5-Day AI Agents: Intensive Vibe Coding Course

Concept-first study guide — understand the ideas, then use them with any tools (June 15–19, 2026)

What this site is for

This site explains the ideas behind the course in plain language. Google tools (Antigravity, Gemini, ADK) are used in the labs, but the concepts work with other products too — like using the same idea of “a backpack” whether you buy from different brands.

Each day page teaches:

  1. Big ideas — what the words mean (agent, IDE, MCP, etc.)
  2. Tool categories — what type of tool you are using, and what else exists in that category
  3. Course materials — links to the official podcast, whitepaper, codelabs, and livestream

Recommended order from the course: podcast → whitepaper → codelabs → livestream.

The story of the whole week (one sentence per day)

  1. Day 1: Learn what an AI agent is and how “vibe coding” lets you build software by describing what you want.
  2. Day 2: Connect agents to the outside world (files, APIs, other agents) using shared plug-in standards.
  3. Day 3: Teach agents how to do jobs without overloading their memory.
  4. Day 4: Check that agents are safe, honest, and actually doing the right steps — not just lucky answers.
  5. Day 5: Turn a working demo into something real people can use reliably in production.

Starter glossary

Short definitions you will see all week. Each day page goes deeper.

AI agent

A program that can think, choose, and act to finish a goal — not just answer one question. It might read files, call tools, run code, and try again if something fails.

Like a helper who can use your computer and the internet, not just chat.

LLM (Large Language Model)

The “brain” that understands and generates language. Examples: Gemini, GPT, Claude, Llama. The agent uses an LLM to reason; other parts give it hands, memory, and rules.

Vibe coding

Building software by describing what you want in normal language and letting an AI write or edit the code. Fast for prototypes; still needs checking before serious use.

IDE (Integrated Development Environment)

A program for writing code — usually a text editor plus extras (file tree, run button, debugger, AI assistant). This course uses Antigravity IDE; the same category includes VS Code, Cursor, JetBrains IDEs, and Xcode.

CLI (Command-Line Interface)

You type text commands in a terminal instead of clicking buttons. Good for automation, servers, and power users. This course uses Antigravity CLI (agy); similar tools include Gemini CLI, Claude Code, GitHub Copilot CLI, and plain shell scripts.

MCP (Model Context Protocol)

A shared plug standard so any AI client can connect to any “tool server” the same way. Stops you from writing a custom connector for every tool × every AI app.

Like USB-C: one port shape, many devices.

Agent framework

Code libraries and patterns for building agents (workflows, tools, memory). This course uses ADK (Agent Development Kit). Other frameworks: LangChain, CrewAI, Microsoft AutoGen, OpenAI Agents SDK.

Daily concept guides

Day 1 — Agents & Vibe Coding

What agents are, how vibe coding changes software building, IDEs, and getting an app online.

Day 2 — Tools & Interoperability

CLIs, APIs, MCP, and standards so agents can talk to the world.

Day 3 — Agent Skills

Memory, playbooks, workflows, and not confusing the AI with too much text.

Day 4 — Security & Evaluation

Trust, testing agent behavior, human approval, and safe coding habits.

Day 5 — Spec-Driven Production

Specs, CI/CD, deployment, and running agents for real users.

Official course links

Kaggle course page kaggle.com/competitions/5-day-ai-agents-intensive-vibecoding-course-with-google Course discussion & daily assignments kaggle.com/.../discussion YouTube livestream playlist youtube.com/playlist?list=PLqFaTIg4myu8AFXUjrVhDkUGp0A9kK8CX Kaggle Discord discord.gg/kaggle