Summer
Robotics

A robot watches a wine glass

In a lab in Campbell, California, a robot arm follows a sheet of glass as it moves. Not a matte box. Not a fiducial-tagged crate. Glass - the thing engineers quietly remove from demos because cameras hate it. The arm tracks it anyway, at 100 frames a second, reacting in roughly the time it takes a housefly to change its mind.

The machinery making this possible is small, the company behind it smaller still. Summer Robotics is about fifteen people. Its product is called Kortx. Its claim is large: give any robot perception that beats the human eye, and a great many jobs robots currently fumble suddenly become possible.

That is the wager. Whether it pays out depends on a single, stubborn problem the founders noticed years before they had a company.

The camera was always a compromise

Here is the inconvenient truth of industrial vision: a camera captures a flat picture many times per second, and then a computer spends precious milliseconds guessing what moved. Add motion and you get blur. Add a shiny part and you get glare. Add a transparent one and you get, helpfully, almost nothing at all. Factories have spent decades engineering the world around this weakness - bright even lighting, matte surfaces, parts bolted into fixtures called jigs so the robot always knows exactly where to look.

It works, in the way that paving over a river works. Expensive, rigid, and faintly embarrassing. Every jig is an admission that the robot cannot actually see; it can only memorize.

Conventional fixes - time-of-flight sensors, structured light, LiDAR - each trade one problem for another. Time-of-flight struggles in bright or changing light. Structured light is slow. None of them love a mirror. The result is an automation industry that can build a car but struggles to let a robot pick up a clear plastic bottle that has tipped over.

A physicist, an optics veteran, and a hunch

Summer Robotics was started in 2020 by Schuyler Cullen and Dirk Smits. Cullen, the CEO, holds a Ph.D. in theoretical physics from Stanford and had spent years leading AI and robotics at Samsung - close enough to the frontier to be unimpressed by it. The bet they made was almost contrarian: stop improving the camera. Replace the idea of the frame entirely.

Their answer borrows from two technologies that had never been formally introduced. The first is the event sensor - a chip that, instead of taking pictures, reports only the individual pixels where light changes, the instant they change, with microsecond timing. The second is a continuously scanning laser. Point one at the other and you get a stream of precisely timed light events bouncing off the world, blur-free by construction, indifferent to ambient lighting.

Summer Robotics calls the marriage laser-event sensing. As far as the company is concerned, it had to invent the category because nobody else had bothered to combine the two.

Kortx, and the end of the blur

Kortx is the platform that packages all of this into something a factory can actually buy. It tracks, recognizes and measures parts in three dimensions at 100-micron precision - roughly the width of a human hair - while updating 100 times a second and reacting within about 20 milliseconds. The company puts end-to-end latency under five milliseconds, and pegs the whole system at five to ten times faster than camera-based rivals.

The headline trick is the materials. Metal, glossy plastic, glass, mirrors - the entire rogues' gallery that breaks ordinary vision - are exactly where Kortx is meant to shine. And because the robot can finally perceive in real time, the fixtures can go. Summer Robotics talks about jig-free assembly, multi-SKU lines that switch products without re-tooling, and setup measured in about thirty minutes rather than days of careful bolting.

Money tends to follow the unsexy problems

It is one thing to claim you have reinvented robot vision. It is another to get the company that builds the machines that build the world's chips to write a check. In September 2025, Summer Robotics announced a Series A led by Applied Ventures, the venture arm of Applied Materials, joined by Solasta Ventures and NAVER's D2SF. Third-party trackers put the company's total funding somewhere between roughly $4 million and $14 million, depending on whose spreadsheet you trust - early-stage money, deployed against a hard problem.

The hire that came with it may matter more than the cash. Rick Van Valkenburg, who spent 28 years at machine-vision firm Perceptron, signed on as Chief Commercial Officer. Founders bring the vision; people like Van Valkenburg bring the purchase orders.

A month later came VoxelSensors, a Belgian photonics specialist, with a strategic R&D and commercial partnership announced during the Belgian economic mission to the United States. The pairing reads like a deliberate division of labor: European optical engineering, Silicon Valley perception software, one shared bet on physically intelligent machines.

Superhuman perception, on purpose

Summer Robotics states its mission plainly: give every robotic system superhuman perception and response, and let entirely new applications fall out the other side. The phrase the company keeps returning to is animal-like perception - vision modeled on biology rather than on photography. Animals do not capture frames and analyze them later. They react to change as it happens. That is, more or less, what an event sensor does.

The markets that follow are the ones where seeing fast actually pays: manufacturing, logistics, and the suddenly fashionable world of humanoid robots, which will need to perceive cluttered, badly lit, unpredictable human spaces without a single jig to lean on.

Back to the wine glass

Return to that lab in Campbell. The robot is still tracking the sheet of glass, untroubled, the way you'd follow a friend across a crowded room. Nothing about the scene is staged for the machine - no special lighting, no matte coating, no fixture holding the glass still and apologizing for the robot's eyes.

That is the whole argument, compressed into one unremarkable motion. For fifty years the automation industry built rooms around the limits of the camera. Summer Robotics is betting it can hand robots a way of seeing good enough that the rooms no longer have to bend. If they're right, a humanoid will one day walk into your kitchen, badly lit and full of glassware, and simply cope.

Fifteen people. One laser. A very old problem, looked at sideways. The glass keeps moving, and for once, the robot keeps up.

See it move