Why AI coding tools need context-rich harnesses beyond simple search

The rapid evolution of AI-assisted development is increasingly driven by the software that manages large language models, not just the models themselves. Many recent advances focus on building sophisticated harnesses that provide richer context for coding tasks. Earlier this summer, Ars Technica spoke with Cat Wu, head of product for Anthropic's Claude Code, about the company's strategy for creating this kind of context-aware coding tool.