Summary

OpenEvidence says it became the default clinical-knowledge system for U.S. physicians within 18 months, with daily use by roughly 40% of doctors and about 20 times the usage of the next-most-used clinical decision-support platform. Daniel Nadler places that growth inside AI’s broader “acceleration and compression” of adoption cycles, but the investor-relevant unlock is distribution: a free professional tool can now spread through medicine at consumer-internet speed.
The product converts paragraph-long patient scenarios into semantic searches across 35 million biomedical publications, then retrieves the three to five trials, guidelines, or passages that actually bear on the case. Nadler’s example is a 44-year-old woman with psoriasis and MS: an IL-17 inhibitor could worsen the MS, while an IL-23 inhibitor is safe and well tolerated. “You have one shot to get it right.”
Trust comes from presenting OpenEvidence as a search engine for physicians, not an autonomous answer engine for patients. It surfaces conflicting evidence, treats citations as “a first-class citizen,” and routes doctors into sources such as The New England Journal of Medicine and JAMA. The social contract resembles a Bloomberg terminal: the professional remains responsible for interrogating the evidence.
OpenEvidence’s go-to-market insight was to treat doctors as consumers rather than “appendages of health systems.” Nadler calls it “a consumer internet company masquerading as a healthcare company”: physicians download a free app onto phones they own, while avid users include senior leaders at Mayo Clinic, Cleveland Clinic, UCSF, MGH, Mount Sinai, and other major systems.
Exploding biomedical knowledge makes continuous decision support structural rather than optional. Medical citations doubled every 50 years in 1950; one published estimate puts the interval today at 73 days, while OpenEvidence’s deliberately conservative top-quartile calculation yields five years. Even reading only the top 10% within one specialty would require about nine hours daily, forcing medical education to invert toward lifelong learning.
Nadler hopes AI will augment physicians and distribute specialist judgment rather than remove doctors from the loop. He invokes planes that can land themselves but still retain pilots, while describing OpenEvidence as a scalable “curbside consult.” One rural Georgia oncologist—one of two within 50 miles, serving a 75% African-American population with median household income of $43,000—uses it as a surrogate panel of colleagues.
For preventive health, Nadler sees more leverage in familiar behaviors than a novel technological answer. His Japanese comparison centers on 70- and 80-year-olds walking 10,000–15,000 steps, continuing purposeful work, and eating until 70%–80% full. His hedge matters: genetics remain powerful, but sustained physical and cognitive activity can mitigate some risks.
Nadler’s founder model prioritizes compulsive motivation over originality or elaborate management systems. He calls OpenEvidence “the most obvious idea in the world,” rejects a simplistic “if you build it, they will come,” and recruits exceptionally intelligent people who already possess an internal “propulsion system.” His desired management posture is to “get out of their way.”

Deep dive

1. OpenEvidence made high-stakes medicine a semantic-search problem

Nadler’s opening qualification: AI adoption cycles that once took five or ten years now compress into one or two. OpenEvidence rode that shift to become, by his account, “the operating system for clinical knowledge” in roughly 18 months—used about 20 times more than the next clinical decision-support platform.
Clinical support is categorically different from paperwork or scribing: administrative mistakes can be corrected, but with a patient, “you have one shot to get it right.” Nadler cites the often-repeated claim that medical error is the third-leading U.S. cause of death, then argues that this understates nonfatal harm: probably 10 to 100 times as many patients experience an aggravated condition without dying.
His specimen case is a 44-year-old woman with moderate-to-severe psoriasis and MS. A dermatologist who trained before IL-23 inhibitors were FDA-approved in 2019 must choose across specialties: Nadler says IL-17 inhibition can worsen MS, whereas IL-23 inhibition is safe and well tolerated—exactly the kind of consequential distinction a keyword search misses.
The entire patient history is therefore the query. OpenEvidence semantically interprets it, searches 35 million biomedical publications, and retrieves the exact three to five trials, guidelines, or snippets that respond—even when the decisive detail sits in an RCT’s methodology or population section rather than its abstract.

2. The product earns trust by routing evidence, not issuing verdicts

Keeping users to physicians is a strategic decision Nadler has repeatedly considered changing but has not. The MD attached to a user’s name creates accountability: like a trader seeing an obviously wrong Bloomberg bond quote, the doctor is expected to notice anomalies, inspect the source, and exercise professional judgment.
Where evidence conflicts, OpenEvidence indicates the ambiguity and routes users to both conflicting phase 3 RCTs in The New England Journal of Medicine and JAMA. “It was never presented as an answer engine. It was always presented as a search engine”—a distinction that defines the interface and keeps the physician in the loop.
References were “a first-class citizen” six or nine months before ChatGPT began providing them. Elad says looking at source material is almost the default behavior. He describes OpenEvidence as one of NEJM’s largest referral sources after Google: physicians know the inputs are leading medical journals, “not tweets”—the clinical version of “gold in, gold out.”

