Pave Robotics Is Sending Robots to Fix Your Roads at 3 AM

The Macro: America’s Roads Are Falling Apart

American infrastructure is in rough shape. The American Society of Civil Engineers gives U.S. roads a D grade. States spend over $70 billion per year on road maintenance, and the backlog keeps growing. The core problem is not money, though money would help. The core problem is labor. There are not enough skilled workers to keep up with the rate of deterioration, and the ones who are available cost a lot and work during the day, which means lane closures, traffic, and angry commuters.

Crack sealing is one of the most important and least glamorous jobs in road maintenance. Small cracks in asphalt let water seep in. Water freezes, expands, and turns cracks into potholes. A crack that costs $1 to seal today becomes a $50 pothole repair next year. Everyone in the paving industry knows this. The economics of preventive maintenance are not controversial. The problem is getting the work done fast enough and cheap enough.

The paving industry has been slow to adopt technology. Most crack sealing is done by workers walking alongside a truck, applying hot sealant by hand with a wand. It is physically demanding, repetitive, and happens in traffic. Turnover is high. Quality varies. And the work can only happen during certain weather windows, which creates bottlenecks.

Robotics has already transformed warehouse logistics, agriculture, and manufacturing. Road maintenance is a natural next target. The tasks are repetitive, the environment is semi-structured (roads are more predictable than, say, forests), and the labor shortage creates genuine pull for automation.

The Micro: A Zipline Engineer and a CrunchLabs Product Lead Walk Into a Parking Lot

Pave Robotics is building Tracer, an autonomous robot that identifies and seals cracks in asphalt. The robot handles detection, navigation, and repair without human intervention. The pitch to paving companies: your robots work 24/7, deliver more consistent results than manual crews, and free up your human workers for higher-value tasks.

The founding team is small but credible. Josh Kelly is the CEO. He was a mechatronics engineer at Zipline, the drone delivery company that actually ships blood and medical supplies in Africa and the U.S. Before that, he was head of engineering for Mark Rober’s YouTube team, which means he has experience building things that work in the physical world, on camera, under pressure. He has a mechanical engineering degree from Stanford. Mason Landon Smith is the CPO, coming from CrunchLabs (Mark Rober’s subscription box company) where he was head of product. He studied industrial design at RISD.

They are a three-person team out of San Francisco, part of YC’s Winter 2025 batch. The contact email is [email protected], which tells you this is still very early stage. They are hiring a founding fabrication technician and a founding robotics engineer, so the robot is in active development.

The competitive landscape is thin. There are a few companies working on autonomous road inspection (like RoadBotics, which uses computer vision to assess pavement condition) but very few attempting autonomous repair. The closest comparison might be autonomous floor cleaning robots in commercial buildings, scaled up to outdoor infrastructure. It is a hard problem, and the fact that almost nobody else is doing it is either a sign that Pave is early or a sign that the technical barriers are higher than they appear.

The Verdict

I think Pave Robotics is tackling one of those problems that is obviously important and obviously underserved. The economics of crack sealing are clear: every dollar spent on prevention saves many dollars on repair. The labor shortage is real and getting worse. And the technology to automate repetitive physical tasks outdoors has matured enough that this is plausible in a way it would not have been five years ago.

The risk is the gap between demo and deployment. Building a robot that can seal cracks in a controlled parking lot is very different from building one that works reliably on a busy road in January in Michigan. Weather, traffic, edge cases, and the sheer variety of road surfaces and crack patterns make this a hard robotics problem. Paving companies are also conservative buyers. They want proof, not prototypes.

Thirty days, I want to see video of Tracer working on a real surface, not a lab floor. Sixty days, I want to know if they have a paving company willing to run a paid pilot. Ninety days, the question is throughput: how many linear feet of cracks can one robot seal per hour, and how does that compare to a human crew? If the numbers work, this company could become the standard for preventive road maintenance. If they do not, it is a very expensive science project.