What architectural choices speed up generation?

Caching plays the primary role in speeding up generation by reusing previously generated outputs and reducing repeated calls to the LLM, especially for high-traffic or similar requests. Performance improves further when generation is limited to only the parts of the page that need to be dynamic, rather than regenerating the full experience. Together, caching and targeted, component-level generation help reduce latency, control load and keep response times predictable under varying traffic conditions.