Split work between ChatGPT architect and Codex executor

Tools

ChatGPT, Codex

Overview

The poster built a bridge where ChatGPT plans and reviews while Codex does the local execution work.

Workflow

Suggested implementation: Use ChatGPT for high-level planning and code review, while Codex handles local file execution via MCP-style access. The workflow involves ChatGPT asking clarifying questions and managing subagents, with Codex performing the actual coding tasks on the local machine. Human involvement is required to verify architectural decisions and review generated code.

Reported result

Actual outcomes are unknown as the source is unavailable (HTTP 403). Proposed acceptance check: Measure time saved on manual file operations versus the overhead of code review. Verify that ChatGPT provides stronger high-level planning capabilities compared to Codex in a controlled test.

Human role

The original post does not say how a person stayed involved.

Limitations

Prerequisites include a Windows environment for Codex and configured MCP local file access. Integration uncertainty exists regarding the stability of computer-use within native apps. Human approval is mandatory for all architectural decisions and code merges. Source reliability is low due to missing original content.

Try it

Act as a development workflow orchestrator. I need to set up a bridge between ChatGPT and Codex for a local project. Inputs: [Project Requirements Document], [Local File Path], [Target Programming Language]. Steps: 1. Perform a dry run: Ask ChatGPT to generate a high-level architecture plan for [Project Requirements Document]. Do not execute code yet. 2. Human Confirmation: I will review the plan. Only proceed if I approve the architecture. 3. Execution: Instruct Codex to implement the approved plan using local file access at [Local File Path]. 4. Review: Send the generated code back to ChatGPT for a security and logic review. Output Format: A structured report containing the Architecture Plan, Execution Log, and Review Findings. Quality Metric: The code must compile without errors and pass basic unit tests. Error Handling: If Codex fails to access files, stop and request permission configuration. If ChatGPT's plan is ambiguous, ask clarifying questions before execution. Note: This is a suggested workflow. Do not claim this was tested in the original source. Ask clarifying questions only if [Project Requirements Document] is missing critical technical constraints.

More cases · Software & development

• Update a code base with Copilot chat and agent mode
• Check code against user intent with review agents and tests
• Build a budgeting app from a single prompt
• Build a mental models framework with a multi-agent dev team
• Build a instant-opening markdown viewer for macOS