AI Coding Practice: Sub-agents Should Support Background Parallelism and Relegation
brandon_galang · x · 2026-08-05
The author shares insights from tweaking AI coding harnesses, arguing that the implementation of sub-agent orchestration patterns significantly impacts performance. Many harnesses treat sub-agents as a single tool call returning a single output, which limits their potential.
He highlights that a more powerful architecture should allow:
- Continuous relegation: Continuously re-delegating tasks to sub-agents (popularized by Cognition).
- Background execution: Running sub-agents in the background rather than the foreground.
- Asynchronous messaging: Allowing sub-agents to message the parent agent without disrupting the orchestrator's flow.
This enables sub-agents to work in parallel while the developer continues driving the parent orchestrator.
More from coding & agent
- NeurIPS 2026 Call for Papers: Focus on Responsible Meta-Agents — RobertTLange · 2026-08-05
- Spring Evals: Benchmarking LLMs on Real-World Spring Boot 4 Code with Hidden Tests — therealdanvega · 2026-08-05
- Navigating SAP's Agent Architecture: MCP for Tool Execution, A2A for Orchestration — SillyDot3305 · 2026-08-05
- YC's Multiplayer Agent Harness: Great Decoupled Architecture, But It Kills KV Cache — Instance_Not_Found · 2026-08-05
- The 'oh-my' Configuration Trend in Open-Source Coding Agents — HarveenChadha · 2026-08-05
- Ex-Google Engineer Releases 3-Hour Free Course on Building AI Agents — ___Mufasaa · 2026-08-05