3. Treating doctors as consumers broke the distribution bottleneck

Nadler says he had “zero interest in building a healthcare company”; the hack was building a consumer-internet company for knowledge workers. Sequoia’s description captures his intent: OpenEvidence is “a consumer internet company masquerading as a healthcare company.”
Doctors had been treated as health-system appendages despite being the “fighter pilots” making high-stakes decisions. OpenEvidence addressed them directly as people who could download a free app onto a phone they owned. This individual-user model now includes senior leadership at UCSF, MGH, Mayo Clinic, Cleveland Clinic, NewYork-Presbyterian, Mount Sinai, and Cedar Sinai.
Elad’s question raises the concern that medical gatekeeping can deny patients useful information. Nadler supports patient agency and physician-generated handouts, while cautioning against unrestricted interpretation: it took his first graduate statistics course at Harvard to understand clinical trials, while efficacy for a patient with another comorbidity may hinge on a p-value that fear or hope obscures.

4. Medical education must invert as knowledge outruns human bandwidth

Medical knowledge measured by citations doubled every 50 years in 1950; Nadler cites a British Medical Journal estimate of every 73 days today, while questioning its inclusion of all publications. OpenEvidence’s more conservative calculation—only the top quartile of peer-reviewed literature—still produces a five-year doubling time.
Narrow the obligation further to the top 10% of literature within a doctor’s own specialty and it still requires about nine hours of reading daily. Even if the realistic burden is three or four hours, the implication survives: continuing education must become the majority of medical education, not a “wink wink” supplement after school.
Nadler says exceptional physicians report learning 90%–95% of what they practice after medical school and often after residency or fellowship; one 70-year-old told him that most of his current practice was learned in the previous two years. He points to avant-garde approaches at Mayo, Cleveland Clinic, and UCSF that encourage evidence-based medicine, “curbside consults,” and a distributed “hive mind.”

5. AI can widen the consult without removing the physician

Nadler calls the technological moment a possible “singularity event horizon,” making 10- or 20-year predictions unreliable. Yet planes have long been able to land themselves without generating a mass movement to remove pilots. When the hosts note that people already form relationships with chatbots, Nadler’s answer is: “They don’t have bodies yet”—personified trust still matters.
Nadler frames the right way to practice medicine in 2025 for a complex patient as involving a cardiologist, neurologist, oncologist, and other specialists deciding together. The constraint is economic and physical: three or four specialists cost more than one, while the supply of oncologists is not accelerating with the expanding treatment set.
OpenEvidence may approximate that distributed judgment at the edge. Nadler says physicians use it in every state, county, and ZIP code, including rural Alaska and southwestern Georgia. One doctor wrote that he was one of two oncologists within 50 miles, serving a 75% African-American population with median household income of $43,000; he uses the product as his “curbside consult.”

6. Preventive health remains mostly behavioral and cultural

Nadler’s “not popular or politic answer” is that there is no undiscovered preventive-health list. In Japan, he sees older adults walking 10,000–15,000 steps, working into their 70s and 80s, rejecting the fetishization of retirement, and eating until 70%–80% full: purpose, movement, diet, and portion size are the behaviors he emphasizes.
On the cultural question, Nadler says political culture had prevented physicians from speaking plainly about evidence and that the pendulum is now swinging toward a more open conversation not reduced solely to identity politics. On neurogenerative disease, he acknowledges a strong genetic component but says no serious neurologist disputes continued brain use as a mitigant; he cites Sanjay Gupta’s examples, such as occasionally writing with the opposite hand, as ways to form new neural pathways.

7. Founder performance comes from propulsion, not idea worship

The portable lesson from physician adoption is psychological: address knowledge workers as people and consumers, speak to them in a way that “hits different,” and entrenched industry adoption limits may break. Medicine is specific; the human response to being treated as an autonomous user is not.
Nadler rejects “if you build it, they will come” as the lesson of either OpenEvidence or Apple. He interprets Steve Jobs through will rather than ideation and calls OpenEvidence “the most obvious idea in the world.” The useful founder question is not which coffee shop produced an idea, but where to find motivation that is “almost compulsive.”
In his own case, Nadler describes an enormous amount of aggression redirected through intellect and luck. He deliberately resists analyzing its roots: “in the analysis and description of something you kill it,” so probing the trauma behind a propulsion system might weaken the very force being examined.
That lens governs recruiting. Nadler says there is “like a 65 correlation” between being smart and output, so intelligence must be paired with an autonomous propulsion system. He seeks people who are obviously exceptionally intelligent and driven, for whom motivation frameworks and constructive-feedback machinery are “entirely redundant”; the best intervention is simply to get out of their way.

Episode 153 - September 5, 2025 Sarah Paine – How Hitler almost starved Britain