Lab 08 - AI Tools

In this lab, we will explore AI-driven development tools and their applications in building modern web applications. You will learn about the following tools:

• Generative AI: The broader category of AI technologies that includes tools like ChatGPT, Gemini, and Claude.
• Prompting strategies: How to ask generative AI the “right kind” of questions — by defining roles, goals, constraints, and output formats — in order to obtain useful, actionable answers.
• V0/Lovable: A tool for rapid UI prototyping and component generation.
• Cursor: An AI powered IDE for coding and debugging.
• ChatGPT API: A conversational AI API for natural language processing.
• GitHub Copilot: An AI coding assistant for accelerating development.

By the end of this lab, you will build a simple Next.js chat application, with UI components created using V0/Lovable and all of this logic implemented using Cursor (or other AI powered IDE).

Generative AI refers to artificial intelligence systems that can generate new content, such as text, images, videos, or audio, based on patterns they’ve learned from existing data. Unlike traditional AI models that classify or predict based on input data, generative AI creates new and unique outputs.

Core Capabilities

• Text Generation: Writing human-like text for applications like chatbots, content creation, and summarization.
• Image and Video Generation: Creating visuals with tools like DALL-E and Stable Diffusion.
• Audio Generation: Composing music or synthesizing realistic human speech.

Applications of Generative AI

• Chatbots and Virtual Assistants: Used in customer service and personal assistants (e.g., ChatGPT, Alexa).
• Content Creation: Writing articles, generating designs, or producing marketing materials.
• Creative Tools: Assisting artists, musicians, and designers in generating innovative ideas and assets.

Overview

ChatGPT is one of the most well-known examples of generative AI, developed by OpenAI. It uses large language models like GPT-3.5 and GPT-4 to understand and generate human-like text. ChatGPT is versatile and can handle tasks ranging from answering questions to writing essays or generating code.

• Human-Like Conversations: Engages in meaningful and context-aware discussions.
• Flexibility: Supports various use cases, including education, entertainment, and business.
• Customization: Can be tailored to specific tasks with system prompts and fine-tuning.

Overview

Gemini is Google's generative AI model, designed to combine advanced language understanding with cutting-edge capabilities in vision and multimodal reasoning. Released as part of Google DeepMind's AI ecosystem, Gemini aims to bring generative AI into various professional and creative domains.

• Multimodal Capabilities: Understands and generates content across text, images, and other formats.
• Enterprise Integration: Tailored for applications in business, including document analysis and automation.
• Advanced Context Understanding: Excels in maintaining context across complex interactions.

Overview

Claude is a conversational AI developed by Anthropic, designed with a focus on safety, ethical behavior, and robust reasoning. It competes with ChatGPT in providing conversational and creative AI solutions.

• Safety-Oriented Design: Built with principles to minimize harmful outputs and prioritize ethical usage.
• Context Handling: Handles long and complex conversations effectively.
• Business Applications: Supports automation, knowledge retrieval, and customer engagement.

Generative AI tools (such as ChatGPT, Gemini, Claude, etc.) are only as useful as the questions you ask them. This section presents simple prompting strategies that help you obtain clearer, more actionable answers — whether you are coding, learning a new topic, writing, or planning a project.

Avoid very generic questions like:

• “Explain this.”
• “Help me with my project.”

Instead, tell the AI exactly what you want:

• “Explain this concept so that a first-year CS student can understand it.”
• “Help me refactor this function to be easier to read and test.”
• “Suggest a study plan for passing an algorithms exam in 3 weeks.”

You can think in terms of:

• Goal – What do you want at the end? (e.g. explanation, plan, piece of code, checklist)
• Domain – What is this about? (e.g. web development, statistics, writing)
• Action – What should the AI do? (explain, summarise, compare, generate, debug, critique, etc.)

The same question can have very different good answers depending on context. For example, instead of:

• “Explain React hooks.”

you can say:

• “Explain React hooks to someone who already knows basic JavaScript but has never used React. Focus on useState and useEffect, and give one short code example for each.”

