Doctors aren’t the problem
with Haresh Patel, Diagnostic MD
Doctors aren’t the problem
Show Notes
Haresh (Resh) Patel spent 12 years building Mercatus - a FinTech platform that aggregated disparate financial records into structured, visual intelligence for institutional experts. The platform was ultimately acquired by State Street with backing from TPG Capital. He burned through $30 million, pivoted multiple times, and emerged with a pattern: if you can solve the fragmentation problem in a domain where experts are drowning in siloed data, you can build something extraordinary. The hard part is not the technology. The hard part is understanding who you serve, how you make money, and why your product deserves to exist.
Now, two years into a deliberate listening phase, Resh is relaunching Diagnostic MD with the same architectural thesis applied to healthcare: fragmented records into structured data into visual intelligence for specialists - but this time consumer-centric, with the patient owning the record rather than the institution. This conversation is a masterclass in the cost of building before you truly understand a domain, and what it looks like to do it right the second time.
The 13.2-Minute Problem: Doctors Are Victims Too
The average American doctor's visit runs 13.2 minutes. Not because physicians are bad at their jobs - most went into medicine out of genuine compassion - but because the system has been engineered for throughput, not for humans. The administrative burden, the EHR overhead, the reimbursement structures: all of it extracts time from the clinical relationship and deposits it into documentation.
Resh's reframe is important: doctors are as much victims of the healthcare system as patients are. The adversarial framing - patient vs. institution - misses the structural reality. A patient with complex, chronic conditions may have records at Stanford, UCSF, and Palo Alto Medical Foundation - each a separate Epic instance, each invisible to the others. The right hand doesn't know what the left is doing. The doctor seeing you in 13 minutes doesn't have the picture needed to see you fully. This is an architecture failure, not a character failure.
Pearl Below the Mattresses: The Buried Root Cause
Resh's central metaphor comes from the fairy tale: the princess feels the pea beneath twenty mattresses. A chronic medical condition's root cause is rarely what presents on the surface, and most clinical practice is optimized for surface-level intervention rather than root-cause resolution. The system treats the symptom because that is what the reimbursement model rewards and what a 13-minute visit allows.
Resh's own story validates the thesis. After years of tinnitus - a condition that had followed him since a roof collapsed and caused a fall and spinal misalignment - conventional medicine treated the symptom. A chiropractor eventually found and corrected the structural cause. The tinnitus resolved. Not because advanced technology was deployed, but because someone was looking through the right layer. Diagnostic MD is built around enabling that kind of layered search systematically: pulling three data streams together so that what is buried becomes visible.
The Three Data Streams: Why No Specialist Can See the Whole Person
The diagnostic gap in healthcare is not a data shortage problem - it is a data integration problem. Diagnostic MD's architecture is built around three streams that no specialist silo currently integrates: longitudinal medical records (what the healthcare system has captured across providers and years), wearable data (continuous biometric information from the body between clinical encounters), and life story - the mind-body layer that includes trauma, stress, relational patterns, and the contextual events that precede or accompany physical symptoms.
The third stream is the one most systematically excluded from clinical care despite being extensively documented in the research literature as medically relevant. A person who develops a chronic condition following a significant life stressor is presenting a clinical signal in their narrative that never makes it into an EHR. Diagnostic MD's consumer-centric model - where the patient owns the record and populates their own history - is designed to capture what institutional records structurally cannot.
Rookie Smarts: The Counterintuitive Hiring Argument
Resh draws on Liz Wiseman's framework from her book Rookie Smarts to make a hiring argument that cuts against conventional wisdom. The right engineer for a domain-disruption problem is not the most experienced one - it is the one who has enough experience to know what good looks like, but has genuinely thrown out their prior methodology and is relearning with modern tools. The worst hire is the expert who arrives in a new domain with seventeen years of muscle memory and refuses to update.
Importantly, "Rookie Smarts" is not a youth argument. A 50-year-old engineer who has rebuilt their mental model around current capabilities is a Rookie Smart hire. A 28-year-old who already thinks they have figured out the only right way is not. The framework pairs this profile with a senior mentor who knows the domain deeply enough to pressure-test the architecture - the combination produces better outcomes than either alone.
Stop, Listen, Relaunch: What $30M Actually Bought
The most instructive admission in this conversation is also the most uncomfortable one. After $30M spent across multiple pivots at Mercatus, Resh did what most founders will not: he spent two years doing nothing but listening. Not customer discovery calls. Not a research deck. Actually listening - to doctors, to patients, to administrators - before deciding what to build next.
His diagnosis: he had been answering questions nobody asked him. He understood the architecture. He did not deeply understand who it was for, how they would pay for it, and why this product should exist rather than someone else's. The framework he carries forward is blunt: technology is the easy part. The hard questions - who do you serve, how do you make money, why does this product deserve to exist - must be answered before you write code, or you will pay for that omission at maximum cost. He also uses Brain Grid (Tyler Wells, a previous show guest) for PRD generation and Claude Code for development, keeping the tech team small and deliberate.
- The Whole Person Diagnostic - Three data streams - longitudinal medical records, wearable data, and life story (mind-body/trauma/stress) - create a complete clinical picture that no specialist silo can provide. Consumer ownership of the record is the key enabler. See Frameworks.
