Match the model to the task, not the hype

Not every step needs your most capable model. Use a top-tier model for planning, tricky reasoning, and ambiguous work — then hand off repetitive or well-defined steps to a faster, cheaper tier to save cost and latency.

- Planning / architecture   → Opus-class (deep reasoning)
- Routine edits / summaries  → Sonnet-class (fast, capable)
- High-volume classification → Haiku-class (cheap, quick)