Life-changing products in days, not decades

Since the 1990s, biology has promised cures. T-cells that hunt cancer. Genes that restore sight. CRISPR edits that fix sickle cell. The science is real. Thirty years later, almost none of it has reached a patient.

The bottleneck isn't the science. It's the ten years between a therapy that works and a patient who gets it. Penicillin was discovered in 1928 and took 15 years to reach mass production. A century on, the timeline has barely moved.

People die of things we already know how to treat. Breakthroughs sit in freezers. This is unacceptable.

Tell a parent their child's therapy exists — just not in time. Tell an oncologist the drug that would have worked is still in phase 1. Tell a rare-disease family their programme was killed in a spreadsheet. The cost of slowness is counted in human lives.

The gap is widening

We can read and write DNA letter by letter. We can pull a patient's own T-cells, engineer them to hunt their own cancer, and cure leukaemias that were fatal within months five years ago. We can deliver working copies of missing genes to the cells that needed them. The FDA approved the first CRISPR therapy in 2023, for sickle cell.

Now look at what reaches patients. The FDA approves fewer than 50 new drugs a year. 10 to 15 are biologics. 4 or 5 are cell or gene therapies — across the whole industry.

7,000 rare diseases exist; fewer than 1 in 20 has a treatment. 300 million people live with one. 10 million die of cancer every year. 55 million live with dementia.

For most of them, the honest answer from their doctor is still: we know what this is. We can't treat it.

A batch size of one

The system was built for drugs made for millions of people. Validate one process, produce identical doses forever. Frameworks, manufacturing, quality — built for that.

Modern biology doesn't work that way. Every product is bespoke. Every batch is its own clinical trial — its own risk, its own scrutiny. Validation doesn't scale with volume. The tooling hasn't caught up.

A CAR-T patient is waiting for their own cells, drawn from their own body, engineered for their own disease. The batch is one person. The scientist, QA reviewer, and operator running it aren't on a production line. They're running a rescue, with tools built for a factory.

Friction is not rigor

Seal started as machine-learning models for bioprocessing and scale-up. Every project hit the same wall: the data was scattered, incomplete, or missing entirely. You can't train models on what isn't there.

So we pivoted to generating the data ourselves — hardware for real-time bioreactor monitoring. We thought more data would fix the problem. We were wrong.

Knowing what happened in the tank didn't matter if we couldn't answer the rest. Who operated it. What media. Which cell line. What happened upstream six weeks ago that might explain this week's failure.

None of it lived in one system. We hunted through paper and disconnected databases. A simple question — what batch is this, who approved it — could take hours.

Every deviation triggered a cascade of documentation. Every change triggered months of re-validation. None of it made the product safer.

Paperwork, patchwork systems, validation documents longer than the software they describe — that isn't traceability. It isn't control. It's theatre. Friction is not rigor.

This doesn't just cost patients. It breaks the people making the medicine.

A QC analyst retypes numbers instead of reading results. A PhD scientist spends Friday night re-typing a batch record that'll be obsolete before it's signed. An operator initials the same form he's initialled a thousand times because the system can't tell he knows what he's doing. Their careers go to paperwork, and the industry loses its best people to fields that use their brains.

The GxP wall

I spent years building flexible data systems for research and saw them take off. Then we hit GxP — manufacturing, validation, patient delivery — and stalled.

The same flexibility that made the tools powerful in R&D made them impossible for quality teams. How do you validate something that changes every week? How do you audit a system users modify themselves?

Handing off from R&D to Manufacturing meant rewriting everything. Nobody invested in digital systems because they'd be obsolete in months. Paper won by default.

What changed is AI. Not a better spreadsheet — a system that reads unstructured knowledge and turns it into structured, auditable records. Flexibility and compliance stop being opposites.

The wall comes down.

The execution platform

What's needed is one platform. One system of record. Discovery to patient. GxP-native. Open. AI-first from the ground up.

In Seal, every change is recorded and every action tracked. Audit trails write themselves. Processes run exactly as defined — precise, repeatable, predictable. Version control and centralised reviews make every change visible and every decision traceable.

Scientists and operators get flexibility. Regulators and management get trust. Patients get what they came for.

The scientist spends her day on science, not reconciliation. The operator is guided through a procedure that adapts to what she sees and is trusted to flag what looks off. The QA reviewer works by exception. The auditor walks through without a document request.

Everyone does the work they trained for.

Every run makes the next one faster

A platform that captures structured data does something paper never could: it learns.

Every run generates linked records — parameters, outcomes, deviations, corrections. Models train on what worked and what didn't. Twenty validation runs become three, because the system simulates the rest. Deviation investigations that took days are pre-empted, because the pattern was seen forty runs ago.

The learning doesn't stay with one product. A deviation pattern from Product A flags a risk in Product B before the first batch runs. A tech transfer that once required weeks of re-validation takes days, because the data already proves the process works. Less testing. Less paperwork. Less re-doing what the system already knows.

The organisations that start first compound the furthest. Every month of structured data is a month of advantage that can't be replicated by buying software later.

From decades to days

In Seal a process is structured, linked data — not a stack of documents. Every unit operation, material, piece of equipment, and acceptance criterion is captured and connected in days, not quarters.

AI reads the data, simulates outcomes, predicts failure modes, and proposes changes that used to take months of physical trial and error.

A tech transfer takes days, not months. A deviation stays at two pages, not forty. A change made on Monday is in production on Tuesday.

At the end of that compression is a person. The parent who gets the therapy in time. The scientist who got to do science. The operator whose judgement mattered. The clinician who stopped explaining timelines and started delivering outcomes.

We are the first generation that can engineer biology. We are also the first with AI strong enough to run the delivery machinery biology has always been stuck in. Those two arriving together, in this decade, is the opening. This is the most important problem in medicine right now.

If you're deciding where to spend your talent, spend it on something at this scale. This is that.

This is how we build the future — in days, not decades.