A new mode of knowledge production, known as the 'organic growth pattern', has emerged within the online learning environment. It exhibits several unique features, such as iterative growth, fuelled by collective intelligence, fuzzy complexity, etc. It is more valuable and adaptable to support knowledge innovation within rapidly changing fields. To reveal its new evolutional production pattern and the interaction mechanism of different generative topics, We take a dual perspective of content and behaviour into consideration. A total of 1855 interactive data points were collected from a connectivist massive open online course. The study employed methods, such as latent Dirichlet allocation, content analysis and lag sequential analysis. Results showed that (1) the paths of knowledge innovation were complex and diverse. The process typically began with an ill-structured question and was driven by collective contribution and cognitive conflict. (2) Knowledge production involved interactive evolution and collaborative growth of multiple generative topics. (3) The four stages of knowledge production were cyclic, with sense-making acting as a bridge between way-finding and innovation. Several implications of the findings are recommended to enhance the quality of knowledge growth, such as designing ill-structured questions, providing discussion activities from easy to difficult, introducing relevant topics, taking advantage of various platforms and mobilizing learners' or participants' quality attention-related behaviours.