📊 View Lecture Slides Full-screen presentation with navigation

Session 1: The AI-Native Workspace

Session Duration: 2 Hours     Block: 1 — Foundations & The Exocortex

Session clock

Minutes Mode Focus
0–50 Lecture Theoretical Foundation & Concepts
50–110 Core lab Environment Setup & Product Ideation
110–120 Checkpoint Pair share / show artifact

Note: Stretch work starts only after the Core checkpoint is completed.

Learning Objectives

By the end of this session, students will be able to:

  • Describe how cloud AI services eliminate local hardware barriers to software development.
  • Set up a functional AI-assisted development workspace using GitHub Copilot and Google AI Studio.
  • Articulate the difference between using AI as a search engine versus using it as a collaborative engineering partner.
  • Define a concrete, scoped AI product idea they will develop and iterate upon throughout this course.

Part 1: Theoretical Foundation — From Local Compute to Cloud Intelligence

1.1 The Old World vs. The New World

For decades, building intelligent software required a heavy investment in resources. Access to expensive hardware, specialised data science teams, and years of mathematical training were prerequisites. A university computer vision project, for instance, might have needed weeks of dedicated GPU time and a team of three PhD students just to build a baseline model. A simple product recommendation engine required data scientists, machine learning engineers, and complex production infrastructure.

That world no longer exists for most practitioners.

The modern reality: We have entered the era of the API-driven foundation model. These massive models, pre-trained on vast datasets, are now accessible via simple, lightweight API calls. You send text, an image, an audio file, or a document. You receive a sophisticated, context-aware response. The model runs on immense server clusters managed by providers like Google, meaning you never touch the underlying infrastructure.

This paradigm shift democratises software engineering. It changes the core question from “do I have the technical skills and hardware to build AI?” to “do I understand the problem well enough to direct an AI system towards solving it?”

1.2 What Cloud AI Actually Means

Understanding the terminology of cloud-based AI is essential for effective product engineering. Here is a breakdown of the core concepts you will encounter daily:

Concept What It Means Practical Implication
Foundation Model A large, versatile model pre-trained on massive datasets (e.g., Gemini). You do not start from scratch; you start with an entity possessing enormous general capability.
API (Application Programming Interface) A set of rules and protocols for building and interacting with software applications. You communicate with the model using structured, programmatic requests (often via JSON and HTTP) rather than a graphical interface.
Token The fundamental unit of text the model processes (often a word or a part of a word). Your operational costs and the model’s memory limits are measured in tokens. Efficient token usage is a key engineering skill.
Inference The act of running live data through a trained AI model to make a prediction or generate a response. You pay per use (per token processed during inference), not for the idle hardware.
Context Window The maximum amount of text (tokens) the model can “see” and process at one single time. This dictates the length of the conversation history or the size of the document you can ask the model to analyze.

1.3 AI as a Collaborative Engineering Partner

There are three levels at which most people interact with AI. Moving from Level 1 to Level 3 is the defining transition of an AI product engineer.

Level 1 — The Consumer (The Search Engine Paradigm)
Example: “Give me a recipe for pasta.”
The AI acts as an oracle. It provides a generic answer, and the interaction ends. The human is a passive recipient.

Level 2 — The Prompt Writer (The Task Delegation Paradigm)
Example: “Act as a nutritionist. Suggest a high-protein pasta recipe for someone training for a marathon.”
The output is significantly better, tailored by context and constraints. However, it is still essentially a one-shot request.

Level 3 — The Engineering Partner (The Iterative Paradigm)
This is the level this course operates at. Here, the AI is a collaborative peer. You engage in a continuous, iterative conversation where you:

  1. Describe a system or component.
  2. Ask the AI to generate a scaffold or draft.
  3. Review the output critically.
  4. Identify flaws, edge cases, or missing logic.
  5. Ask for a specific revision, test the result, and repeat.

The critical shift is understanding that AI does not replace the engineer’s judgment — it accelerates the engineer’s execution. As an AI Product Engineer, you are the architect. You still need to define the problem precisely, evaluate whether the output is functionally correct, anticipate failure modes, and ultimately integrate these AI-generated components into a cohesive, secure system.

1.4 The Course Framework: THINK → KNOW → BUILD → CONNECT → EQUIP → EVALUATE → SHIP

This module is organised around seven engineering stages that mirror how a professional team builds an AI-powered application from zero to production:

Stage What You Do in This Stage
THINK Use AI to explore problems, brainstorm features, and reason about architectural solutions.
KNOW Build a personal, structured knowledge base (your “exocortex”) that the AI can later retrieve information from.
BUILD Use AI as a pair-programming partner to scaffold the frontend and backend of your applications rapidly.
CONNECT Wire your applications to cloud AI APIs (like Google Gemini) to bring them to life.
EQUIP Give your AI tools and agentic capabilities, allowing it to execute code, search databases, or fetch real-time data.
EVALUATE Systematically test your AI system for reliability, hallucination rates, and security vulnerabilities.
SHIP Polish and deliver a functional, demonstrable AI prototype to end-users.

