What are the four building blocks OpenAI recommends for prompts?

OpenAI structures prompts around goal, context, output format, and boundaries. None are required; the company recommends adding them only for bigger tasks.

What is the difference between Chat and Work in this guide?

Chat is for quick questions and rewording, while Work is for tasks that pull in multiple sources, make changes, or produce larger deliverables like reports. Work tasks burn more credits but are recommended for time-saving or decision-support work.

What new controls are available in Codex?

Codex introduces "Steer" (adding a message to redirect the current run) and "Queue" (lining up a message for the next run), plus slash commands like "/plan" (propose an approach before changes), "/goal" (set a higher-level objective), and "/review" (run code reviews locally or via GitHub).

AI Business & IndustryTHE DECODERJul 13, 2026Jul 14, 2026, 04:01 JST

OpenAI tells users: stop overthinking prompts, start with what you want

Send on LINE

3 Key Points

What happened
OpenAI released a new prompting guide for ChatGPT and Codex that simplifies how users write instructions. Instead of detailed step-by-step scripts, the guide recommends starting with the desired result, adding optional building blocks (goal, context, output format, boundaries) only when needed, and using one or two hard constraints to block unwanted behavior rather than elaborate procedures.
Why it matters
The guide shifts away from complex prompt engineering toward a more intuitive approach—users spend less time crafting perfect instructions and more time iterating through follow-ups to refine output. For business users handling recurring tasks, the framework suggests testing prompts manually first before automating, and using ChatGPT's verification feature (e.g., checking that action items have owners and deadlines) for high-stakes work. This may make AI tools faster to adopt for teams without specialized prompt-writing skills.
What to watch
The guide distinguishes between Chat for quick questions and a product called "Work" for multi-source tasks and larger deliverables—Work consumes more credits but targets time-saving and decision support. For Codex (the coding assistant), new mid-run steering and queuing options, sandbox protections, and slash commands like "/plan" and "/review" let developers manage complex, multi-step projects.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, in your inbox every morning.

Get news like this every morning, free

Context & Analysis

OpenAI's guide reflects a deliberate shift in how the company frames AI use for non-technical business users. Rather than the developer-focused documentation for GPT-5 and GPT-5.5—which emphasized API parameters, reasoning-effort levels, and elaborate schemas—this end-user guide strips away jargon and complexity. The core message is that short, outcome-focused prompts often outperform lengthy scripts, and that iteration through follow-ups is the expected workflow, not a failure.

The introduction of the "Work" tier alongside Chat suggests OpenAI is segmenting by task weight and resource cost. Quick rewording stays in Chat; anything requiring multiple data sources, changes, or producing substantial output moves to Work. This framing lets users match tool choice to stakes and ROI—particularly useful for teams running recurring tasks that benefit from upfront manual refinement before automation.

For Codex users, the sandbox protections and mid-run steering options address a practical pain point in multi-step coding work: the ability to pause, redirect, or propose a plan before executing changes. The addition of GitHub integration ("/review" via @codex comment) extends the assistant's reach into existing development workflows, lowering the friction of adoption.

FAQ

What are the four building blocks OpenAI recommends for prompts?: OpenAI structures prompts around goal, context, output format, and boundaries. None are required; the company recommends adding them only for bigger tasks.
What is the difference between Chat and Work in this guide?: Chat is for quick questions and rewording, while Work is for tasks that pull in multiple sources, make changes, or produce larger deliverables like reports. Work tasks burn more credits but are recommended for time-saving or decision-support work.
What new controls are available in Codex?: Codex introduces "Steer" (adding a message to redirect the current run) and "Queue" (lining up a message for the next run), plus slash commands like "/plan" (propose an approach before changes), "/goal" (set a higher-level objective), and "/review" (run code reviews locally or via GitHub).

THE DECODERRead Original Article

Send on LINE

AI news that matters for your work, delivered every morning.

Pick your industry and the AI tools you use, and get news related to your work every day.

Get it free

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.