Learning Timeline
Key Insights
Agent Mode vs. Standard Chat
Unlike standard chat mode which only generates one page at a time, Agent mode can build an entire application structure with multiple interconnected files and logic within the sandbox.
Language and Visual Considerations
Since GLM-5 is developed by Zhipu AI, text within generated images may occasionally appear in Mandarin, or visual errors may occur with complex data (such as the Periodic Table). You may need to provide a corrective prompt if this happens.
Automatic Progress Tracking
The Agent automatically builds the logic to track user progress (e.g., 'Course Completed' or checkmarks on the lesson list) without you needing to write the code manually.
Prompts
Generating a Complete Educational Course App
Target:
GLM-5 Agent
Plan a chemistry course structure with multiple kid-friendly lessons. Generate engaging chemistry images and build out the lessons and exercises, add interactive visual exercises, and then test and finalize the application in a sandbox environment.
Step by Step
Building a Multi-Page Course App with GLM-5 Agent
- Access the GLM-5 platform and ensure you select 'Agent' mode (not standard chat mode) to enable multi-step autonomous tasks.
- Enter your main prompt in the input box, specifying the course topic, target audience (e.g., children), and required elements like quizzes and images.
- Monitor as the AI automatically generates a 'to-do list'; the Agent will plan the course structure, generate images using the GLM-Image model, and build interactive exercises.
- Allow the AI to work within the 'sandbox environment' to build the code for every page (index, lesson pages, exercises) simultaneously.
- Wait until the Agent completes the 'testing' phase to ensure there are no errors in the generated application.
- Click the 'Start' button on the application displayed in the preview to begin the course.
- Navigate through each chapter by clicking 'Next Lesson' to test the flow between pages.
- Test interactive elements such as quiz buttons, drag-and-drop animations, and feedback pop-ups to ensure the application logic is fully functional.
