The paper pitch is a nice one because the speedup is not sold as permission to loosen validation. Splitting JSON into SIMD structural scanning and later value parsing makes the cost tradeoff visible, especially around UTF-8, escaped quotes, and parser disagreement at trust boundaries.