Problem

General Trajectory addresses the gap between impressive robot demonstrations and economically useful physical labor. Humanoid robots may perform athletic maneuvers, but dexterous manipulation—reliably grasping, placing, and handling unfamiliar objects—is identified as the main bottleneck to useful work. The pain is greatest in environments such as warehouses and manufacturing facilities, where mixed inventory and changing scenes make conventional, task-specific automation expensive, brittle, and difficult to scale.

The clearest initial use case is logistics automation: mixed-SKU palletization, picking, sorting, and packing. The company also demonstrates box packing and folding and placing unseen food items into fulfillment bags. These workflows combine repetitive labor with enough variation and edge cases that a robot must generalize rather than simply repeat a fixed motion.

Product / Service

General Trajectory is developing AI models for robots, including a foundation model and a vision-language-action architecture that connects high-level task reasoning with low-level dexterous manipulation. Its approach fine-tunes vision-language models to produce robotic controls, trains them to reason through multi-step physical tasks, and improves them using simulated rollouts scored by verifiable rewards. The company’s manipulation policy acts as an action expert that handles grasping, placing, and manipulating objects the robot has not previously seen.

For dexterous grasping, the system combines 3D visual perception, a multi-stage grasp-generation pipeline, human demonstrations, and a preference-trained reward model. Human experts compare candidate grasps; the model learns those preferences and uses them to improve the underlying policy. General Trajectory reports testing across 58 real-world scenes, with more than 63% gains on difficult objects such as reflective bottles while maintaining near-perfect performance on standard objects. Public materials describe the technology and research demonstrations but do not specify a finalized commercial delivery model, such as a software license, robot deployment package, or managed service.

Market

The company competes in embodied AI, robot foundation models, and industrial robotics automation, with stated application areas spanning advanced manufacturing, scientific R&D, and autonomous defense. Its logistics focus places it alongside automation providers and robotics companies such as Dexterity, Ambi Robotics, and Locus Robotics, although those companies may differ in whether they sell robotic systems, warehouse automation, or underlying intelligence. General Trajectory’s distinctive positioning is the attempt to make general-purpose robots useful in unpredictable environments rather than optimizing a single fixed workflow.

As of the available 2026 information, General Trajectory is an active San Francisco seed-stage company founded in 2024 and part of Y Combinator’s Winter 2025 batch. Public company databases report approximately $500,000 raised in one round, while the company’s disclosed traction is primarily technical: demonstrations of generalization to unseen objects, autonomous packing and fulfillment tasks, and reported gains over baseline models. Public sources do not disclose revenue, named commercial customers, or paid deployments, so the company should be viewed as early-stage and research/product-validation focused rather than as a business with publicly established commercial traction.