Every session maps to one or more of these stages. By Session 15, you will have navigated the full lifecycle of an AI product.

Part 2: Practical Labs — Environment Setup & Product Ideation

Lab 1.1 — Set Up Your AI-Native Development Environment

Your first task is to set up a professional-grade workspace. We will use GitHub Copilot as our in-editor assistant and Google AI Studio for testing foundational models and managing our API keys.

Step 1: GitHub Account and Copilot

  1. Create or log in to your GitHub account at github.com.
  2. Activate GitHub Copilot. It is available for free for students via the GitHub Education Pack, or you can sign up for a free trial.
  3. Install Visual Studio Code (VS Code) if it is not already installed on your machine.
  4. Open VS Code, navigate to the Extensions Marketplace (Ctrl+Shift+X / Cmd+Shift+X), and install the GitHub Copilot extension.
  5. Verify Copilot is active: Create a new file named test.js. Type function calculateTax(amount) { and pause. Copilot should offer a ghost-text suggestion to complete the function. Press Tab to accept it.
  6. Important Prerequisite: Ensure you have Node.js version 20 or higher installed. Open your terminal and run node -v. We will be building our backend with Node and Express starting in Session 5.

Step 2: Google AI Studio

  1. Navigate to aistudio.google.com in your web browser.
  2. Sign in with your Google account.
  3. Familiarise yourself with the interface by creating a new prompt: select “Create new prompt” → “Freeform prompt”.
  4. Get your API key: Click “Get API key” in the left sidebar and create a new key. Store this safely. You will need it in Session 7 to connect your application to Gemini. Crucial security note: Never paste API keys directly into public chats, GitHub repositories, or client-side HTML.

Step 3: Workspace Verification Checklist Before moving on, ensure you have:

  • VS Code installed with the GitHub Copilot extension active.
  • Node.js 20+ installed (node -v returns v20.x.x or higher).
  • A Google AI Studio account with a generated, securely stored API key.
  • A GitHub account ready for storing your project code.

Lab 1.2 — Define Your AI Product Idea

You will build one cohesive project throughout this module. Choose your idea wisely: it must be specific enough to be built as a prototype in 30 hours, but meaningful enough to solve a real problem.

The Product Idea Framework To define your product, you must answer these four foundational questions clearly:

  1. Problem: What specific, concrete problem does this application solve? Avoid generic statements.
  2. User: Who exactly is using this product? (e.g., “First-year engineering students” is much better than “people”).
  3. AI Role: What specific task does the AI perform in this product? Does it summarise? Does it classify? Does it retrieve answers from a specific document?
  4. Success Metric: How will you objectively know the product is working correctly?

Examples of Weak vs. Strong Definitions:

Question Weak Answer Strong Answer
Problem “Helps people learn things faster.” “Students waste hours re-reading dense lecture notes to find specific concepts; this tool instantly answers questions directly from those notes.”
User “Anyone who reads.” “First-year BSc Computer Science students at my university.”
AI Role “Uses AI to be smart.” “Embeds lecture notes into a vector database and answers natural language questions using only retrieved context to prevent hallucinations.”
Success “The app works well.” “The AI correctly answers 8 out of 10 benchmark test questions drawn from the notes, and cites the correct page number.”

Recommended Project Archetypes for this Course: If you are struggling for an idea, choose one of these proven archetypes:

  • Course Knowledge Assistant: Upload complex lecture notes; users ask questions and get cited answers.
  • Internal Document Navigator: Search through a collection of corporate reports, policies, or manuals.
  • CV-to-Job Matcher: Users upload a CV and a job description; the AI compares them, highlights missing skills, and explains gaps.
  • Multi-Document Summariser: Users paste multiple news articles or research papers; the AI outputs a structured, comparative synthesis.
  • Personal Research Assistant: Chat with a curated, local collection of saved web articles and PDFs.

Action Item: Record your four answers in a text document. You will refine this idea in every session as you learn new capabilities.

Checkpoint (10 min): Pair up with a neighbour. Pitch your Problem and User in 60 seconds. The partner must identify one potential weakness or area of ambiguity. Swap roles.

Key Takeaways

  • Cloud AI services provide access to massive, pre-trained foundation models via simple API calls, eliminating the need for expensive local hardware and specialised ML training.
  • AI is most powerful when utilised as an iterative, collaborative engineering partner (Level 3), not just a one-shot oracle for answering questions (Level 1).
  • This course follows a structured seven-stage engineering framework: THINK → KNOW → BUILD → CONNECT → EQUIP → EVALUATE → SHIP.
  • A strong AI product idea is defined by a specific problem, a tightly scoped user base, a clear and concrete role for the AI, and a measurable criterion for success.

Further Reading & Resources

  • Google AI Studio Documentation: ai.google.dev - Explore the official guides for Gemini capabilities.
  • GitHub Copilot for Students: education.github.com - Information on getting free access via the Student Developer Pack.
  • “The Pragmatic Programmer” by Andrew Hunt and David Thomas - A foundational text on the software engineering mindset, highly relevant even in the AI era.