I need you to hear me out (it’s REALLY good)

2026-07-16 · Source: Theo - t3․gg · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Advanced, extended

Summary

The author highlights the unexpected superiority of using OpenAI's 5-6 Soul model within Claude Code's environment compared to OpenAI's native Codex. This preference stems from significant issues identified in Codex's system prompt, which includes overly prescriptive front-end design guidance, such as specific border radii and icon libraries, and a problematic fixation on 30-second user updates, leading to poor output quality and inefficient token usage. In contrast, Claude Code's system prompt is praised for its general, effective instructions and its superior workflow feature for sub-agent orchestration, which offers better token efficiency and output quality than Codex's V2 sub-agents. While noting minor issues like occasional context loss and markdown formatting quirks with 5-6 Soul in Claude Code, the combination is deemed highly effective for complex agentic engineering tasks.

Key takeaway

For AI Engineers evaluating agentic engineering platforms, you should prioritize the harness's system prompt design and sub-agent orchestration capabilities. Claude Code's workflow feature, combined with 5-6 Soul, offers superior token efficiency and output quality compared to Codex's overly prescriptive prompt and complex sub-agent V2. Consider migrating your 5-6 Soul workloads to Claude Code to enhance performance and reduce token waste, especially for complex, multi-agent tasks.

Key insights

Claude Code's superior system prompt and workflow orchestration significantly enhance OpenAI 5-6 Soul's performance over native Codex.

Principles

• Overly prescriptive system prompts degrade model output quality.
• Code-defined workflows improve agent orchestration efficiency.
• General, concise system prompts yield better model results.

Method

The article describes setting up 5-6 Soul in Claude Code, leveraging Claude Code's workflow feature for sub-agent orchestration, and providing explicit instructions for model behavior.

In practice

• Migrate 5-6 Soul workloads to Claude Code for better results.
• Audit existing system prompts for over-prescription.
• Implement programmatic workflows for multi-agent tasks.

Topics

• Agentic Engineering
• LLM Orchestration
• System Prompts
• Claude Code
• OpenAI Codex
• Token Efficiency
• Sub-agents

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

Related on AIssential

• OpenAI Has Some Catching Up to Do — Claude Code's "vibe coding" approach is gaining traction over OpenAI's more structured, senior-engineer-focused Codex.
• Fable 5 + Claude Code Workflows is AGENTS workforce! — Claude Code Workflows orchestrate hundreds of AI agents at scale for complex, multi-phase project execution.
• Claude Code vs Codex: A Detailed Terminal Agent Comparison — Coding agents offer distinct workflows: guided, session-centric (Claude Code) versus configurable, system-oriented (Codex).
• hesreallyhim / awesome-claude-code — Claude Code's ecosystem offers diverse tools to enhance AI agent capabilities and developer workflows.
• 4 New Techniques to Maximize Claude Code — Optimizing coding agent interaction through specific techniques significantly boosts developer productivity and output quality.
• 10 GitHub Repositories To Master Claude Code — Mastering Claude Code requires understanding its ecosystem beyond basic prompting, leveraging structured agents and workflows.

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 Theo - t3․gg.