Learning with LLMs
This course covers how to use a large language model as a study and teaching tool with Google Gemini. It starts with the five elements of a good prompt, persona, objective, audience, context and boundaries, together with the Master Prompt strategy and the follow-up prompts that close a specific knowledge gap. It then covers the limitations of LLMs, including hallucination, training data cutoff and the memory ceiling, and the habit of verifying key claims against reliable sources. The second part moves from single prompts to reusable tools: a Gemini Gem holds a fixed system instruction written in Markdown so that a phase-based study activity behaves the same way every session, while the Canvas tool and Google Sheets with Apps Script turn a conversation into a working web app that can be deployed and shared. The course finishes with a full build cycle on a self-chosen HKDSE topic, in which participants ask Gemini for a plan-only proposal, cut the features that do not help, then generate, refine and submit the tool.
Learning objectives
- Write structured prompts using persona, objective, audience, context and boundaries, and refine the output with follow-up prompts.
- Identify where LLMs fail, including hallucination, knowledge cutoff and the memory ceiling, and verify key claims against reliable sources.
- Build and share a learning tool as a Gemini Gem, a Canvas web app or a Google Sheets Apps Script web app.
Lesson plan
Prerequisites
- No coding experience is required
- A Google account for signing in to Gemini and Google Drive
Software
- Google Gemini (Fast / Thinking / Pro modes, Gems, Canvas, Create Image)
- Google Sheets with Apps Script