Useful types of context:

• Your current level: beginner / intermediate / advanced
• What you already know: “I know X, but I am confused about Y”
• Where it will be used: exam prep, documentation, class presentation, code in a specific project

The more relevant context you give, the more tailored the answer will be.

You can ask the AI to “act as” a particular kind of expert. This helps it tune the style and level of the response.

Examples:

• “Act as a university teaching assistant for a programming course.”
• “Act as a senior backend engineer reviewing my Java code.”
• “Act as an English teacher helping me improve my writing.”

This does not change the underlying model, but it often leads to more useful answers, especially for explanations or feedback.

If you know how you want the answer to look, say it explicitly.

Examples of constraints:

• Length: “Keep the answer under 10 bullet points.”
• Style: “Use simple language and avoid jargon.”
• Time/space: “Explain this in a way that would fit on a slide.”

Examples of output formats:

• “Answer as a list of bullet points.”
• “Output a step-by-step plan numbered 1, 2, 3…”
• “Return the result as JSON with fields `name`, `description`, and `difficulty`.”
• “Show the final answer first, then a short explanation.”

This is especially useful when you plan to copy the answer into code, documentation, slides or another tool.

You can ask the AI to structure the answer in multiple layers:

• “First give a short summary (3 bullet points), then a more detailed explanation.”
• “List the main ideas first, then examples for each idea.”
• “Provide a high-level explanation, followed by a concrete example.”

This helps you quickly get an overview, and then dive deeper only if needed.

Good use of AI is rarely “one prompt and done”. Treat the AI as a collaborator and refine the answer:

• “This is good, but explain point 3 in more detail.”
• “Can you give a simpler example?”
• “Rewrite this explanation as if I were 15 years old.”
• “Can you show the same function, but in TypeScript instead of JavaScript?”

Each follow-up can:

• clarify something you did not understand,
• adapt the answer for a different audience,
• or focus on the part that matters most to you.

Generative AI can be wrong or incomplete. You can explicitly ask it to show its own uncertainty:

• “List the assumptions you are making in this answer.”
• “Tell me what information you are missing that could change the answer.”
• “Highlight which parts of this answer might be less reliable.”

This is useful when you use AI for decisions, research, or learning and want to be aware of possible gaps.

Below are several example prompts you can adapt for your own use.

a) Learning a new concept

Act as a university tutor in computer science.
I am trying to understand [concept].
1) Explain it in simple terms.
2) Give a small concrete example.
3) Mention the most common mistakes students make with this concept.
Keep the explanation under 15 lines.

b) Asking for code help

Act as a senior [language/framework] developer.
I will paste a piece of code.
Tasks:
1) Explain what the code does.
2) Point out any potential bugs or edge cases.
3) Suggest improvements for readability and structure.
4) Show an improved version of the code.

c) Improving a piece of writing

Act as an editor.
I will paste a short text in English.
Improve grammar, clarity, and style, but keep my original meaning and tone.
Then briefly explain the most important changes you made.

d) Planning and organisation

Act as a study coach.
I need to prepare for an exam in [subject] in 3 weeks.
I can study about 1.5 hours per day.
Create a weekly study plan with concrete tasks for each day.
At the end, list 3 tips on how to stay consistent.

These examples illustrate the typical structure of a good prompt:

• define a role (“act as …”),
• provide enough context (level, constraints, domain),
• specify clear tasks, and
• optionally request a specific output format.

Using these strategies will make generative AI a much more effective assistant for your projects, studies, and everyday tasks.

V0 is a visual development platform that enables rapid prototyping and development of user interfaces. It allows developers and designers to create UI components visually, reducing the need for manual coding of UI elements. (similar tools: lovable, bolt.new, Replit AI)

V0 emerged in the landscape of low-code/no-code development tools, which have gained popularity in recent years to bridge the gap between design and development. The goal of such tools is to accelerate the development process and make it accessible to a wider audience, including those with limited coding experience.

