MemoriesUpload launches to teach robots from first-person video of subconscious human corrections
kimmonismus · x · 2026-11-10
- MemoriesUpload is launching: the startup aims to supply internet-scale first-person video of how people actually do tasks, capturing the small, unthinking corrections—adjusting a grip, shifting a part, retrying—that robots rarely see in staged demos.
- Actions are labeled in first-person video; the team says collecting this at quality requires hardware, models and infra working together, which they've spent over a year building.
- Claim: 3 of the top 5 frontier robotics labs are already training models on this data. More varied examples could help robots handle real-world work beyond neat demo setups.
More from Embodied
- Niantic Spatial adds object interaction in Gaussian Splats for embodied AI real2sim training — Scobleizer · 2026-11-11
- SpaceX unveils Starlink Mini 2 with built-in 71Wh battery, up to 4.5 hours runtime — Scobleizer · 2026-11-11
- VLK generates large-scale synthetic data for humanoid loco-manipulation without teleoperation — pabbeel · 2026-11-11
- MacroData Labs launches MacroHands, SOTA 3D hand models from egocentric video — gui_penedo · 2026-11-11
- Otto One launches a home box where AI agents live locally with cloud-model fallback routing — rohanpaul_ai · 2026-11-11
- NVIDIA fits Gemma 4 E4B robot brain on an 8GB Jetson with vision, chat and tool calls — huggingface · 2026-11-11