Codex Knowledge Base
The Harness Effect: Why the Same Model Scores 16 Points Higher in a Different Tool
Claude Opus running inside Cursor scores 93% on Terminal-Bench 2.0. The same model running inside Claude Code scores 77%. That is a 16-point differential.
The ExecPlan Pattern: Structuring 7-Hour Codex Sessions with PLANS.md
The single biggest determinant of whether a Codex session completes a complex task or drifts into incoherence is not the model — it is the planning.
OpenAI Codex CLI Official Documentation Guide (2026)
The official Codex CLI documentation has grown considerably since the tool's open-source debut in mid-2025.
Learned Capability Governance: What Aethelgard Means for Codex Permission Profiles
A summarisation task receives the same shell execution, subagent spawning, and credential access capabilities as a code deployment task. Sidik and Rokach.
From ChatGPT to Codex CLI: What Changes When Your AI Can Actually Run Code
If you already use ChatGPT to help you write code — pasting in error messages, asking for function implementations, copying suggestions back into your.
Formal Architecture Descriptors: Cutting Codex CLI Navigation Overhead by a Third
Your AGENTS.md file tells your coding agent what to do. But does it tell the agent where things are in a way that actually reduces navigational overhead? A.
Engineering Pitfalls in AI Coding Tools: What 3,864 Bugs Reveal About Codex, Claude Code, and Gemini CLI
When an AI coding agent produces wrong code, developers blame the model. When it crashes mid-session, they blame the tool. A new empirical study from York.
Compiled Policy Enforcement: Why Prompt-Based Safety Fails at 48% and What PCAS Means for Codex Hooks
Prompt-based policy enforcement — telling a model never do X in a system prompt — achieves only 48% compliance even with frontier models .
Slash Command Queueing: Fire-and-Forget Workflows in Codex CLI
The single most frustrating friction point in any interactive coding agent is waiting. You know exactly what you want to do next — toggle fast mode, compact.
Codex CLI for the Sceptic: Honest Answers to 'Why Should I Bother?'
Every team has one: the developer who rolls their eyes when someone mentions AI coding tools. Perhaps you are that developer. You've seen the hype cycles.