OpenAI's data agent shifts RAG retrieval from raw logs to pre-curated, high-density context

OpenAI 带来的上下文工程范式反思：为何离线编撰能让在线 RAG 看得更少、更准？

OpenAI's internal data agent serves 3,500+ users across 600 PB of data with a single GPT-5.5 model and ~13 tools online. The real work happens offline: Codex reads pipeline code to infer table semantics, turning raw metadata into structured descriptions that online RAG retrieves. Engineer Emma Tang notes that giving the model less but more accurate context yields better results. Six context layers address four pain points: code holds true meaning, query history is noisy, metric definitions live in docs, and correction memory can go stale. Staleness is patched by live schema checks at runtime. The model still overconfidently miscalculated ChatGPT active users as 5 million. No accuracy or ablation data disclosed.

Why it matters: First systematic breakdown of OpenAI's internal Data Agent engineering—offline enrichment + lightweight online RAG is directly relevant to teams building enterprise agents. Deduction because this is a third-party analysis, not a first-party release, and some details come from ...