A Thousand New Doors

Something beautiful is happening in science.

For most of human history, every great leap in biology began when humans learned to see something that had always been there. Microscopes revealed cells. Sequencing revealed the code of life. Each new instrument changed not only what scientists could see, but what they could investigate. Francis Bacon had a name for this: novum organum, the new instrument, the tool that changes not what is true but what can be known. We believe AI can become such an instrument for biology. But its potential depends on solving a problem that more powerful models alone do not resolve: having access to biological information is not the same as being able to reason with it.

Our goal is to build a future where disease is no longer destiny. Where biology becomes something we can read, reason about, and eventually program with intention. Where diseases that feel immovable today become solvable tomorrow, and the balance of power begins to shift back to us: humans.

That is what makes this moment extraordinary. Biology has never lacked possibility. What has been missing is our ability to navigate enough of it at once. Today, ideas can meet evidence faster. Signals separated by disciplines, datasets, assays, and years of research can begin to illuminate one another. Our partners are already making critical program decisions up to 32 times faster, an acceleration that has already translated, for a single partner, into tens of thousands of hours and millions of dollars. The point was never speed for its own sake. It is to give scientists more chances to ask the right question, run the right experiment, choose the right program, and spend their time where it can matter most.

A useful biological finding is more than a connection between two entities. It matters what was measured, what was perturbed, in which biological setting, and whether a relationship was directly observed or inferred. A rise in a gene's expression is not the same as an increase in the activity of the protein it encodes, and a system reasoning about a drug's mechanism has to know the difference. Connecting findings without preserving those distinctions can make an explanation sound coherent without making it scientifically sound. We are building Mithrl to bring that evidence into a shared knowledge system while preserving the context behind it, so that scientists and AI can compare possible explanations, examine the evidence for each, and identify what still needs to be tested.

Some of the most exciting things happening through Mithrl were never on our roadmap. Our partners are finding patterns in biology that neither they nor we knew were there. At least half a dozen of those discoveries have already become patent filings, and those are only the ones we know about. We never designed Mithrl as a patent engine. It emerged. Give scientists a new way to see across biology, and they begin finding things nobody thought to look for. That is the progress we care about most: not doing known work faster, but expanding the space of what can be discovered at all.

That is also why, we believe, productivity is far too small a measure for AI in biology. Biology is an action-oriented domain. An answer matters because of what a scientist can do with it next. Did it reveal a better target? Change which lead advances? Surface a pattern worth protecting? Strengthen a program decision? Give an asset a better chance of becoming a medicine? Those outcomes are our north star, and they make Mithrl's value legible not only to the scientist using it, but to the program team, the sponsor, and the organization behind them: a clear view into what changed, why it changed, and why that change mattered.

That has shaped how we build. No two diseases are the same, and no two drug programs should be forced into the same mold. Every partner brings different data, assays, biological context, standards of proof, workflows, and ways of deciding. So we deploy inside our partners' environments, under their governance, and work directly alongside their teams. Our scientists and engineers build the custom workflows, agents, analyses, and infrastructure each partner actually needs, while the platform underneath preserves the evidence, provenance, knowledge, and context that make that work durable. Scholars call the discipline of returning to original evidence ad fontes, to the sources; it is the habit we build into every answer. The goal is infrastructure that bends around the science itself: custom where the biology demands it, rigorous where trust demands it, and reusable enough that every program makes the next one stronger.

Next week, we launch Mithrl's Biomedical World Model, built on validated biology. It connects experimental results with structured biological knowledge to help scientists investigate mechanisms, compare explanations, and decide what to test next. It works with the frontier models our partners already want to use, while keeping the supporting evidence and the context of each research program available throughout the analysis. Our ambition is for that foundation to stay with a program as its questions evolve, from investigating a target to evaluating a lead and making translational decisions, without losing the evidence and reasoning behind earlier work.

We raised this capital because we believe we are still in the first few minutes of this story. The most important things this infrastructure will make possible may not be things we can name today. A partner filing a patent was not on our roadmap. Neither was watching months of scientific work collapse into a fraction of the time. Those possibilities appeared because scientists were handed a new instrument and asked questions we had never imagined.

The most exciting part is not that we can already see what is possible. It is that we cannot. The map is still being drawn, the signals are still coming into focus, and somewhere beyond what we know today are discoveries, programs, and medicines that have not yet had their chance.

We’re incredibly excited about the future of biology. If you are too, perhaps you should consider joining us.