OpenRoboto launches decentralized network for egocentric robotics data on Bittensor

OpenRoboto, the team behind Bittensor Subnet 80, has rolled out a decentralized data-collection platform called Shift. The goal: pay people to record video of real-world environments like factories and kitchens, then transform that footage into structured training data for vision-language-action robotics models. How Shift works within the Bittensor ecosystem Shift sits on top of SN80, which went live on the Bittensor mainnet on July 29, 2026. The subnet has been running model-improvement competitions where miners refine open-weight robotics models, upload their weights to Hugging Face, and compete against randomized LIBERO-Pro benchmarks. Only models judged to be strictly better than the current champion earn the right to replace it. Emission distribution among top performers typically follows a 70/20/10 split, concentrating rewards at the top of the leaderboard. As of early September 2026, a full 85% of emissions were flowing toward physical-robot competitions. Shift adds a new dimension to this setup. Rather than relying solely on miners competing to improve models, the platform creates a parallel track for data collection. Contributors are rewarded for submitting egocentric vide...