BenchSci
In 2015, PhD student Tom Leung lost rare patient samples because of a bad antibody.
Verdict: BenchSci's mission promises patients faster medicines, but its homepage sells EMET as biology's answer to Cursor.
A Narrative Authenticity Framework analysis of BenchSci. NAF score 8.4 out of 10. Weakest component: Real Aspiration, 7.0.
Originally published on LinkedIn, May 5, 2026, with a score of 8.3. Facts checked and updated November 2026.
BenchSci's own company page starts the story plainly. In 2015, a PhD student named Tom Leung lost rare patient samples because of a bad antibody. The evidence to pick the right one existed somewhere in the literature. He couldn't find it in time.
Leung, a University of Toronto epigenetics researcher, teamed up with David Q. Chen and Elvis Wianda. Liran Belenzon, an MBA student at the Creative Destruction Lab, joined as co-founder and CEO. Their first product read scientific papers to show which antibodies actually worked in which experiments.
Ten years on, BenchSci serves 16 of the top 20 pharma companies. In June 2026 it launched EMET, an AI research environment for drug discovery. The launch post pitched it this way: "Biology never got its Cursor. Until now."
The company page says the mission is fewer failed drugs and medicines reaching patients faster. The homepage leaves the patients out.
Where BenchSci stands
The founders. Liran Belenzon, CEO, with co-founders Tom Leung, David Q. Chen and Elvis Wianda.
The year. EMET launched in June 2026. Former Pfizer R&D chief Mikael Dolsten became the company's first board chairman, Google Cloud became EMET's main infrastructure partner in August, and argenx signed a two-year deal in September. Merck extended its agreement and Sanofi signed a three-year license in late 2025.
The brand line. The homepage reads "EMET, the agentic research environment for R&D," promising "every scientist an agentic team of expert scientists." The company page adds: "we measure ourselves by one thing: whether the science moves faster because we exist."
The NAF scorecard
| Component | Score | One-line read |
|---|---|---|
| Friction | 9.0 | Good science lost to bad information |
| Lived Stake | 9.0 | A founder who lost samples he couldn't replace |
| Distinctive Qualification | 8.5 | A decade of evidence mapping, in a crowded new field |
| Real Aspiration | 7.0 | A patient promise on the company page, not the front page |
| Composite | 8.4 |
Scores as of publication. Current standings at seekargus.com/leaderboard.
Score change: BenchSci rises from 8.3 to 8.4. Real Aspiration moves from 6.5 to 7.0 because the company page now states a mission with patients in it: fewer drugs fail, fewer years are wasted, and more patients get the medicines they need faster.
Friction: 9.0
The problem is specific and costly. Scientists make decisions without evidence that already exists in millions of papers, and experiments fail for avoidable reasons. The homepage now cites that 70 percent of drugs fail in the clinic.
It stops short of 9.5 because the homepage states the failure as an industry number. The lost samples, the version a reader feels, sit on a different page.
Lived Stake: 9.0
Leung lost material he could never get back, and the company tells that story in its own words on its company page. That meets the 9 anchor.
It stays at 9 because the loss was a research setback, and the story doesn't appear where most buyers meet the brand.
Distinctive Qualification: 8.5
BenchSci has spent a decade turning published experiments into a structured map of disease biology. The homepage cites more than 38 million publications, including 16 million closed-access papers, curated with PhD scientists. Big pharma keeps renewing.
The field got crowded in 2026. Anthropic and NVIDIA both launched agent tools for scientific work this year, and BenchSci now competes on their turf.
Real Aspiration: 7.0
The mission names a changed world: fewer failed drugs and faster medicines for patients. That's a real step up from the old "accelerate breakthroughs" line.
The homepage frames EMET as a better tool for scientists, modeled on a coding product. A 9 would put the patient at the top: the medicine that reaches someone sooner because a team didn't spend two years chasing a dead end.
The story BenchSci should tell
In 2015, one of our founders lost patient samples he could never replace, because the right answer was buried in papers no one could read in time. We've spent ten years making sure scientists find it. Fewer failed experiments means more medicines reach patients sooner.
Three moves to raise the score
1. Bring the patients onto the homepage (Real Aspiration 7.0 to 8.0)
Lead the homepage with the mission the company already wrote: fewer failed drugs, medicines to patients faster. Keep EMET as the how. Measure it by how often press and customers repeat the mission line.
Prior-work check: Partial. The mission is on the company page. The homepage leads with EMET's capabilities and the Cursor comparison, as of November 19, 2026.
2. Pair the 70 percent with the lost samples (Friction 9.0 to 9.5)
Next to the homepage's "70% of drugs fail in the clinic," add the two-sentence version of Leung's story, so a buyer sees both the industry cost and the human one. Measure it by time on page and demo requests from that section.
Prior-work check: Partial. The company page tells the founding story, and the homepage cites the 70 percent figure. They don't appear together, as of November 19, 2026.
3. Publish EMET results with named customers (Distinctive Qualification 8.5 to 9.0)
Release outcome numbers from EMET deployments, such as time saved per research question or programs redirected, with customers named. Measure it by new enterprise deals that cite the results.
Prior-work check: Partial. The homepage quotes a customer saying EMET was "probably 10x faster," and press releases announce the Merck, Sanofi and argenx deals. No published outcome data with named customers found, as of November 19, 2026.
| Move | Closest existing program | Status |
|---|---|---|
| Patients on the homepage | Company page mission | Partial |
| 70 percent next to the lost samples | Company page founding story | Partial |
| EMET results with named customers | Homepage testimonial; deal releases | Partial |
Projected composite: 8.9 (Friction 9.5, Lived Stake 9.0, Distinctive Qualification 9.0, Real Aspiration 8.0). These checks cover public sources only and can't see BenchSci's internal plans.
The LinkedIn original suggested making the patient sample story the front door. BenchSci has since put it on its company page, so that move was reshaped.
Sources
- BenchSci company page: founding story and mission, viewed November 19, 2026
- BenchSci homepage, viewed November 19, 2026
- BenchSci: introducing EMET, June 8, 2026
- BenchSci names Mikael Dolsten chairman, July 2026
- BenchSci partners with Google Cloud, August 2026
- BenchSci and argenx agreement, September 2, 2026
- University of Toronto: BenchSci's founding
- Campus Technology: Anthropic and NVIDIA agent tools for science, July 2026
Tim Donovan · SeekArgus