01
Medicine is running out of people. The WHO projects an 11 million health worker shortfall by 2030. In our too-small-to-matter home country Chile, 2.6 million patients sit on a public waiting list today. The median wait for a specialist or surgery is 242 days.
Health problems do not go stale. People will need help for as long as people exist. Building here is building something people need. For a hundred years.
02
Something else is happening at the same time.
Three years ago, medical AI was failing physician benchmarks. This year Med-Gemini is at 91.1% on MedQA. Google's AMIE reached 81.7% diagnostic accuracy against 53.3% for primary care doctors. Utah has approved an AI copilot that can prescribe roughly 190 chronic medications. China has deployed autonomous clinical AI across 260+ hospitals in 93.5% of its provinces. ARPA-H is funding the first agentic-AI clinical pilot in the United States.
The curve is legible, and it is not slowing down.
03
The bigger question of them all: What happens when AI can actually practice medicine?
Our long term position on medical superintelligence:
1. AI will surpass human doctors in diagnosis and treatment. Not assist - match and exceed. Already happening on benchmarks, but with a long road ahead for last-mile adoption.
2. The world will be forced to let it happen. The medical workforce gap puts increasing pressure on the health system. We believe regulation will follow necessity - telemedicine went from frowned upon to universal when COVID hit.
04
The bottleneck is not the AI itself; frontier labs purposefully push this forward. We see everything else as the bottleneck: clinical integrations, liability and accountability changes, and deployment infrastructure.
The bottleneck is everything around medical superintelligence for it to reach a patient. Clinical integrations. Liability architecture. Data standards. Enough trust from hospital administrators to let it run. A regulatory scaffolding that did not exist six months ago and has to exist in six years. That is the layer we are building. We call it the harness. The road the car will eventually drive on.
When both medical superintelligence and regulations arrive, patients will be reached through us.
05
A company cannot be built on a 5-15 years bet alone, and you cannot learn a landscape no one has ever touched.
So we climb in stages. Each base camp is a product that adds its own value: solves pressing problems for its own users and - bottom line - earns its own revenue. Quietly, we leave a piece of the harness in the ground behind it.
06
We believe AI will take over most of the cognitive work of medicine. We also believe the practice of medicine, as most people experience it, will become more human, not less. The scarce good in a world where diagnosis is abundant is presence. A pediatrician whose voice is half the medicine. An oncologist who sits with a family through the worst year of their life. Those people do not get replaced. They become the part of medicine that patients and societies will pay more to protect.
Fixing the shortage requires medical superintelligence; truly treating patients requires the human touch.
07
The software medicine runs on was built by people who did not love the craft and it shows. Our work sits at the intersection of process engineering, user experience improvement and compliance-grade infrastructure. Medicine deserves better engineering; even more so in the age of medical superintelligence.
Chile is a small country with bimodal healthcare - first-world care for a few, developing-world access for the rest. We got lucky: access to the former while questioning the latter. Always thinking about it, since medicine was always in the room for us.
We come from families of medics. From surgeons and doctors to mental health professionals, our families have been always at the front lines of patient care. One could argue it simply “runs in our veins”.
Looking backwards, this is the problem we were always going to come back to.
Pablo Díaz & Víctor Perl - Superposition Labs