.png)
The Cursor moment for biology
In software, platforms like Cursor haven’t erased the need for skilled engineers — they’ve shifted the work from wrestling with syntax to designing and iterating on systems.
Biology needs its own Cursor moment.
Imagine diving into an unfamiliar oncology dataset and immediately seeing patterns, collaborators’ notes, and possible hypotheses — without spending weeks deciphering schemas, merging file formats, or engineering pipelines.
The expertise stays where it matters: in the science.
It’s about freeing scientists to focus on designing the next experiment, not plumbing the data. Collaboration becomes instant, exploration becomes fluid, and unknown datasets become starting points, not roadblocks
AI as a semantic interface for science
Researchers think in science: Can tyrosine kinase inhibitors help patients with the exon 21 L858R point mutation?
But verifying that hypothesis with data means leaving the world of science and entering the world of data wrangling, wrestling with schemas, databases, data dictionaries, and code… lots and lots of code. I believe AI can be the bridge between these worlds.
By translating natural language into the technical operations required to explore, join, and analyze data, AI can keep researchers in the semantic space of science while quietly handling the details of data wrangling underneath.
Radically transparent AI
At the same time, I believe the AI user interface should be explicitly designed for radical transparency and control.
Every AI-driven action should provide a verifiable intermediate artifact such as tables, queries, code, or data frames with a clear, step-by-step context.
Researchers should be able to pause at any time, inspect the work, modify it, and fully understand each result. Just as Cursor allows you to inspect the code and “accept” each code difference, we believe the same should apply to scientific analysis.
This level of transparency is not just a feature; it is fundamental to scientific integrity. Scientists must be able to trust, reproduce, and share their work with confidence. They cannot do that without this transparency.
But transparency alone is not enough. Research requires flexibility, the ability to pivot when you need more control or other tools. I believe that any good AI interface needs explicit off-ramps built into every workflow. From any mode, researchers should be able to export and continue in tools like Jupyter, R, SAS, SQL, WDL, or Nextflow. The idea is to start your analysis with AI, but then seamlessly transition to the high-code tools you know and trust when you need deeper control.
With these principles in mind, AI accelerates work; it never becomes an opaque decision-maker. You can edit and extend all artifacts such as code, queries, and data frames without feeling locked into working only with the AI.
Vibe coding :: Vibe research
In software, vibe coding tools like Vercel’s V0 or Figma Make allow creators to explore ideas quickly before committing to production code. They open the door for far more people to prototype applications and generate useful insights without writing production-grade code or waiting on a developer.
Life sciences research should have the same early-exploration mode. Researchers should be able to “trial and error” before committing to in-depth, publication-quality work.
Today, even a simple feasibility question, like estimating how many patients have a certain biomarker profile and have received two or more lines of therapy, can require a small army: IT to permission and extract the data, a data analyst to query it, a clinician to define criteria, and often a statistician and epidemiologist to interpret the results. All this just to learn whether the target population even exists in a dataset.
AI can collapse that process from months to minutes, putting feasibility checks and exploratory analysis in the hands of far more researchers. Once they see a meaningful signal, they can move seamlessly into deeper statistical analysis and validation.
Speed and ease in the service of impact
The measure of AI’s value in life sciences is not novelty, it is impact.
We want to make it faster to get started, easier to iterate, and simpler to collaborate so that more high-impact science happens more often and by more people.
That is the future we are working toward: not replacing scientists, but expanding opportunities for them and for the science we all depend on.