London, UK- FLock.io has collaborated with the Aptos Foundation to advance AI capabilities in generating Move programming language code. This partnership introduces a large language model (LLM) that surpasses previous standards by providing more accurate and efficient code generation for the Aptos network.
FLock.io leverages a dataset composed of community-derived Move code to boost the performance of its decentralized AI model. It aims to refine Aptos-specific coding nuances and improve code clarity and functionality. In initial tests, this model has outperformed ChatGPT-4o, handling various tasks, from basic scripts to complex applications like yield tokenization with AMM trading capabilities.
Jiahao Sun, CEO of FLock.io, emphasized the importance of decentralized AI, stating that the model’s success illustrates the potential of community collaboration in creating specialized and sophisticated tools. FLock.io plans to expand its dataset with expert insights and establish benchmarks for evaluating performance, aiming to launch a production-ready model for Aptos developers. This future model will be hosted on FLock.io’s federated learning platform, ensuring continuous enhancement through decentralized training.
Bashar Lazaar from the Aptos Foundation highlighted that while Move on Aptos is recognized for security and efficiency, current mainstream LLMs lack strong support. The development of a Move-tuned LLM aims to promote decentralized AI advancements and make Aptos technology more accessible across the developer community.
