The tools used to cure disease should belong to the people doing the work.
Not rented back to them.
Not hidden from them.
Not trained on their judgment and then sold as someone else’s intelligence.
They should belong to the scientist, the lab, the company, the hospital, the patient community, and the field itself.
Because biology is not merely an industry. It is one of the ways humanity tries to understand life, suffering, decay, and repair. The instruments we use for that work matter. They shape not only what we find, but what we are allowed to ask.
Why this matters now
AI will change biology.
Not in some vague, distant way. It is already beginning to do the work once reserved for entire teams: reading literature, planning analyses, choosing methods, running code, interpreting results, and proposing the next experiment.
Used well, these systems can make science faster, cheaper, and more honest. They can help a small lab do work that once required an institution. They can help a patient population become legible. They can help a scientist follow more leads, make fewer clerical mistakes, and see patterns that would otherwise stay buried.
That is precisely why the question of ownership matters.
The danger is not that AI will be useless. The danger is that it will be useful enough for science to depend on it before science has any real control over it.
A few companies are now building closed systems that sit between researchers and their own discoveries. They own the infrastructure where the data runs. They own the software that decides how the work is done. They own the model that reasons over the results. They own the memory of how the work unfolded.
And these systems do not become intelligent in the abstract. They become intelligent by watching work.
They watch the datasets people bring.
They watch the hypotheses people test.
They watch which methods experienced scientists trust and which they reject.
They watch the quiet judgment that does not appear in papers but lives in the hands of people who have spent years doing the work.
The expertise of a whole field goes in.
Someone else's product comes out.
Science has watched this happen before. Researchers write the papers. Researchers review them, without pay. Researchers edit the journals. And then their libraries buy the results back through a subscription they cannot cancel, because the work of the field had quietly become someone else's inventory. The field has spent thirty years trying to undo that, and it is not done yet.
The same shape is now forming one layer deeper. Not around the papers this time, but around the judgment that produces them: absorbed into private systems, sealed behind interfaces, and rented back to the people it came from.
That is not progress. It is dependency with better typography.
A closed scientific system does not merely answer questions. It shapes the questions that can be asked. It decides which methods are natural, which assumptions are invisible, which errors are easy to miss, and which workflows become standard by default.
Those are not neutral choices. They are methodological choices, technical choices, and commercial choices, made by people whose incentives are not the same as yours.
When they are made inside a closed company, behind a wall no scientist can inspect, modify, or contest, the direction of biology quietly stops being governed by science alone. It becomes governed by product strategy.
We do not think the people building these systems are villains. Many are serious, careful, and honest. But good intentions are not a governance model. A field as important as biology should not place its future inside systems it cannot see, cannot change, cannot leave, and cannot hold to account.
So we are building the opposite.
Inflexa is an open-source orchestrator for computational biology, from preclinical research through translational work. The analytic engine and the provenance substrate are open under Apache 2.0. They run on your own machine. They work with the model you choose. They can be read, changed, inspected, forked, and kept.
Your data stays yours.
Your hypotheses stay yours.
Your workflows stay yours.
Your results stay yours.
And your judgment is not harvested to strengthen a machine you do not own.
The three layers of control
The emerging play in AI-for-biology is simple: own all three layers.
Own the infrastructure.
Own the harness.
Own the model.
Each layer can be useful on its own. Each can be a legitimate product. But when all three are closed and held by the same company, the result is no longer just software. It becomes a private operating system for scientific work.
Infrastructure is the layer where your data lives and your compute runs. Whoever owns it controls the meter, the location, the permissions, and often the practical boundaries of what you can do.
The harness is the layer that plans the analysis, chooses the tools, runs the steps, records the outputs, and decides what counts as the work. This is not a thin wrapper. This is where science becomes procedure. It is the layer that turns intent into action.
The model is the layer that reasons. It reads your question, your data, your methods, your results, and often your next hypothesis. If the model is welded to the harness, you do not choose the intelligence that enters your scientific process. It is chosen for you.
Closed together, these layers become a system that watches the field, learns from the field, improves because of the field, and remains answerable only to the company that owns it.
The community supplies the knowledge.
The company keeps the control.
We reject this pattern.
Not because companies should not build tools. They should. Not because paid products are immoral. They are not. Not because every piece of software must be free. It need not be.
We reject it because the deepest machinery of scientific reasoning should not become an invisible dependency, owned by a small number of companies that the field has no way to hold to account.
A scientist should be able to open the instrument.
A lab should be able to alter the workflow.
A company should be able to protect its hypotheses.
A field should be able to understand the systems that increasingly shape its knowledge.
Without that, the tool is not merely serving science. It is governing it.
What we believe
Knowledge built by a field should not be captured by a vendor.
Science advances because knowledge accumulates outside the skull of any one person and outside the walls of any one institution. A method is published. A dataset is shared. A mistake is found. A result is reproduced. A tool is improved. Someone else begins where the last person stopped.
That is the moral structure of science: knowledge becomes more powerful when it can be examined, corrected, and reused.
AI should strengthen that structure, not quietly reverse it.
If the workflows of thousands of scientists train a private system, and that system becomes inaccessible to those same scientists except by subscription, something has gone wrong. The living memory of the field has become a proprietary asset.
We are against that.
A tool you cannot inspect is a tool you cannot fully trust.
Trust in science is not a feeling. It is not a brand. It is not a dashboard with a clean interface.
Trust comes from the possibility of inspection.
Can you see what happened?
Can you rerun it?
Can you change the method?
Can you find the bug?
Can someone else check the logic?
Can the system survive disagreement?
A closed system can ask you to trust it. An open system can be tested.
That difference matters.
The more powerful a scientific tool becomes, the less acceptable it is for it to be opaque. When a system only formats tables, opacity is annoying. When it chooses methods, interprets results, and guides experiments, opacity becomes dangerous.
