OpenAI recently released a new prompt guide for general users, aiming to lower the interaction barrier for non-developers. The guide integrates the regular ChatGPT interface with Codex into a single framework, with its core philosophy advocating "start small and directly express your needs." This is completely different from the style of the recently released GPT-5 and GPT-5.5 developer documentation, which focuses on API parameters and complex prompt patterns.
Not long after the release of this guide, OpenAI launched an independent product ChatGPT Work, built on Codex technology and the new GPT-5.6 model. To support this layout, the new guide revolves around four optional modules: goals, context, output format, and boundaries, suggesting users "replace step-by-step scripts with constraints," directly show the final result, and only introduce third-party data sources like spreadsheets or PDFs when necessary.
Additionally, the guide clearly distinguishes between the "chat" mode for quick questions and the "work" mode for handling heavy tasks with multiple data sources. On the code assistant Codex side, OpenAI introduced "Steer (turning)" and "Queue (queuing)" intervention mechanisms, and made it run in a sandbox environment that limits file and network access. It also added /plan and /goal commands to enhance multi-step project planning and code review.
This guide not only standardizes the user operation path but also marks that generative AI is accelerating its transition from "command fine-tuning" to a practical stage driven by intent.
