haven’t really looked at it, but the blurb mentions self recursion a little https://www.reddit.com/r/MachineLearning/comments/11o97on/p_gitmodel_dynamic... [P] GITModel: Dynamically generate high-quality hierarchical topic tree representations of GitHub repositories using customizable GNN message passing layers, chatgpt, and topic modeling. Decompose Python libraries and generate Coherent hierarchical topic models of the repository. https://github.com/danielpatrickhug/GitModel The ability to bootstrap its own codebase is a powerful feature as it allows for efficient self-improvement and expansion. It means that the codebase is designed in such a way that it can use its own output as an input to improve itself. In the context of GitModel, this feature allows for the efficient improvement and expansion of its own codebase. By using its own output to generate hierarchical topic trees of GitHub repositories, it can analyze and extract insights from its own codebase and other codebases to improve its functionality. This can lead to more efficient and effective code generation, better semantic graph generation, and improved text generation capabilities. I spent around 10 hours today on a major refactor creating a simple pipeline abstraction and allowing dynamic instantiation from yaml configs. It now also supports multiple GNN heads. Please try it out and let me know what you think! Example: https://github.com/deepmind/clrs