A research project brief. Pick one weakness of standard RAG, such as poor recall on multi-hop questions, design a change to the pipeline (query rewriting, hybrid search, graph-based retrieval or adaptive chunking, for example) and measure it against a baseline on a public benchmark. The goal is a result you can publish along with your code.
By the Build Fast with AI teamPublished Updated
The brief
Context
Keep building
Build triage, handoff, and tool-calling workflows with the OpenAI Agents SDK.
Work with GPT-5.5 and GPT-5.5 Pro for long-context reasoning, coding, and agentic workflows.
Use adaptive thinking, effort controls, and long-horizon coding patterns with Claude Opus 4.7.