• Rapid Prototyping: V0 allows for quick creation of user interfaces using a visual editor, which speeds up the design and iteration process.

• Collaboration: Designers and developers can work together seamlessly, as the tool provides a common platform for UI development.
• Code Export: V0 can export code in various formats (e.g., React components), which can be integrated into existing projects.

In this lab, you will use V0 to design the frontend of your chat application, creating components like the chat window, input field, and send button. This will allow you to focus on functionality without spending excessive time on UI coding.

Cursor is an AI-powered Integrated Development Environment (IDE) designed to enhance developer productivity through intelligent code suggestions and assistance. Cursor uses machine learning models to provide context-aware code completions, refactoring suggestions, and error detection.

The development of Cursor is part of the broader trend of integrating AI into software development tools. With advances in machine learning and natural language processing, tools like Cursor aim to augment the capabilities of developers, allowing them to write code more efficiently and with fewer errors.

• AI-Assisted Coding: Provides real-time code suggestions based on the context of your current work.

• Error Detection and Debugging: Helps identify potential issues in your code before runtime.
• Productivity Boost: Automates repetitive tasks and reduces the cognitive load on developers.

You will use Cursor to implement your chat application. The AI assistance will help you write efficient code.

GitHub Copilot is an AI-powered code completion tool developed by GitHub in collaboration with OpenAI. It uses OpenAI's Codex model, which is trained on a vast dataset of public code repositories, to provide code suggestions and generate code snippets based on comments and existing code. Now copilot support multiple LLM's from OpenAI, Antropic or Google

Released in technical preview in June 2021, GitHub Copilot represents a significant advancement in AI-assisted programming. It aims to assist developers by suggesting code snippets and functions, thereby reducing the amount of boilerplate code they need to write manually.

• Intelligent Code Suggestions: Generates code based on context, comments, and function names.
• Support for Multiple Languages: Works with various programming languages, including Python, JavaScript, TypeScript, Ruby, and Go.
• Learning from Context: Adapts to the coding style and patterns of the developer over time.

While GitHub Copilot is not a primary tool in this lab, it can be used to assist with coding tasks, providing suggestions and speeding up development as you work on your project.

ChatGPT API provides developers with access to OpenAI's powerful language models, such as GPT-4, GPT-4.1, GPT-5, GPT-5.1, enabling the integration of natural language understanding and generation into applications.

OpenAI released GPT-3 in June 2020, which was followed by the development of ChatGPT, a model fine-tuned for conversational interactions. The ChatGPT API was made available to developers to harness the capabilities of these models in their own applications, allowing for a wide range of use cases, from chatbots to content generation.

• Natural Language Understanding: Capable of interpreting and processing human language input.
• Conversational Responses: Generates coherent and contextually appropriate responses in a conversational format.
• Customization: Developers can guide the model's behavior using system prompts and parameters.

1. Download the generated project se-lab8-tasks.zip (run `npm install` and then `npm run dev`).
2. Read the section “Using Generative AI Effectively: How to Ask Good Questions” from this lab. Then, interact with at least one generative AI tool (ChatGPT, Gemini, Claude, etc.) and:
   • use it to refine the requirements and UX for your chat application, and
   • use it to generate or improve a part of your work (e.g., code snippet, test, error message copy, or help text)
3. Use V0 to design the frontend of your chat application:
   • chat window
   • message list
   • input field
   • send button
4. Implement the chat logic of your application using Cursor / V0 / ChatGPT / other tools:
   • you may use a simple local/mock backend or hardcoded responses,
   • you are encouraged to use AI tools while coding,
   • you do NOT have to integrate any external API (e.g., ChatGPT API) in your project.
5. Add a Privacy Policy / Terms of Service section to your application using a generative AI tool.
   • Clearly prompt the AI with the context of your app (what it does, what data it handles).
   • Adapt the generated text to your project if needed.
6. Bonus (optional): Use AI-generated images (Leonardo.AI, DALL·E, Stable Diffusion, etc.) for your chatbot avatar or UI elements.