- Rookie Smarts - The best technical hire is not the youngest or the most experienced, but the one who has enough experience to know what good looks like and has genuinely thrown out old assumptions to relearn with modern tools. Pairs with a senior domain mentor for maximum effect. See Frameworks.
- Stop, Listen, Relaunch - Spend serious time listening to customers, practitioners, and domain experts before rebuilding. Technology is the easy part. The hard questions - who do you serve, how do you make money, why does this exist - must be answered first. See Frameworks.
- Brain Grid - PRD generation tool built by Tyler Wells (previous show guest) that Resh uses to translate product thinking into structured requirements before development begins.
- Diagnostic MD - Consumer-centric health platform integrating longitudinal medical records, wearable data, and life story to surface root-cause diagnoses that specialist silos cannot reach. Patient owns the record.
- Pearl Below the Mattresses - A metaphor for the buried root cause of a medical condition: the real origin is hidden beneath layers of symptoms, partial records, and institutional blind spots. Effective diagnosis requires systematically removing each layer. See Glossary.
- Longitudinal Health Record - A continuous, unified record of a patient's medical history across all providers and time periods, as opposed to the fragmented, institution-specific records most patients currently have. See Glossary.
- Mind-Body Connection (Clinical) - The medically documented relationship between psychological factors - trauma, stress, relational patterns - and physical health outcomes. Systematically excluded from EHRs despite strong research support. See Glossary.
- Digital Twin (Medical) - A computational model of a specialist physician's diagnostic methodology, enabling that expert's reasoning pattern to be applied at scale without requiring the expert's direct time. See Glossary.
- Rookie Smarts - The cognitive posture of an experienced person who has deliberately shed prior methodology and is relearning with fresh tools and assumptions. Not a youth argument - a mindset argument. See Glossary.
- Data vs. Information (Healthcare) - Raw clinical data (labs, notes, images) becomes information only when it is integrated, structured, and interpreted in context. Most healthcare systems generate data; few generate information. See Glossary.
- Patient Portal Centric - A product architecture in which the patient - not the institution - owns, controls, and contributes to their health record. Enables integration of data across provider silos and inclusion of patient-generated context. See Glossary.
Why are doctors victims of the healthcare system, not adversaries?
Because the system was not designed for the patient-physician relationship - it was designed for throughput, reimbursement optimization, and institutional risk management. Doctors went into medicine for compassion. What they encountered was 13.2 minutes per patient, EHR overhead, and administrative burden that leaves no time for the clinical depth their training prepared them for. The adversarial framing misses this. Doctors, like patients, are operating inside a system that does not give them what they need to do their jobs well. Resh's architecture is designed to help both - not to route around physicians but to give them the integrated picture they currently cannot access in the time they have.
What are the three data streams, and why do you need all three?
Longitudinal medical records (what the healthcare system has captured), wearable data (continuous biometric signals between clinical encounters), and life story - trauma, stress, relational context, and the narrative events that precede or accompany symptoms. The third stream is the most systematically excluded from clinical care, despite being well-documented in the research literature as medically relevant. A patient who develops a chronic condition following a significant life stressor is presenting a signal that never makes it into an EHR. You cannot build a complete picture without it. No specialist silo integrates all three. A consumer-owned record can.
What is the 'pearl below the mattresses' and how does it apply to medical diagnosis?
It's the buried root cause - the pea beneath twenty mattresses that the princess still feels. Chronic conditions typically don't originate where they present. Most clinical practice is optimized for surface-level intervention: treat the symptom, manage the episode, move to the next patient. The structural origin - a spinal misalignment, a chronic stressor, an unresolved trauma - gets missed because no one is looking through the layers systematically. Resh's own tinnitus resolved after a chiropractor found and corrected a spinal misalignment from a fall he'd had years before. Not advanced technology - just someone looking at the right layer.
What does Rookie Smarts mean for building a technical team?
It means the best hire for a domain-disruption problem is not the most experienced engineer and not the youngest - it's the one who has enough experience to know what good looks like, and has genuinely thrown out prior methodology to relearn with current tools. The worst hire is the expert who arrives with seventeen years of muscle memory and refuses to update. A 50-year-old who has rebuilt their mental model around modern AI capabilities is a Rookie Smart hire. A 28-year-old who already thinks they have it figured out is not. The framework pairs this profile with a senior domain mentor - someone who knows the domain deeply enough to pressure-test the architecture. The combination outperforms either alone.
What are the three questions you must answer before writing code?
Who do you serve, how do you make money, and why does this product deserve to exist. Resh spent $30M and multiple pivots at Mercatus learning that technology is the easy part. He knew what the architecture should be. He did not deeply understand who it was for, how they would pay, and why his product should exist rather than a competitor's. The two years of listening that preceded Diagnostic MD's relaunch were not wasted time - they were the most productive phase of the work, because they produced answers to these questions rather than better answers to questions no one had asked.
How should founders evaluate investors, and what does a good investor criteria list look like?
Resh's frustration with the VC process centers on the blank page request: 'tell me what you're building and I'll react.' That dynamic puts all the intellectual labor on the founder and produces inconsistent, emotionally driven feedback. His benchmark for a useful investor interaction came from a Denver VC who showed up with ten explicit yes/no criteria before the pitch began: this is what we invest in, and these are the conditions that must be true. That level of transparency is useful regardless of outcome - it tells you immediately whether you're talking to the right person. Founders should seek investors who have done their own thinking, not those who outsource their diligence to the founder's enthusiasm.