The AI Product Going Viral With Doctors: OpenEvidence, with CEO Daniel Nadler

Growth & Adoption Metrics

• Open Evidence currently serves over 100,000 physicians monthly in the United States alone.
• Global monthly active users exceed 300,000 to 400,000, with over 200,000 actively querying the system.
• This represents approximately 10% to 25% of all active physicians in the U.S., scaling from near zero users just one year prior.
• The growth trajectory relies entirely on word-of-mouth and app store discovery rather than enterprise sales or traditional marketing budgets.

Product Value Proposition & Clinical Impact

• The tool addresses the "long tail" of medical knowledge, specifically handling complex patient cases a doctor may encounter only once or twice in a career.
• It solves the "firehose" problem of medical information, which doubles in volume every five years according to internal conservative modeling.
• The system enables doctors to retrieve comparative safety and efficacy data for rare comorbidities (e.g., treating psoriasis in a patient with Multiple Sclerosis) that standard search engines cannot resolve.
• Co-founder Daniel Nadler estimates the platform could save 300,000 to 800,000 lives annually; a projection of 1 million lives saved is set for approximately 2034.
• Specific reported outcomes include preventing the worsening of comorbidities and identifying life-threatening conditions like pulmonary embolisms via AI-guided reasoning.

Data Strategy & Training Architecture

• The models are trained exclusively on peer-reviewed medical literature, explicitly excluding the public internet, health blogs, or social media to prevent hallucinations.
• A strategic partnership with the New England Journal of Medicine (NEJM) provides full-text access to one of the world's most prestigious journals, a dataset not used by competing AI models.
• The NEJM partnership was initiated organically by senior editorial board members who became power users, rather than through traditional enterprise sales pitches.
• The technical architecture utilizes an ensemble of smaller, specialized models focused on retrieval and ranking rather than a single massive foundation model.
• This "specialized model" approach, validated in their 2023 best paper "Do We Still Need Clinical Language Models?", outperforms larger general-purpose models in medical domains.

Operational Philosophy & "Open" Strategy

• The "Open" in the company name signifies a direct-to-consumer (direct-to-pro) go-to-market strategy that bypasses hospital gatekeepers, IT committees, and long procurement cycles.
• This model democratizes access for community doctors, rural practitioners, and veterans (e.g., at Walter Reed) who cannot afford expensive enterprise SaaS subscriptions ($10,000+).
• The company generated significant traffic to medical journals by linking AI answers directly to original source texts, creating a symbiotic relationship with publishers.
• Unlike traditional AI tools, every answer is grounded in references that allow doctors to drill down and verify the source material.

Team Composition & Hiring Philosophy

• The founding team consists of PhD-level scientists from elite programs (Harvard, MIT), selected for "high neuroplasticity" rather than just pedigree.
• Nadler defines "high IQ" in this context as the specific ability to efficiently acquire and assimilate completely new information rapidly, a trait he equates to the adaptability of historical military leaders like Napoleon.
• The company avoids repeating the "A players want to work with A players" dynamic of his previous venture, Kensho, emphasizing high-velocity learning as the primary driver of progress.

Forward-Looking Statements & Vision

• Nadler predicts that the "wow factor" costs of foundation models will converge to zero, shifting competitive advantage entirely to specialized applications.
• He asserts that humanity has already reached AGI, noting the moving goalposts often equate to a desire for machine consciousness rather than functional capability.
• The long-term vision is "hyper-personalized medicine," where every patient's specific fact pattern is matched against the entirety of medical knowledge to formulate a unique care plan.
• This level of personalization is projected to expand life expectancy ceilings, potentially allowing humans to reach ages of 120–130 through continuous biological optimization ("Theseus's ship" approach).

Technical Constraints & Hallucination Management

• The system is designed to prevent hallucinations by strictly limiting its context to peer-reviewed data, ensuring no unverified internet content influences clinical decisions.
• Unlike generative art where hallucination is a feature, in medicine, hallucination is treated as a critical failure point.
• The architecture includes a "grounding" mechanism that forces the AI to cite specific pages and sections of source journals, enabling human verification.