GitHub Copilot CLI combines model families for a second opinion

honestly? combining model families just for a 'second opinion' sounds like massive over-engineering. it's fine for toy projects, but for production, you just need a reliable prompt chain, not a model selector. the goal isn't novelty; it's reducing hallucinations and context drift. we're just slapping a wrapper on existing LLMs. it's a nice feature, but don't mistake complexity for necessity.

we need to focus on grounding the output in the actual codebase, not just having the AI pick a flavor. if the context is garbage, changing the model family doesn't fix the fundamental error. it's an incremental improvement, not a paradigm shift.

my take is that this is neat for internal tooling, but we're overcomplicating the setup just to play with the inputs. save the complex routing for when we actually need reliable, audited results, not for fun.

actionable: Stop chasing model family combinations and focus on robust retrieval augmented generation (RAG) setup.
impact:medium