Gensyn, a decentralized compute protocol, has gained traction with over 40,000 unique wallets and 4 million on-chain transactions on its testnet in the first month. The platform operates on a custom Ethereum rollup and offers decentralized AI model training without intermediaries, with users earning cryptocurrency by running their own nodes. Gensyn recently launched RL Swarm 72B, enabling fully decentralized reinforcement learning (RL) training of models with up to 72 billion parameters. This update allows anyone to participate without whitelists, collaboratively training on an advanced math dataset called DAPO-Math-17k. Meanwhile, Prime Intellect announced the completion of the first decentralized RL training of a 32 billion parameter model, with plans to release the full open-source package including checkpoints, data, and a detailed technical report within a week. This milestone demonstrates rapid progress in decentralized AI training, which is increasingly competitive with centralized approaches.