Working with AI Assistants

About this course

A practical course on using AI assistants for real work, built around one idea: understanding what the tool actually is predicts what it will do badly, and that is most of the skill. It starts with the mechanism, which explains almost everything surprising about these tools: why they are fluent and sometimes wrong, why they invent citations confidently, why they agree with you too easily, and why the same question can produce different answers. Then an honest account of what they are genuinely good at and where they reliably fail. The middle covers prompting: the four parts that carry the weight, why negative constraints are so effective, what actually helps and what is folklore, and how to iterate without going backwards. It also names the tells that make AI-written text recognisable, because avoiding them is part of using these tools well. The last section is about real work. Generated code that looks right and is subtly wrong, with an in-browser exercise reviewing exactly that. Then what not to paste in, when disclosure matters, where assistance becomes misrepresentation, and how to keep your own judgement sharp while using the tools daily. No specific product is named, and nothing here depends on one. The material is about the category, and it applies to whichever assistant you have in front of you.

What you will learn

• Explain what a language model does, and predict where it will fail
• Judge when output is verifiable enough to trust and when it is not
• Write prompts with context, task, constraints and format, and iterate on them productively
• Recognise the writing patterns that make AI-generated text obvious
• Review generated code for the failures it specifically tends to have
• Handle privacy, disclosure, and the line between assistance and misrepresentation

Before you begin

• No experience with these tools needed
• The code review exercise assumes basic JavaScript, and the rest of the course does not
• Access to any AI assistant is useful for practice, and is not required