Crunch, a decentralized machine learning intelligence layer based in Westfield, New Jersey, announced on January 19th, 2026, that it is enabling its community members to mine Bittensor, a decentralized AI blockchain network. This development opens the door for over 11,000 machine learning engineers and more than 1,200 doctorate-level researchers, primarily from enterprise and academic settings, to participate in a decentralized AI ecosystem without needing in-depth blockchain knowledge.
The company plans to facilitate the process by operating coordinator nodes that manage the technical infrastructure necessary to mine various Bittensor subnets, specialized AI task-specific networks within the greater blockchain. This coordinator service abstracts the blockchain-related complexities, such as mining slot management and staking mechanics, allowing contributors to focus squarely on machine learning model development.
Bittensor itself operates as an open-source blockchain network that incentivizes users by creating a marketplace for sharing computing power, data, and AI models globally. Mining on Bittensor entails contributing computational resources and ML models to subnetworks competing to solve distinct AI challenges. Crunch's model aligns well with Bittensor’s subnet architecture, which supports collective intelligence deployment effectively.
Rather than recruiting miners already familiar with cryptocurrency technicalities, Crunch seeks to onboard ML scientists without blockchain expertise by providing infrastructure support and bundling compatible opportunities. According to Jean Herelle, CrunchDAO's founder, the platform's goal is to simplify participation in decentralized AI for the majority of its members, who are data scientists and engineers aiming to contribute without having to master blockchain intricacies. This approach maintains the decentralized mission while enhancing accessibility.
Crunch employs a meta-modeling technique based on principles of collective intelligence. By aggregating hundreds of independent ML models from community contributors into ensemble models, Crunch generates intelligence streams that have consistently outperformed individuals' predictions. This methodology leverages the wisdom of crowds, producing more reliable and diverse forecasting. Within the Bittensor network, Crunch Coordinators utilize this approach to amalgamate submissions into superior ensemble models fed back into the subnets.
This strategy effectively introduces a meta-layer or “super-miner” concept, where coordinators or subnet creators establish markets for model selection that acquire mining slots and deliver enhanced AI services. This innovation addresses a significant bottleneck hindering decentralized AI growth: while many ML scientists in academia and institutions possess advanced modeling expertise, navigating decentralized infrastructure complexities has limited their involvement.
By abstracting blockchain technicalities through coordinator infrastructure, Crunch empowers scientists to focus exclusively on crafting better models. This not only offers additional monetization avenues for these practitioners but also channels vital specialized intelligence into decentralized AI networks, which currently face talent shortages.
This bridging of deep academic AI expertise with decentralized technology opens access to intelligence networks without compromising decentralization tenets. It enables a new wave of contributors to engage with decentralized AI by lowering barriers and creating a more inclusive environment.
Subnet operators interested in launching coordinator nodes under the Crunch framework can direct inquiries to [email protected]. Additional information about the Crunch Protocol Testnet, ongoing community interactions, and updates is accessible via Discord and X.
CrunchDAO itself connects more than 11,000 data scientists and machine learning engineers with organizations across diverse sectors such as finance and genomics to solve predictive challenges. Through open modeling markets, clients access thousands of competing predictive models evaluated strictly on performance metrics, while contributors earn rewards based on their models' success. The platform has deployed over 35,000 models, serving well-established clients including ADIA Lab and the Broad Institute of MIT & Harvard.