Run a planner and cheap executor subagent coding pipeline

The commenter uses Codex with GPT-5.5 as planner and a GPT-5.4 Mini agent as executor so cheaper models handle tasks that would otherwise cost too much.

What They Did

In the thread, the commenter says they built their own workflow instead of relying on Codex’s default agent flow. They use GPT-5.5 only to plan tasks, then hand those tasks to a GPT-5.4 Mini agent that does most of the actual work, specifically including file editing; the person chose and assembled this routing themselves.

How It Went

The claimed result is lower-cost handling for most coding work because GPT-5.4 Mini covers most tasks while the pricier model is reserved for planning. The comment gives no metrics, examples, or failure cases.

Worth Knowing

The main catch is that they had to build the workflow themselves; the comment links a GitHub project but gives no setup time or cost numbers.