Lots of potential counter trends to the "god model" thesis:
- proliferation of models at the app layer, even for the same task (e.g. coding)
- consensus view on the growing importance of context engineering.
- OpenAI agent framework including relatively constrained environment, and workflow guidance
- Adoption rates of models that produce components (e.g. speech, image, music, 2D video) relative to models that do more (e.g. real time video with input handling)
@balajis my very well be right that the convergence will be polytheistic.