Running local and cloud models in the same coding agent: what actually ships in 2026

Someone on r/LocalLLaMA asked a question this week that turned out to be harder to answer than it...