surgical robotics

The recent AI boom has made a lot of previously hard problems feel surprisingly easy.

Tasks that used to require entire teams, specialized workflows, or years of software development can now be compressed into a prompt, an agent, or a model call. Writing, coding, design, research, support, analytics, operations, and many other software heavy workflows are becoming dramatically cheaper and more accessible.

But that does not mean all hard problems disappear.

In fact, I think the opposite happens. As AI eats the formerly hard software tasks, the truly difficult problems become more obvious. Over time, incumbents and a handful of new unicorns will absorb many of the obvious AI applications. The frontier will move toward problems where intelligence is only one layer of the stack: energy, robotics, manufacturing, healthcare, transportation, biology, materials, and other domains where atoms matter.

To me, two of the largest remaining categories are energy and robotics.

Energy is foundational: generation, storage, transmission, and cost. Robotics is the other side of the same coin: the ability to act in the physical world. AI has given us a much better interface to machines. We can now talk to software, tools, and agents in natural language. But the deeper opportunity is not just speaking to machines. It is giving machines better ways to perceive, decide, and act.

That is where robotics becomes interesting.

There are many domains where robotics will matter even without frontier AI. Warehouses, factories, farms, construction sites, logistics networks, homes, and hospitals all contain repetitive, difficult, dangerous, or precision sensitive physical work. AI accelerates this, but the need already existed.

When you ask where robotics could have the largest human impact, healthcare comes up very quickly.

You can imagine Baymax like robots for patient care: machines that help monitor patients, assist nurses, move supplies, clean rooms, deliver medication, or handle the endless busywork that makes hospitals expensive and overburdened. That alone is a massive category.

But then there is treatment itself.

In almost every futuristic science fiction world, healthcare is more autonomous than it is today. A machine can assess you, diagnose you, and provide treatment with extraordinary precision. In Alien, we see surgical pods. In other imagined futures, medicine looks less like a crowded hospital system and more like an automated, always available infrastructure layer for human health.

That brings me to surgical robotics.

Fully automated surgery is a multi decade vision. It is not something that appears all at once. Surgery is too complex, too high stakes, too regulated, and too dependent on edge cases for anyone to honestly pretend otherwise.

But long term visions matter because they create direction. A north star can pull many useful discoveries out of the process, even before the final goal is reached.

SpaceX is the obvious example. The mission was Mars, but the path toward Mars produced reusable rockets, lower launch costs, and Starlink, a technology that is already transformative on its own. The intermediate products were not distractions from the mission. They were part of the compounding curve.

Surgical robotics could follow a similar pattern.

The north star may be fully autonomous surgery. But the first step is not a general purpose robot surgeon. The first step is better procedure specific surgical robots.

Start with minimally invasive surgery. Build robotic systems that make common procedures cheaper, safer, more accessible, and easier to perform. Improve teleoperation so that a surgeon's reach is no longer limited to the hospital they happen to be standing in. Make the platform smaller, faster to set up, more affordable, and easier to deploy across hospitals that cannot justify today's expensive robotic surgery systems.

Then, in parallel, make the robot a data engine.

Every procedure becomes a structured learning environment. The system can record instrument motion, camera views, force feedback, surgeon decisions, anatomical landmarks, errors, corrections, and outcomes. Over time, that data becomes the foundation for training surgery specific models.

At first, those models do not need to perform the surgery. They can assist. They can stabilize the camera, identify anatomy, warn about unsafe movements, suggest next steps, enforce no go zones, measure performance, and help train surgeons. Then they can automate narrow subtasks. Then larger portions of procedure specific workflows. Eventually, maybe, the field gets closer to general surgical autonomy.

The progression might look something like this:

procedure specific teleoperation
        ↓
better surgeon reach and lower cost robotic access
        ↓
robotic systems as data collection platforms
        ↓
surgery specific AI assistance
        ↓
partial autonomy for narrow tasks
        ↓
procedure level autonomy
        ↓
general surgical models

That compounding will not be obvious in a day. It may not even be obvious in a year. But over several years, a platform that improves hardware, collects data, trains models, and expands procedure coverage can start to compound quickly. Over a decade, the curve could become exponential.

This is why surgical robotics is so exciting to me.

It sits at the intersection of the hardest remaining categories: robotics, healthcare, AI, hardware, regulation, human skill, and real world deployment. It is not a pure software problem. That makes it harder. But it also makes it more durable.

The near term opportunity is not to replace surgeons. It is to build tools that make surgeons more capable, procedures more accessible, and hospitals more efficient. The long term opportunity is much larger: a world where high quality surgical care is not limited by geography, hospital resources, or the availability of specialized human labor in a specific room at a specific time.

Fully automated surgery may be far away.

But the path toward it is already valuable.

The best way to predict the future is to create it.