Antioch, a fledgling simulation startup founded in May 2023, closed an $8.5 million seed round that places its valuation at $60 million. The financing, led by venture firms A* and Category Ventures, also saw participation from MaC Venture Capital, Abstract, Box Group and Icehouse Ventures. Co‑founder Harry Mellsop told TechCrunch that the capital will be used to tighten the "sim‑to‑real" gap – the disparity between virtual training environments and the unpredictable conditions of the physical world.

Physical AI promises engineers the ability to program robots the way they write software today. In practice, the promise stalls because real‑world data is scarce and building mock‑up warehouses is prohibitively expensive. Antioch’s platform offers a different path: detailed digital twins of warehouses, factories and outdoor spaces where autonomous agents can practice, learn and generate data at scale. By connecting simulated sensors to a robot’s software stack, developers can probe edge cases, run reinforcement‑learning loops and produce training sets without ever moving a physical machine.

The startup’s approach draws a direct comparison to Cursor, the AI‑driven coding assistant that has reshaped software development. "We want to be the Cursor for physical AI," Mellsop said, emphasizing the need for a turnkey toolchain that lets smaller companies iterate quickly without the deep pockets required for building sensor‑laden test tracks. The platform already incorporates models from Nvidia and World Labs, layering domain‑specific libraries that simplify physics configuration for users.

Industry insiders see simulation as a critical safety and cost‑reduction measure. Adrian Macneil, an angel investor and former Cruise executive, noted at the Ride.AI conference that “simulation is really important when you’re trying to build a safety case or dealing with very high‑accuracy tasks.” Waymo’s reliance on Google DeepMind’s world model to test autonomous driving software illustrates the broader trend: virtual environments can dramatically cut the miles of real‑world driving needed before a vehicle hits the road.

Antioch’s early customers include both startups and multinational corporations that are already pouring money into robotics. By offering a shared simulation backbone, the company hopes to generate a data flywheel similar to what powered the SaaS revolution for software developers. "We genuinely think anyone building an autonomous system for the real world will do so in software primarily in two to three years," Mellsop added.

Beyond robotics, the platform is attracting academic interest. MIT researcher David Mayo is using Antioch’s simulator to evaluate large language models that design and test robots in a virtual sandbox, even staging contests where AI‑crafted bots compete to push each other off platforms. Such experiments could pave the way for new benchmarking standards for physical AI, mirroring how AI news automation tools have transformed content creation.

With the fresh funding, Antioch plans to expand its sensor‑perception modules, broaden its library of domain‑specific physics, and deepen integrations with cloud providers. The goal is clear: give robot builders the same off‑the‑shelf development experience that developers enjoy with platforms like GitHub, Stripe and Twilio, but for the physical world.

Used: News Factory APP - news discovery and automation - ChatGPT for Business