Entity: OpenAI Symphony
OpenAI Symphony is an internal multi-agent management system and project board framework developed by openai to coordinate parallel long-running agent execution across complex codebases.
Overview & Context Switching Problem
During large-scale internal software development projects—such as the three-engineer build of a 1,000,000+ line codebase—OpenAI discovered that engineers experienced severe context-switching fatigue when managing more than 3 to 5 simultaneous terminal codex sessions. Engineers frequently forgot session statuses, lost track of task progression, and spent excessive time context-switching between terminal windows.
To address this bottleneck, OpenAI created Symphony to move agent state out of transient terminal sessions and into a structured project board interface (similar to Linear or Jira).
Operational Mechanics & Impact
- Ticket-Based Execution: In Symphony, work items are defined as discrete tickets holding deliverables, current status, and active guidance. Autonomous worker agents pull tickets directly from the project board to execute tasks.
- 500% Productivity Surge: Within three weeks of deploying Symphony, landed pull requests among OpenAI engineering teams rose by 500%.
- progressive-context-shaping Integration: Symphony acts as an organizational manifestation of progressive context shaping. By updating the ticket on the board, engineers change the state and direction for the next agent run without needing to hold session state in memory or restart entire agent conversations.