How to Use Codex for Work: My Complete System at OpenAI | Jason Liu

2026-07-26 · Source: Behind the Craft · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Intermediate, extended

Summary

Jason Liu, a DevEx engineer on OpenAI's Codex team, details his comprehensive system for integrating Codex into his daily workflow, as shared in a recent episode. He utilizes Codex to create a "chief of staff" thread that autonomously reviews and prioritizes communications from Slack, email, Twitter DMs, and Linear at 9 am, 1 pm, and 5 pm, even drafting replies. Liu distinguishes between Codex, a coding agent, and ChatGPT Work, a UX-focused interface, primarily using "sole medium" for organizational tasks and "extra high" or "ultra" for complex app development. His system involves organizing projects as pinned threads, maintaining an Obsidian vault for context, and automating tasks like flight check-ins. He also trains Codex to adapt its writing style for various audiences and regularly prompts it to clean up unused skills. Liu highlights the power of browser and computer use for tasks such as video editing and research, and demonstrates building a personalized drum learning application through iterative goal-setting and specific feedback. He notes the ongoing shift of Codex automations to the ChatGPT Work cloud experience.

Key takeaway

For Automation Engineers or Prompt Engineers seeking to maximize AI productivity, integrate Codex or similar AI agents as a central operating system. Establish persistent "chief of staff" threads for communication management and leverage AI's memory to build self-improving skills. Focus on providing specific, iterative feedback to refine AI outputs and automate complex tasks, from drafting emails to developing custom applications, ensuring your AI adapts to your unique workflow and preferences.

Key insights

Codex can automate complex workflows and adapt to personal preferences through iterative feedback.

Principles

• AI agents thrive on specific, iterative feedback.
• Persistent context (pinned threads, vaults) enhances AI utility.
• Distinguish AI tools by their core UX/capabilities.

Method

Build AI-driven workflows by defining clear goals and plans, then iteratively refine the AI's output with specific feedback, allowing it to learn and self-improve over time.

In practice

• Create a "chief of staff" AI for daily comms triage.
• Train AI to mimic your writing style for different audiences.
• Use AI to build custom learning or prototyping apps.

Topics

• OpenAI Codex
• AI Agent Development
• Workflow Automation
• Prompt Engineering
• Personal Productivity
• Browser Automation

Best for: AI Engineer, Automation Engineer, Prompt Engineer

Related on AIssential

• Full Workshop: Setting Yourself Up for Success —Jason Liu, OpenAI Codex — Codex enables AI agents to manage complex tasks and information through persistent memory and cross-application control.
• Codex for (almost) everything — Codex now functions as an autonomous, context-aware agent across the entire software development lifecycle.
• Codex-maxxing — AI agents become powerful when integrated into durable operating loops with shared memory and continuous interaction.
• OpenAI's Codex can now watch you work once and repeat the task forever — OpenAI's Codex enables AI agents to learn and automate complex workflows through single-demonstration recording.
• Inside How OpenAI Uses Codex to Do Product Work | Rohan Varma — OpenAI's Codex transforms product management by automating routine tasks, enabling rapid prototyping, and fostering ambitious, agent-driven development.
• The Codex App Has Upended My Daily Workflow — The Codex app transforms coding into managing AI agents, enabling parallel work and sustained context through a desktop interface.

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 Behind the Craft.