Mixing LLMs like Lego bricks seems smart because models have different...

https://technivorz.com/the-hidden-tax-of-multi-model-architectures-why-more-models-often-means-less-intelligence/

Mixing LLMs like Lego bricks seems smart because models have different strengths and failure modes. It helps spot errors or resolve disagreements. But watch out. Automating synthesis often masks critical dissent or leaks private data across providers