthKit integrations instead of paying API middlemen, which collapses operating costs. The low price is also designed for global accessibility, particularly in regions where modern EHR infrastructure does not exist.
How does Savva protect user privacy?
Data is stored locally on the user's phone, not in the cloud. Savva has no user accounts, meaning data cannot be tied to an individual identity. Users can put the app in airplane mode and verify that queries do not leave the device.
Is Savva FDA-approved as a medical device?
No. Like Oura and most consumer health products, Savva is intentionally not classified as a medical device. It explicitly does not provide medical advice and prompts users to consult physicians for clinical decisions.
When is Savva launching publicly?
Savva was approved by the Apple App Store in April 2026 and is targeting a public launch around Earth Day, April 22.