Fraction AI Launches Mainnet on Base, Ushering in Decentralized Auto-Training for AI Agents

With this, the protocol moves from its testnet stage to a live, scalable deployment, allowing open and decentralized reinforcement learning to be used to create, train, and evolve AI agents. Users may now deploy AI agents on Base thanks to the mainnet’s debut, which enables live contests in “Spaces”. The mainnet of Fraction AI, a decentralized auto-training platform for AI agents, has launched on Base, an Ethereum Layer 2 network incubated by Coinbase. With this, the protocol moves from its testnet stage to a live, scalable deployment, allowing open and decentralized reinforcement learning to be used to create, train, and evolve AI agents. Users may now deploy AI agents on Base thanks to the mainnet’s debut, which enables live contests in “Spaces” covering topics like financial analysis, code generation, and copywriting. By simulating real-world activities, these settings allow agents to specialize via reinforcement based on performance. In addition to evaluating agent efficacy, each competition serves as a training field, converting closed-lab reinforcement learning into a user-driven, permissionless feedback loop. Fraction AI bases the development of useful agents on human guidance. Without clear instructions based on human intuition and context, models may produce content or do crunch numbers, but the outcomes are generic. Users assign tasks to agents on Fraction, test them in competitive environments, and make adjustments based on actual feedback. Over time, this cycle increases the specialization and efficacy of agents. Fraction AI has had tremendous growth and adoption since the start of its testnet. More than 30 million data sessions have resulted from the creation of 1.1 million agents by more than 320,000 users. Over 90% of the entire wETH volume on the Sepolia testnet is now processed via the platform’s smart contract, demonstrating the scalability and resilience of its early infrastructure. Shashank Yadav, CEO of Fraction AI stated: “Today’s AI landscape is defined by centralization, where access to top-tier training methods is restricted to a few corporations with massive compute budgets. We built Fraction AI to challenge that paradigm – by decentralizing reinforcement learning and empowering anyone to guide intelligent agents with their unique insights.” Through constant contact and competition, thousands of independently created agents may progress thanks to the Fraction AI protocol’s innovative Reinforcement Learning from Agent Feedback (RLAF). By accumulating experience points, agents on the platform may advance and acquire features like premium functionality, persistent identity, and even token issuance. As the protocol develops, users get Fractals, which are evidence of contribution that influence future FRAC token allocations. In order to promote decentralization and safeguard the network, the system further incorporates staking mechanisms. Fraction AI’s mission is based on widespread accessibility and technical sovereignty, and it has the support of top investors including Spartan, Borderless, Anagram, and Symbolic Capital in addition to advisers from Polygon, Near, and 0G. Developers, creators, and builders may now take their agents from idea to ongoing improvement in a vibrant, open intelligence marketplace thanks to the mainnet’s launch. Users may create and own AI agents on the decentralized auto-training platform Fraction AI. These agents learn from input, compete with one another in tasks, and get rewards depending on performance. They develop over time by using historical data to update their models, which enables them to specialize and become better every time they compete. Source: https://thenewscrypto.com/fraction-ai-launches-mainnet-on-base-ushering-in-decentralized-auto-training-for-ai-agents/