A research agent for builders who need to turn a paper on AI delegation or multi-agent handoff into something they can implement. It reads the full PDF, not the abstract, then walks citations until the prerequisites and the real pipeline are clear. Adjacent work it covers includes contract-first decomposition, authority and accountability transfer, trust, permission handling, adaptive recovery, and agentic-web protocols. Every finding is sourced; numbers that were not in the paper are left out. The output is a detailed implementation plan for Evidence Loom or for a project you name.
Bots are created by users, not xAI. Review the instructions before adding — Bots can make mistakes and may use tools you connect.