How Preload launched Preload on X
Captures internet-scale training data for physical AI from expert work like surgery
What Preload launched

@PreloadAI (YC F26) is redefining the data layer for physical AI: internet-scale training data from cameras, force-sensing gloves, and EMG, captured from expert work like aircraft maintenance and surgery. Built by ex-Neuralink, Meta, and Paradromics engineers. Demo + careers in https://t.co/mTxjQiV7Rp
How LaunchVideo.com rated the launch
LaunchVideo.com rating
About the launch
Preload records how people use their hands and turns those recordings into training data for robotics and physical AI teams. A head-mounted camera, gloves that measure pressure across the fingers and palm, and a forearm EMG band run in sync, so each demonstration includes the worker's view, the contact, and the muscle effort behind the grip. The San Francisco company is in Y Combinator's Fall 2026 batch, and it is aimed at labs training dexterous policies, vision-language-action models, and world models. Video still dominates robot demonstration data, and it leaves grip force and contact unmeasured. Motion-capture rooms also stay too expensive and too slow for expert work outside a lab. This launch is Preload's attempt to close that gap with wearable capture that can leave the room. The launch presents the stack as a way to capture expert work such as aircraft maintenance and surgery. What the company says is already in hand is more than 350 hours from real-world, non-surgical settings, plus letters of intent from major hospitals. Surgery is the longer-term focus, and Preload has posted a role to run collection with surgeons and clinical partners. In the demo thread, the founders showed a portable bimanual recorder with stereo video, tactile maps for both hands, and EMG traces in one view, and a pick-and-place segment where EMG-derived force estimates rose with heavier objects. They said a policy using those EMG features showed greater force and contact awareness than a vision-only policy, and that they test the signals on their own dexterous hand and robotic arm. The founders are Leo Li, Surya Pandiaraju, and Runming (Tony) Wang, former PhD candidates at Caltech and Johns Hopkins. Li was a neuroengineer at Neuralink and a software engineer at Amazon, with a robotics degree from the University of Toronto. Pandiaraju worked at Paradromics, NeuroPace, and iota Biosciences. Wang builds neural recording and stimulation hardware and previously studied motor-unit coordination with EMG at Emory and Georgia Tech. Their YC profile also lists Meta among the team's brain-computer interface backgrounds.
Tags
Questions
Frequently asked questions
The Preload launch post reached 33K views on X, with 347 likes, 34 reposts, 44 comments, 195 bookmarks.
Preload launched on X on November 18, 2026 with the message "Captures internet-scale training data for physical AI from expert work like surgery".
Preload records how people use their hands and turns those recordings into training data for robotics and physical AI teams.
LaunchVideo.com rated the Preload launch 3.00 out of 5: hook 3/5, editing / creativity 3/5, copy 4/5, sentiment of launch 4/5, distribution strategy 1/5.







