1. Get the right context, instantly Big repos? Cursor lands on the right file in <0.5s (vs 1–3s + 3–5 wrong files). 2. Move faster in large codebases Search stays sub‑second and predictable, even as repos grow. 3. Cut manual context wrangling Add 0–1 extra files per request (vs 2–4), because Cursor starts in the right place. 4. Use fewer tokens 800–1.5k tokens per request vs 2–3k of mixed, partially relevant code.