Implementing the capstone system
Throughout this syllabus we have been working toward a single goal: learning how to build AI agents. Rather than beginning with complex systems, we started with Python fundamentals and gradually added the pieces needed to create autonomous software.
What we learned along the way
The syllabus followed a clear progression from basic programming toward full AI-enabled systems.
We began with Python foundations, learning how programs execute, how data is represented, and how functions and modules help structure real software.
We then moved into practical programming, where Python programs interact with the outside world through files, libraries, and network communication.
Next we explored data science and AI tools, introducing numerical computing, data analysis, and basic machine-learning ideas that appear frequently in modern software systems.
From there we built classical agents using long-running loops, program state, decision logic, and tools that perform actions.
We then extended those ideas into LLM agents, where language models assist with reasoning, planning, and selecting tools.
After that we examined the OpenAI Agents SDK, which provides structured abstractions for defining agents, tools, and memory.
Finally we explored multi-agent systems, where separate agents coordinate their work through clearly defined roles.
The capstone application combines all of these ideas.
The capstone multi-agent application
The capstone system is a manager-worker multi-agent application written in Python.
At the center of the system is a manager agent. The manager asks a language model to plan a set of tasks that will achieve a goal. These tasks are then assigned to worker agents.
Workers carry out tasks by invoking tools, which are Python functions that perform concrete actions such as writing files or generating HTML.
Within this system, the language model helps plan tasks, research topics, and generate text, while the Python code manages execution and artifact creation.
The generated solar system website
The result of the capstone system can be seen by visiting the generated website:
Let’s Explore the Solar System
The pages are generated by the agents, which research the topics and produce the HTML content automatically.
The website itself is intentionally simple. The goal of this project has been to build the AI system that generates the content rather than to design a sophisticated web layout.
A small amount of manual styling has been added to the published version of the site. Automatically generating a well-designed website is a more advanced challenge and lies beyond the scope of this capstone system.
The generated solar system website is a demonstration that we can learn to build AI agents by building them step by step.