Grounded Launches Robotics Data API with SOTA Hand Tracking and SLAM
On September 17, Grounded (GroundedSI) launched the Grounded API, a data infrastructure layer for robot training centered on hand tracking and SLAM, with multiple SOTA benchmarks. Several commentators position it as an open alternative to Meta Project Aria's closed ecosystem.
Confirmed
- Hand tracking achieves SOTA with sub-centimeter error on three benchmarks: OakInk2 (8.66mm), DexYCB (5.7mm), and Show3D (13.6mm)
- SLAM benchmarks also reach SOTA
- Real-world egocentric data can be augmented into training data within minutes
- The API integrates with Hugging Face, LeRobot, and rerun, targeting the BitRobot Network ecosystem
Why it matters
- Per BitRobot Network as relayed by @chrisjpaxton, better robot models come from better data, requiring both hardware that fully captures data and tools that convert it into training signals; GroundedSI had already built the former, and the Grounded API completes the latter, closing the loop from capture hardware to training signal
- @iamrobotbear notes the API is positioned as an open alternative to Meta Project Aria, giving the community an egocentric data processing option outside a closed ecosystem
- @databoydg, @micoolcho, and @pabbeel highlight the efficiency of turning in-the-wild ego data into training data within minutes, lowering the barrier for robot data preparation
- Per @TheZachMueller, Lambda founder stephenbalaban recalled a 2013 camera-equipped baseball cap as an early related attempt, reflecting the continuity of egocentric data capture
2026-09-17 ~ 2026-09-17 · 6 related posts
- Grounded API launches with SOTA hand-tracking (<1cm) and SLAM benchmarks — databoydg · 2026-09-17
- GroundedSI launches Grounded API, turning robot data capture hardware into training signal — chris_j_paxton · 2026-09-17
- Grounded launches robot-data API; Lambda founder recalls 2013 camera-cap as first product — TheZachMueller · 2026-09-17
3 near-duplicate retellings: pabbeel · iamrobotbear · micoolcho