From Prompts to Agents: Master Modern AI

ChatGPT, Gemini and Claude started out answering questions. Today, AI agents can plan, use tools and finish tasks on their own. This course covers all three stages: how Generative AI works, how to prompt it well, and how AI agents take action. Learners start with the foundations. They learn how AI, machine learning and deep learning differ, and how large language models from OpenAI, Google and Anthropic use tokens and context windows. They also learn why AI sometimes gives confident but wrong answers. Next, they build prompt engineering skills. They learn the six parts of a strong prompt (role, task, context, constraints, examples and output format) and practise few-shot, persona and chain-of-thought prompting, debugging and prompt chaining. They then explore Retrieval-Augmented Generation (RAG), embeddings and vector databases. From there they move into Agentic AI: memory, planning, tools, APIs, multi-agent systems and the Model Context Protocol (MCP). Real examples include Khan Academy's Khanmigo, GitHub Copilot and Waymo. Responsible AI, bias and prompt-injection risks are covered throughout. In the final project, learners build an AI study assistant. They start with a basic prompt and develop it into an agent that plans and adjusts a study schedule. Learning Outcomes > Explain how Generative AI and large language models power ChatGPT, Gemini and Claude. > Design structured prompts using role, task, context, constraints, examples and format. > Apply few-shot, persona, structured-output and chain-of-thought prompting. > Improve and evaluate AI outputs through refinement, debugging and prompt chaining. > Explain how RAG, embeddings and vector databases connect AI to external data. > Understand how AI agents use goals, memory, planning, tools, APIs and MCP. > Identify real-world uses and risks of AI, including bias and prompt injection. > Build an AI study assistant, from basic prompt to agent design. > Earn a certificate of completion.