Decentralized Physical AI (DePAI) is emerging as the next big paradigm shift in artificial intelligence, where decentralized networks power the training, validation, and inference of AI models on real-world, physical data. Jensen Huang, CEO of NVIDIA, recently highlighted that the shift from generative AI to agentic AI marks a turning point toward Physical AI (i.e. autonomous cars and robots).
At NATIX, we're building precisely this future. With over 250K drivers and 160M km of streets mapped, we’re leading the world's largest camera-based Decentralized Physical Infrastructure Network (DePIN), capturing continuously updated real-world imagery and video data through our expansive data curation network. By turning millions of kilometers of street-level visual data into actionable insights, NATIX is foundational for training advanced, evolving AI models essential for Physical AI (autonomous cars and robots) and smarter maps.
Our flagship products include Drive&, a smartphone app that rewards users to collect street-level data simply by driving, and VX360, a small device (built in collaboration with Grab, Southeast Asia’s leading super app) with the VX360 Companion App that leverages Tesla vehicles’ 4 built-in cameras for 360° imagery data collection. Together, these products form the backbone of a powerful global data collection network, continuously feeding real-world street-level imagery into decentralized AI models for the training, testing, and validation of autonomous cars and robots, as well as for better mapmaking efforts.
We’re thrilled to announce the launch of StreetVision Subnet (Subnet 72) on Bittensor, the ultimate decentralized AI network. By combining NATIX’s curated real-world data with Bittensor’s crowdsourced intelligence framework, we’re contributing to Physical AI by creating a trustless, ever-evolving ecosystem that powers advanced autonomous driving and smarter map-making applications.
Bittensor aligns perfectly with the NATIX philosophy: decentralize the infrastructure, reward the contributors, and let the best models win. The StreetVision Subnet ingests real-world data streams, processes them to generate critical insights for map-making and Physical AI, and continuously outputs progressively refined AI models. The competitive aspect of decentralization ensures miners constantly refine these models, enhancing their accuracy and utility. These improved models are seamlessly integrated back into our Edge Network (smartphones and dashcams), enabling cost-effective, real-time street-level analysis and reinforcing a powerful cycle of continuous improvement and real-time detections.
Detecting and classifying roadwork accurately and dynamically in real-time is crucial for mapping updates, navigation safety, and autonomous vehicle reliability. NATIX feeds real-world data into the subnet, where miners compete to detect roadworks and continuously publish improved detection models. These models are optimized to run efficiently on the AI-accelerated chips of our Drive& smartphone network, enabling real-time roadwork detection. Additionally, we will publish and monetize these models, harnessing decentralization to build the most accurate and reliable roadwork detection model available.
Beyond roadwork, NATIX will rapidly expand to other use cases such as detecting potholes, road signs, littering, and infrastructure monitoring. Our subnet will later enable decentralized analysis and classification of driving videos into “scenarios” that range from routine traffic conditions to rare “edge cases” (sudden weather changes, unexpected obstacles, unusual traffic behaviors such as jaywalking, or complex urban traffic patterns). The diversity and complexity of scenario classification are especially important for autonomous driving due to EU regulations that dictate that automated driving systems must be validated against a minimum set of traffic scenarios relevant to their operational design domain. The process of scenario generation will be fed directly into simulation tools, allowing autonomous driving companies to rigorously train, test, and validate their systems based on real-world data to meet the necessary standards and build safer autonomous vehicles.
Bittensor is a decentralized compute protocol that enables open, permissionless training and inference of AI models. Unlike centralized AI platforms, Bittensor incentivizes a global network of miners and validators to build and refine machine learning models, all while competing to deliver the most accurate and useful outputs. With the StreetVision Subnet on Bittensor, incubated by Yuma (a DCG Company) and inspired by BitMind’s innovative open-source AI architecture, we’re turning this massive, decentralized data stream into useful insights for map making, infrastructure monitoring, and autonomous driving.
StreetVision Subnet is specifically tailored for robust image and video analysis, enhancing both Physical AI applications and dynamic map-making through a decentralized pipeline:
Our subnet operates under Bittensor's Dynamic TAO ($dTAO) model, aligning emissions and robust incentives directly with network performance. With substantial TAO and Alpha token existing in the pool since the launch and ongoing emissions, we ensure a vibrant trading and participation environment.
The subnet issues 2 dTAO tokens per block (approximately 14,400 dTAO per day), shared by miners, validators, and the subnet owner. Additionally, it competes for global TAO emissions based on network utility and performance, providing substantial incentives for continuous model optimization and accuracy. The amount of TAO our subnet earns, which goes directly into the dTAO-TAO pool as liquidity, depends on its incentive ranking, validator voting, and the price performance of our dTAO relative to TAO, signaling the market’s valuation of the subnet's utility and health within the Bittensor ecosystem.
To secure committed and high-quality participation and in order to ensure the security of NATIX Network, the following requirements apply:
The requirements above are subject to change and are not necessarily the final ones. We would like to work with the Bittensor Foundation and community, especially during the 4-week grace period, to create a healthy and functional subnet that serves our vision for physical AI.
To support long-term growth and stability of the subnet, NATIX commits to not liquidating any dTAO emissions under our control for the first three months after launch. Although powered by NATIX’s data, we believe that our subnet can serve as its own ecosystem and generate considerable value. We are still finalizing the dTAO value accrual mechanism, nevertheless, it’s clear that we will work on monetization of miner-generated insights and models, and such revenues will fund $dTAO buybacks and burns, creating deflationary pressure and ensuring long-term value accrual. Additionally, to create a fair model, NATIX will charge the subnet for access to its raw data streams, which form the layer for generating valuable outputs foundational within the subnet.
Decentralization isn’t just about blockchain—it’s about empowering global collaboration to build smarter infrastructure and safer autonomous systems. StreetVision Subnet is pioneering the future of crowdsourced, decentralized AI-powered maps and autonomous driving solutions.
Become a part of StreetVision:
This is the future—powered by Physical AI, decentralized intelligence, and continuous innovation. Stay tuned for details on joining and earning with StreetVision Subnet. For technical issues regarding the NATIX Subnet on Bittensor, visit Subnet 72 on Bittensor Discord. For NATIX-related issues, you can read more on NATIX Network, join the NATIX Discord, or follow NATIX on Twitter.
Decentralization isn’t just about blockchain—it’s about empowering people to contribute to global infrastructure in new ways. This is just the beginning. From roadwork detection to future AI-powered mapping insights, we’re pioneering a new standard for decentralized intelligence and opening up the potential of decentralized physical infrastructure networks.