OpenAI Symphony Autonomously Picks Up Linear Tickets via Codex
OpenAI's Symphony agent watches a Linear project board, autonomously picks up TODO-state tasks, progresses them through in-progress and review Kanban stages, and re-reads comments to act on feedback — without manual triggering. The agent runs Codex locally. Token cost is substantial: approximately 300,000 tokens to create a basic markdown file, compared to roughly 14,000 tokens for direct Codex use.
Why It Matters
Symphony demonstrates the shape of autonomous software-development agents: continuous project-management monitoring plus conversational feedback loops embedded in existing tools. The 20× token overhead vs direct Codex is the current cost of that orchestration layer — meaningful for teams deciding whether conversational continuity inside Linear justifies the inference spend.