Horizon by Outworld Labs

Not yet independently verified. Dated from the Y Combinator launch post of 12 November 2026; the maker's site gives no launch date, no pricing and no availability date for Horizon, and its accuracy claims for the physics engine are its own. We will update this when it can be confirmed, and remove this note.

Policies that grasp and place perfectly in a simulator and then fumble on real hardware are the problem Outworld Labs has built Horizon around. The young company, from two UC Berkeley EECS graduates in Y Combinator's autumn 2026 batch, argues that robot training only transfers when the simulator matches the robot, and that today it rarely does.

A physics engine for hands, checked against the robot

At the centre of Horizon is a physics engine written from the ground up for dexterous manipulation, the fine contact work of fingers and grippers where general-purpose engines tend to drift from reality. Outworld's method is to measure how a particular robot actually behaves, fit a model to that evidence and report how closely simulation and hardware agree, so a team knows how far to trust a result before it touches the robot.

Environments, data and evaluation in one place

Around the engine, the company is building tools to create realistic environments, reconstruct and augment real-world recordings, and run simulation, training and evaluation workloads at scale. The aim it states is more varied training data and dependable evaluation across tasks, objects and settings, with fewer physical robots, human operators and manual resets.

Who it is for

Horizon is aimed at teams building robot foundation models and at robotics groups whose training and testing is limited by hardware time. Outworld Labs invites those teams to describe their workflow; it publishes no price or general availability date yet.

Key facts

Maker: Outworld Labs Founders: Matei Gardea and Daniel Grant Launched: Y Combinator launch post, 12 November 2026 Product: Horizon, a simulation platform for robot training and evaluation Backed by: Y Combinator