You aren't using Codex like me...

2026-07-14 · Source: Matthew Berman · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Robotics & Autonomous Systems · Depth: Advanced, long

Summary

The article details advanced strategies for optimizing Codex usage, particularly with the new GPT-5.6 model family. It highlights that while Soul offers the highest performance for demanding tasks, Luna Max provides superior performance-to-cost ratio compared to Terra. Users should avoid "fast" mode due to its high cost-to-speed ratio. Key optimization techniques include advanced thread management for delegating tasks and monitoring parallel operations, regularly reviewing the "agents.md" file for outdated rules, and integrating plugins like Zapier for extended functionality. The Codex browser is presented as a powerful tool for complex real-world tasks, including system administration. The guide also emphasizes adding useful skills, utilizing "goals" (loops) for sustained, iterative task completion, and enabling mobile control. Crucially, it outlines security measures, such as implementing pre-tool use hooks to prevent catastrophic file deletions and using the "approve for me" setting for sensitive commands.

Key takeaway

For AI Engineers optimizing their Codex workflows, strategically selecting GPT-5.6 models and configuring advanced features is crucial. Prioritize Soul for complex problems and Luna for routine tasks, avoiding "fast" mode due to its cost. Implement pre-tool use hooks and "approve for me" settings to safeguard your system against accidental file deletions, ensuring secure and efficient autonomous operations. Regularly review your "agents.md" file and explore plugins and goals to maximize productivity.

Key insights

Optimizing Codex with GPT-5.6 requires strategic model selection, advanced features, and robust safety configurations for peak performance and security.

Principles

• Model choice impacts cost and performance.
• Iterative goals drive autonomous progress.
• Proactive security prevents catastrophic errors.

Method

Configure Codex with pre-tool use hooks to block dangerous commands. Set "approve for me" for sensitive actions. Review "agents.md" and select GPT-5.6 models (Soul for hard, Luna for easy) based on task complexity and budget.

In practice

• Use GPT-5.6 Soul for hardest problems.
• Use GPT-5.6 Luna for easier tasks.
• Install Matt Pocock's coding skill.

Topics

• Codex Optimization
• GPT-5.6 Models
• AI Safety
• Autonomous Agents
• Workflow Automation
• LLM Security

Best for: AI Engineer, Software Engineer, Machine Learning Engineer

Related on AIssential

• GPT 5.4 is a big step for Codex — GPT 5.4 marks a practical leap for AI agents, excelling in reliability, speed, cost, and precise instruction following.
• GPT-5.6 SOL is HERE — GPT 5.6 offers superior efficiency and agentic capabilities at a lower cost than frontier models like Fable.
• This is absolute chaos... — GPT56's high usage necessitates strategic prompting and configuration to manage costs and rate limits effectively.
• How to use GPT-5.6 — OpenAI's GPT-5.6 introduces tiered models and thinking levels, demanding strategic usage to manage resource consumption and optimize task performance.
• Codex + GPT-5.5 = SUPER APP! Build and Do ANYTHING! — OpenAI's GPT 5.5 and Codex offer a superior, more efficient AI coding and automation workflow.
• [AINews] Codex usage up >10x in 6 months to 7M users, +1M in the past ~day; did Codex overtake Claude Code?? — Rapid AI tool adoption and agentic system evolution underscore a shift towards outcome-driven, efficient, and secure AI deployments.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Matthew Berman.