dispatching-parallel-agents-codex

Dispatching Parallel Agents

Overview

When you have multiple unrelated tasks (different subsystems, different bugs), working them sequentially wastes time. Each thread is independent and can happen in parallel.

Core principle: Dispatch one codex_implement per independent problem domain. Let them work concurrently.

When to Use

digraph when_to_use {
    "Multiple independent tasks?" [shape=diamond];
    "Are they independent?" [shape=diamond];
    "Single codex_implement handles all" [shape=box];
    "One codex_implement per domain" [shape=box];
    "Can they run without shared state?" [shape=diamond];
    "Sequential codex_implement" [shape=box];
    "Parallel dispatch" [shape=box];

    "Multiple independent tasks?" -> "Are they independent?" [label="yes"];
    "Are they independent?" -> "Single codex_implement handles all" [label="no - related"];
    "Are they independent?" -> "Can they run without shared state?" [label="yes"];
    "Can they run without shared state?" -> "Parallel dispatch" [label="yes"];
    "Can they run without shared state?" -> "Sequential codex_implement" [label="no - shared state"];
}

Use when:

• 3+ failing test files with different root causes
• Multiple subsystems broken independently
• Each problem can be understood without context from others
• No shared state between threads (no same-file edits)

Don't use when:

• Failures are related (fix one might fix others)
• Need to understand full system state first
• Threads would edit the same files

The Pattern