Back to directory
AI & ML · AI infrastructure / MLOps / evals

Antioch

Develop autonomy at the speed of software.

Enabling the autonomous development of autonomous systems.
New York, USA652 followers
TLVC Rating
Hook
Editing / Creativity
Copy
Sentiment of launch
Distribution strategy
Community Rating
No ratings yet
Your rating
Sign in to rate this launch.

About

Antioch Agent is a browser-based simulation environment for robotics and physical AI teams, letting developers onboard a robot, build high-fidelity scenes, and run closed-loop tests against the full autonomy stack without leaving the tab. The pitch is straightforward for any team working on autonomy: compress field testing cycles that normally take weeks into hours by spinning up digital instances of hardware wired to simulated sensors, then iterate on policies, edge cases, and training data inside the same loop. Antioch lets robotics developers spin up digital instances of their hardware connected to simulated sensors, enabling reinforcement learning, edge case testing, and training data generation without building physical test environments. CEO Harry Mellsop says it's 'the first time you can have autonomous agents iterate on a physical autonomy system, and actually close the feedback loop.' The launch matters now because robotics development remains gated by hardware, and most teams either build mock warehouses or buy expensive data collection runs to train policies. Antioch is targeting the sim-to-real gap, the difficulty in ensuring that robots trained in virtual environments can perform reliably in the physical world, with an agentic workflow that lets autonomous systems iterate on themselves inside the simulator. Use cases already in the wild span autonomous vehicles, agricultural drones, and academic research, including an experiment where MIT CSAIL researcher David Mayo directs AI models to design robots, then uses Antioch's simulator to validate their real-world viability . The company is based in New York and led by CEO Harry Mellsop, previously of security startup Transpose, with cofounders and early team drawn from Google DeepMind and Meta Reality Labs . Antioch recently raised an $8.5 million seed round at a $60 million valuation, led by A* and Category Ventures, with participation from MaC Venture Capital, Abstract, Box Group, and Icehouse Ventures. Antioch Agent is the first public product built on that thesis, aimed at founders and operators who want to develop physical autonomy at something closer to the cadence of shipping software.
Tags
<500KSeedProduct launchExplainerB2BGlobalAI-generatedUSVertical AI
Comments (8)
Sign in to join the discussion.
Nadia Okafor7d ago

Browser-based sim for the full physical stack is a flex, but I want to know what happens when my laptop fan starts impersonating a jet engine at scene 3.

Yuri K.7d ago

Closed agentic loop in the browser sounds great until the WebGPU context dies and your robot has a mid-task existential crisis. What's the recovery story?

Devesh M.7d ago

Docs page or it didn't ship. Also, is any of this open or are we all just admiring the lobby?

Priya R.7d ago

That cold open on the launch video did more for sim-to-real than any whitepaper I've skimmed this year. Whoever edited that deserves equity.

marco7d ago

Tweet copy buries the lede a bit. 'Weeks to hours' should be the hook, not paragraph two.

Tomiwa F.7d ago

Adjacent space here, we've been wrestling with scene fidelity vs iteration speed forever. Curious how you handle contact dynamics without melting the tab.

Linh N.7d ago

Every robotics demo since 2019: 'simulate in hours.' Every robotics team in practice: still calibrating the gripper at 2am. Rooting for you to actually break the curse.

Saoirse Brennan7d ago

The category is about to consolidate hard and a browser-native loop is the kind of wedge that makes ops teams switch without asking permission.