How can it compete with vast amount of trained codebases on Github? For LLMs, more data equals better results, so people will naturally be driven to better completion with already established frameworks and languages. It would be hard to produce organic data on all ways your technology can be (ab)used.