So far, you’ve learned how to use Codex to write, edit, and organize code through natural language and session commands. You’ve seen how it can refactor files, create new ones, and follow your project’s rules automatically.

Now, it’s time to go a step further. In this lesson, you’ll explore how Codex can see what you see and find what you need — by interpreting visual inputs and connecting to real-time web information. You’ll discover how to give Codex visual context with images and diagrams, and how to use web search integration to bring in the latest knowledge.

These capabilities make Codex a truly adaptive assistant that understands your work in full context — both on your screen and beyond it.

GPT-5-Codex can process both text and images, allowing it to understand the visual side of your development work. When you attach a screenshot, diagram, or chart, Codex analyzes it and combines what it sees with your text prompt to give a more accurate response.

This makes it easier to troubleshoot errors, interpret visual outputs, or even translate diagrams into code structures.

Example prompts:

• “Here’s a screenshot of my terminal — can you explain the error at the bottom?”
• “Look at this system diagram and tell me how the data flows.”
• “Analyze this chart — what’s causing the spike in response time?”

Visual context helps Codex act like a true collaborator — one that can “see” what’s on your screen. When you share a terminal error, a data visualization, or a wireframe, Codex can reason about what’s happening and guide you to a solution.

You might upload a stack trace image, ask Codex to identify the root cause, and get a fix ready in seconds. Or you could show a UI mockup and have Codex generate a matching React component layout.

Example prompts:

• “Here’s my chart — what can you tell about the performance trend?”
• “This is my UI wireframe — generate React components that follow this layout.”
• “Analyze this screenshot of test results and suggest what to fix first.”

Codex can also connect to the web to access real-time information. This means it can look up current documentation, the newest command syntax, or recent best practices — all within your coding session.

If you’re using a library that changed recently, you can ask Codex to find the latest examples and apply them. Or, if you need to configure a new tool, Codex can pull in the newest setup instructions.

Example prompts:

• “Search for the latest FastAPI version and show what changed in request handling.”
• “Find current best practices for async Python functions and refactor my code.”
• “Look up the most recent Docker Compose format and update my file.”

This integration lets Codex combine what it already knows about your project with the newest information available online — keeping you current and efficient.

In this lesson, you discovered how GPT-5-Codex can use both visual and web context to understand your work more deeply. You can now show Codex images or diagrams to give it visual awareness, and ask it to search the web when you need the latest information.

Next, you’ll practice using these features in action — attaching visuals and using live search to debug, design, and improve your projects with Codex.