CrunchDAO
Decentralized prediction protocol connecting data scientists and AI engineers
Performances
RD Popularity Index
RD Growth Index
Comparison
Follow Updates
Follow Lists
People
About CrunchDAO
CrunchDAO is a decentralized AI research community that connects data scientists with institutions through prediction competitions (Crunches). Institutions publish real-world problems, and over 11,000 ML engineers and 1,200 PhDs worldwide build models. The system automatically evaluates submissions and distributes USDC rewards.
Its core competitiveness lies in lowering participation barriers: researchers can submit models without blockchain knowledge, data remains private through Trusted Execution Environments (TEE), and model aggregation improves accuracy. The platform runs on Bittensor subnets and uses the CFI index to help researchers choose suitable competitions, addressing issues like data silos, limited perspectives of internal teams, and lack of resources for independent researchers.
In the past six months (March to September 2026), the platform iterated continuously: fixed submission issues for users in China, added a GPU request mechanism (submissions not requiring GPU must request access), and now displays real-time metrics and durations. New competitions like broad-obesity-3 and numinous were launched, with numinous rewarding only the signal track. Additionally, the CFI index was introduced on Bittensor, combining task fit, profitability, competition density, etc., to help researchers efficiently select subnets. These updates improved usability and transparency, but GPU resource limits and adjustments to some competition rewards also sparked discussions.
Major progress since March 2026: On March 31, CrunchDAO integrated Numinous (SN6), a Bittensor forecasting subnet, pooling predictions from 11,000+ ML engineers via consensus mechanisms. Top contributors earned ~$100,000 TAO monthly. However, the Numinous subnet was officially discontinued by August 2026, with the platform pivoting to other coordinators like Synth for probabilistic financial forecasting. Future focus: implementing Trusted Execution Environments (TEEs) for encrypted model submissions, enabling institutional participation while protecting IP; and lowering entry barriers for researchers through Crunch's coordinator abstraction.
Co-founders: Jean Hérelle (CEO), founder of CrunchDAO/Crunch Lab, former quant researcher at DataCrunch; Peter Cotton, co-founder and Chief Scientific Officer.
Core executives: Philipp Wassibauer (CTO), ex-Director of Engineering at Dune Analytics and Head of Engineering Europe at TripActions, led platform and on-chain protocol design; Jennifer Morrison (CMO); Grégoire Colcombet (Chief of Staff).
The community spans 12,000+ data scientists and 1,200+ PhDs, trusted by ADIA Lab and the Broad Institute.