From multi-detector CT to high-field MRI, from PET/CT to molecular imaging, the last 20 years have pushed radiology to the very heart of medicine. It is essential to oncology, catching cancers early, guiding therapy, and tracking response. It is essential to cardiology, diagnosing coronary disease, cardiomyopathies, and valve disorders. It is essential to emergency care, providing the rapid imaging that makes triage possible. Indeed, radiology has become essential to every clinical specialty.

But just as it cements its place as the foundation of modern care, radiology itself is under threat. Nearly 7 in 10 radiologists report their practices are short-staffed, and 67% report being so burnt out that it is hurting their personal life and straining their relationships ^[1]. Imaging use per Medicare patient in the US has risen ~13% in recent years ^[2], with demand projected to grow in the double digits over the next decade ^[3]. Meanwhile, radiologist training capacity has been nearly flat for decades, increasing the workforce ~1–2% per year ^[4].

The clinical consequences are stark. In emergency care, patients who need imaging are 4.4 times more likely to remain in the department beyond four hours and this delay can be fatal ^[5]. In acute stroke, every 15-minute treatment delay, often tied to imaging bottlenecks, increases the risk of death or disability ^[6]. In oncology, COVID-era imaging backlogs shifted breast cancers to later stages and are projected to cause roughly 2,500 excess deaths in the United States over the next decade ^[7,8]. Radiologist burnout itself is independently linked to higher rates of medical error, in many cases with devastating consequences ^[9].

Radiology's centrality means its fragility is the system's fragility. Without urgent action, patients will face longer waits, later diagnoses, higher error risk, and avoidable deaths.

We want to enable human radiologists to accomplish super-human feats not currently thought possible. Practically, this means reading very large numbers of scans, very accurately, and without breaking a sweat. You're given full reports. The reports are correct. Every finding is easy to check. You can move on. It's a delight.

We can see that, time and time again, generative AI has a positive effect on productivity in very highly-skilled work ^{[10, 11]}, with higher adoption rates being tied to greater productivity gains ^[10].

So this is what we're building: Foundation models that are trained to generate full reports directly from scans. We want to handle the repetitive grind of high-volume cases and help with the complex ones. We don't want to just preserve radiologists' capacity, we want to expand it, enabling radiologists to cover more patients with less burnout.

Radiology once reinvented medicine by making the invisible visible. Today, by augmenting human expertise with scalable intelligence, we believe radiology can define the future of medicine once again.