Scientists should be able to steer the instruments they depend on.
A tool that cannot be changed eventually stops being a tool and becomes a constraint.
Biology is too varied for one company’s assumptions to fit every problem. A cancer lab, a rare disease group, a pharma translational team, a computational biology core, and a hospital research unit do not all work the same way. They should not have to bend their science around a closed workflow built for someone else’s priorities.
If the tool does not fit the problem, the scientist should be able to change the tool.
That is not a luxury. It is the difference between agency and dependency.
Reproducibility without openness is fragile.
A closed company can give you a provenance record. It can show you a log. It can export a report. It can promise traceability.
But if the system that produced the record is closed, the record is still a favor. And favors can change with pricing, policy, acquisition, regulation, or strategy.
Real provenance is not merely a receipt. It is the ability to understand the thing that produced the receipt.
The code should be readable.
The methods should be visible.
The parameters should be recoverable.
The artifacts should be verifiable.
The chain of reasoning should not depend on corporate permission.
Without openness, reproducibility becomes a customer feature. With openness, it becomes a property of the system.
Your ideas are yours.
Your data is not raw material for someone else’s moat.
Your negative results are yours.
Your failed attempts are yours.
Your prompts are yours.
Your hypotheses are yours.
Your experimental intuition is yours.
Your hard-won judgment is yours.
None of it should quietly become training material for a private system that later sells a generalized version of your expertise back to the world.
A scientific tool should help you think. It should not feed on the act of thinking.
What we are building
Inflexa is built around a simple principle:
The part that does the science must remain open.
Not partially open.
Not open until the business model changes.
Not open as a teaser for the real product.
Open in the parts that matter.
The analytic engine is open.
The orchestration layer is open.
The provenance substrate is open.
These are the parts that decide what happens, how it happens, and how the result can be trusted. These are the parts that must not become a hostage layer.
Inflexa runs locally as a terminal application. The sandbox runs with Docker or Podman, so the environment your analysis executes in is the same one your colleague gets, without turning every installation into a dependency fight. It can run on a laptop, a workstation, or a cluster. It does not require an account. It does not require permission. It does not require you to tell us what you are working on.
You bring the model you trust.
If you want to use one model today and another tomorrow, that should be your decision. The harness should not be welded to a single vendor’s intelligence. Biology is too important to route through one model, one provider, one policy regime, or one commercial incentive.
Every result carries its history: the method, the parameters, the source, the content hash, the model that reasoned about it, and a cryptographic signature, expressed in a way aligned with the W3C provenance standard. Signed, so that any later change is detectable.
Not a private log inside a vendor’s platform.
Not a decorative audit trail.
A record that can be checked.
The point is not merely to make biology faster. Speed alone is not enough. A bad system can make wrong work faster. A closed system can make dependent work faster.
The point is to make scientific work faster while keeping it inspectable, portable, and owned by the people doing it.
What we commit to
The core is open, and it stays open.
The analytic engine and provenance substrate are Apache 2.0. You can install them for free, run them yourself, read the code, change the code, and build on top of them.
This is not a trial.
This is not a funnel.
This is not “open” in the theatrical sense.
It is open because the core machinery of scientific reasoning should be open.
We do not harvest your work.
Your analyses are not training data for a model you cannot see. Your workflows are not quietly absorbed into a private system. The open core improves in public, through code, review, discussion, and contribution.
It does not improve by watching you work in private.
Inflexa runs where you are.
You can run it locally. You can run it in your own environment. You can bring your own compute. You can bring your own model. You can keep your data inside the boundary you choose.
There is no account required.
Download it and run it. No sign-up. No gate. No permission. No need to disclose your disease area, your target, your dataset, your hypothesis, or your failure.
Bring the model you trust.
The reasoning layer is your choice. The harness is not tied to one company’s model. You decide what reads your data and your hypotheses.
Provenance is open and verifiable.
Every result should carry the story of how it came to be. Not as marketing, but as structure. The record should be inspectable by anyone who needs to trust the result, not only by those with access to a vendor’s dashboard.
Where the paid product fits
We are a company.
That means we need to survive. We need to pay people. We need to support the work. We need to build something durable enough to matter.
Inflexa Cloud is how we do that.
Cloud is for teams: shared workspaces, shared memory across projects, pooled datasets, collaboration, governance, and the ordinary conveniences that matter when real groups do real work together.
But there is a line we will not cross.
We will not move the core scientific machinery behind a paywall.
The parts that decide what your results are, reason about them, and prove where they came from stay open source.
Not until it becomes inconvenient.
Not until investors ask for a tighter moat.
Not until enterprise customers make closed features tempting.
Not until growth makes the old promise expensive.
Always.
Cloud should be worth paying for because it helps teams work better together, not because it holds the science hostage.
If you never pay us, you should still have a complete, honest, useful tool that belongs to you.
We will get many things wrong. Every serious builder does. We will change our minds about implementation details, interfaces, defaults, and priorities.
But we will not change our mind about the central question:
Who owns the instrument?
Our answer is simple.
The people doing the work.
The invitation
If you are trying to cure disease, you should not have to surrender the tools of discovery to systems you cannot see, cannot change, and cannot leave.
You should be able to run the work yourself.
You should be able to inspect the logic.
You should be able to choose the model.
You should be able to keep the record.
You should be able to carry the tool with you.
You should be able to improve it.
So install Inflexa.
Run it on your own machine. Use your own data. Bring your own model. Read the code. Change what does not fit. Break it. Tell us where it broke. If it is useful, help make it better.
Not behind a wall.
Not inside a private platform.
Not as rented intelligence.
Out in the open.
The knowledge to cure disease belongs to the people doing the work.
The tools should too.