OpenAI's Symphony turns issue trackers into AI agent orchestrators

OpenAI’s engineering team hit a wall with coding agents: they were fast, but humans couldn’t manage more than five sessions without spiraling into context-switching hell. Rather than squeezing more out of the agents, they flipped the script. They built Symphony, an open-source orchestrator that treats your project management tool (like Linear) as the control center. Instead of babysitting individual coding sessions, agents autonomously pick tasks from the board, work on them continuously, and flag results for human review. The payoff? A 500% spike in landed pull requests in some teams. The key insight? The bottleneck wasn’t the AI—it was human attention. Symphony removes the need for engineers to act as overworked project managers for their AI coworkers, letting them focus on reviewing and steering high-quality outputs. Why it matters: This shifts AI coding from a tool you use to a system you manage, potentially reshaping how software teams scale their output without scaling their headcount.