Cohere research: coding agent teams beat single agents, right strategy cuts cost 70%
AI Engineer · youtube · 2026-11-07
Cohere senior research scientist Nouha Dziri shares research on teams of coding agents, where a lead splits up work and workers explore, build and check in parallel. Teams beat a single agent on long software tasks but cost several times more — so across hundreds of experiments on ProgramBench and Commit0 she maps where extra workers pay off:
- Delegate the stuck stage: invest workers only where the task stalls;
- The lead should wait for workers: a lead that runs ahead underperforms one that waits;
- Strong lead + cheap planning workers: gets close to an all-frontier team at about 70% lower cost.
She also examines how agents message each other and what larger agent teams will need.
More from coding & agent
- Gemini /teamwork agents caught live: four parallel task logs all running that second — RyanSpunk · 2026-11-13
- Codex CLI 0.168.0 adds model provider picker, side-by-side TUI workspace and animated GIF generation — github-actions[bot] · 2026-11-13
- Free MCP server lets Claude read Altium, Allegro, KiCad and Gerber PCB files directly — woodisruff · 2026-11-13
- a16z Partner Clones Open-Source Game with Fable 5.2 Agent, Which Finds Its Own Horse Animation Asset — FinanceYF5 · 2026-11-13
- OpenAI team shares how it runs and governs fleets of persistent agents internally — romainhuet · 2026-11-13
- Perplexity's Decider scanned 1B tokens across 7,000 agent sessions, found 14 security fixes — denisyarats · 2026-11-13