Use what you learned in the previous lesson to solve real-world problems.

Check what you understood with a short quiz.

Build a prompt from a task, audience, constraints, and desired format. You will see how “write a summary” changes when you add length, tone, purpose, and output shape.

Use what you learned in the previous lesson to solve real-world problems.

Add background facts, source text, or user preferences that the model should use while answering. You will distinguish context that guides the answer from the actual task you want done.

Use what you learned in the previous lesson to solve real-world problems.

Check what you understood with a short quiz.

Use one or two examples inside a prompt to demonstrate the pattern you want. You will separate in-prompt examples from training data the model saw long before your request.

Use what you learned in the previous lesson to solve real-world problems.

Read a chat as a stack of messages with roles such as system, user, and assistant. You will see how role labels help set durable instructions, user requests, and generated replies apart.

Use what you learned in the previous lesson to solve real-world problems.

Check what you understood with a short quiz.

Follow how a follow-up question can rely on earlier turns in the same chat. You will identify when “make it shorter” or “use the same style” depends on conversation history as context.

Use what you learned in the previous lesson to solve real-world problems.

Treat tokens as the small chunks of text a model reads and writes, not as exact words. You will estimate why a short sentence, a long document, and generated output use different token amounts.

Use what you learned in the previous lesson to solve real-world problems.

Check what you understood with a short quiz.

Reason through what fits inside the model’s context window: your prompt, chat history, source material, and the answer being generated. You will spot why very long conversations or documents may need trimming.

Use what you learned in the previous lesson to solve real-world problems.

Recognize model parameters as learned internal numbers that shape behavior after training. You will avoid confusing them with facts, files, prompts, or settings you can directly edit during a request.

Use what you learned in the previous lesson to solve real-world problems.

Check what you understood with a short quiz.

Tell internal model parameters apart from request settings such as maximum output length, temperature, or image size. You will name settings as controls on a single request, not changes to the trained model itself.

Use what you learned in the previous lesson to solve real-world problems.

Choose whether a request needs a text model, image model, or multimodal model that can handle more than one kind of input or output. You will compare capability, speed, cost, and fit for the job in plain language.

Use what you learned in the previous lesson to solve real-world problems.

Check what you understood with a short quiz.

Ask for output in a usable shape such as bullets, a table, JSON, a caption, or a short paragraph. You will connect format instructions to what you plan to do with the result next.

Use what you learned in the previous lesson to solve real-world problems.

Create an image prompt by naming the subject, action, setting, style, composition, and important constraints. You will see how small wording changes can steer what appears in the generated image.

Use what you learned in the previous lesson to solve real-world problems.

Check what you understood with a short quiz.

Compare image outputs made from the same or similar prompts and name what changed: subject details, style, layout, lighting, or unwanted artifacts. You will treat each image as a candidate result rather than a guaranteed match.

Use what you learned in the previous lesson to solve real-world problems.

Spot hallucinations as confident-sounding outputs that are false, unsupported, or invented. You will practice flagging fake citations, wrong details, and answers that fill gaps instead of admitting uncertainty.

Use what you learned in the previous lesson to solve real-world problems.

Check what you understood with a short quiz.

Lower hallucination risk by giving source material, asking the model to use only that source, requesting uncertainty when needed, and checking important claims elsewhere. You will learn what prompting can improve and what still needs verification.

Use what you learned in the previous lesson to solve real-world problems.

Treat prompts as data you send to a service, especially when they include names, secrets, private documents, or business information. You will decide what to remove, anonymize, or avoid sharing before making a request.

Use what you learned in the previous lesson to solve real-world problems.

Check what you understood with a short quiz.

Review this chapter with practice based on your mistakes.