Opinions mine based on learning from scratch this space in the last couple months only.

I feel like these architectures built on top of last gen LLMs are mostly useless now.

The current gen jump was significant enough that creating a complex chain of thought with RAG on last gen usually is surpassed by 0 shots on current gen.

So instead of spending time and money building it it's better to focus on 0-shot and update your models to the latest version.

Feeding LLM outputs into other LLM inputs IMHO just will increase the bias. Initially I expected to mix and match different models to avoid it but that didn't work as much as I expected.

It depends a lot on your application honestly.