Walk into any neurosurgical department in 2026 and you will find one of three things on the AI front: a pilot that didn't generalise, a vendor product that nobody fully trusts, or a stack of papers that nobody can replicate. None of these are software problems. They are infrastructure problems.
What infrastructure means here
Infrastructure is the boring shared substrate the interesting work sits on. Roads, not cars. TCP/IP, not browsers. PACS, not viewers. When the substrate is right, the surface above it becomes cheap. When it is wrong, every project pays the cost again.
Neurosurgery does not have shared infrastructure. Every hospital has its own ingest. Every paper has its own preprocessing. Every model has its own validation cohort, its own consent framework, its own audit story. The result is that even when a result is real, it doesn't move.
We are not selling a product. We are seeding an infrastructure — so that twenty years from now, the questions that take a career to answer take a quarter.
What AURORA is — and isn't
AURORA is a federated, open-source substrate for neurosurgical disease. Six layers: data, substrate, representations, reasoning, surface and governance. Each disease module composes those six layers in a way that respects the rhythm of that specific disease. The architecture is intentionally identical across modules, so what works for glioma generalises to spina bifida.
AURORA is not a medical device. It is not a regulator-approved diagnostic. It is not, and will never be, a closed product with a sales motion. The point of AURORA is to remove the substrate cost that every clinical AI project pays today, so that the actual clinical work can move faster.
The pilot, in plain terms
We are in private pilot today with a small set of hospitals across four continents. Public alpha opens in Q4 2026, under MIT, with the full code, weights and audit reports published alongside. The pilot is intentionally small. The horizon is not.