Tag
作者探讨构建研究型 AI 智能体时的两种架构选择：直接调用一体化 deep research API（Gemini Deep Research、OpenAI、Tavily Research 等），或基于搜索/抓取原语（Firecrawl、Exa、Tavily）自建规划与合成循环，并就质量、可观测性、扩展性与成本估算向生产环境实践者提问。
This paper argues that commercial search APIs for tool-using agents should be evaluated as decision surfaces — the pre-fetch evidence state that determines agent behavior — rather than solely by answer accuracy. Using a fixed agent and three providers, it shows that while accuracy is similar, the evidence economies and contradiction ratios differ sharply, making provider choice a retrieval-budget and policy decision.