Partnering with Antioch to Close the Loop for Physical World
The development platform for autonomous machines
We’re excited to expand our partnership with Antioch through its $32M Series A led by Greylock. We led the company’s pre-seed just over a year ago. Since then, we’ve watched them change the way some of the world’s biggest companies build and test hardware.
When we first met Harry Mellsop and Alex Langshur over coffee in New York, the company had no product or name. Yet Harry and Alex charted a clear vision, built on a deep belief about the future of physical AI development and the power (and unmet promise) of simulation’s role therein.
Harry and Alex had recently come off a successful exit, selling their previous blockchain company, Transpose, to Chainalysis, where their product became one of the core revenue drivers of the business. They could have stayed in the category and run the playbook again. Instead, they chose to start over together in robotics.
Over that first coffee, they told us why. Robotics had always been the thing they wanted to build. Harry had worked on Tesla’s Autopilot team, and both had been building and researching robots since their days at Stanford. Now they had decided to devote themselves to it completely. By the time we met, their apartment had become a makeshift robotics lab.
What they wanted to build was a single place where physical systems could be developed like software. Rather than have every robotics team wire up its own hardware and book time on physical test rigs, Antioch would let a developer build and test a physical system in simulation before it ever touched the real world.
Just as impressive as their product vision was their Day 0 clarity on team building. Before they had raised a dollar, they already had a list of the smartest people they knew, many of them from Stanford, whom they intended to go get, one by one. That discipline around talent has held as the company has scaled.
Closing the software engineering loop for the physical world
Software development cycles are fast because the feedback loop is tight. A developer makes a change, tests it, learns, and repeats, often many times a day. It is this loop that has also enabled agents like Devin and Cursor to autonomously operate in the domain. Physical systems lack an equivalent. Testing robots, drones, and sensors in the real world is slow, expensive, and sometimes dangerous. Even large test facilities cover only a fraction of the edge cases that matter.
Antioch closes that loop in simulation. The platform builds digital twins of hardware and generates the environments to test them against (including weather, lighting, and rare and dangerous scenarios that are hardest to stage), then runs those tests in parallel in the cloud and feeds the failures back. And by closing that loop, Antioch makes autonomous agents for the physical world possible—agents that can develop, test, learn, and iterate on real machines.
It is already working. At Amazon’s Ring, Antioch’s simulations have shown how closely they can match physical test results and faithfully reproduce how real systems behave. Customers across security, industrial, aerospace, and defense, alongside partners like NVIDIA and Nebius, are putting Antioch to work on some of the hardest problems in physical autonomy.
Software ate the world when building became a fast, continuous cycle. We believe the same thing is about to happen to the physical world—and Antioch is building the platform that makes it possible.
