🛰️ What is AntiGravity
O AntiGravity is a code editor made by Google, powered by AI agents. The difference from traditional editors is that you don’t type code line by line: you describes it in natural language what you want, and the agent writes, edits, and organizes files directly on your machine.
🧭 The core idea
Think of AntiGravity as a developer who understands Portuguese: you give the instruction, it operates at the computer level (creates folders, writes files, runs the project) and shows you the result.
- •Runs locally — the code stays on your machine, not in a locked-down cloud.
- •Uses powerful models (such as Gemini 3 Pro, Claude, and others) — you choose the best one for each task.
- •Has built-in features: managing multiple agents, browser control, and self-correction.
runs on your machine
run for you
you just describe
chooses the best
🎂 The 3-layer cake
The easiest way to understand where AntiGravity fits is to think of a three-layer cake. Each layer solves the same problem — automating and building — but with a different balance between ease e control.
Top — visual tools (Make, n8n)
Easier, less control
You drag modular blocks. Great for simple, predictable automations. It’s a polished interface on top of code.
In between — cloud vibe coding
You speak, the cloud hosts
Platforms that build and host for you without touching your computer. They simplify things, but use fixed structures.
Base — local environment (AntiGravity)
Full control and functionality
The code itself, on your machine. It’s the “final layer”: everything the others do, you can do here without limits.
💡 Practical tip
Absolute beginner? Start by experimenting in the cloud layer with something simple to get a feel for "vibe coding." Then move down to local (AntiGravity), where you get full control.
drag blocks
the AI hosts
full control
ease × power
⚖️ Deterministic vs. agentic
The most important decision before building anything is: is the problem fixed, or does it change? This separates the deterministic world from the agentic world.
✓ Deterministic (no-code)
- ✓Repetitive, predictable task
- ✓Steps always in the same order
- ✓Ex.: "completed form → adds to the email list"
- ✓Ideal tool: Make, n8n
✗ Don’t force no-code when…
- ✗The problem is complex and changes
- ✗Needs its own interface and logic
- ✗Ex.: an app that receives an image and transforms it to fit your needs
- ✗Then the way forward is agentic (AntiGravity)
In mode agentic, AI doesn’t follow a fixed script: it reasons about the goal, writes the code, tests it, sees the error, and fixes it—repeating until it works. That’s what unlocks problems a block-based workflow can’t solve.
→ deterministic
→ agentic
vs. reasoning
test and fix
🧱 Code is the final layer
Every new AI tool ultimately generates or manipulates code. That's why, when you learn to work at the code layer (with AI doing the heavy lifting), you reach the foundation — and you won’t be at the mercy of the next trendy tool.
🛡️ Why this protects you
Visual tools come and go. Code remains the foundation of everything. Mastering the final layer is a antidote to FOMO: you don’t have to chase every release, because you’re already where anything is possible.
And don’t be alarmed: you don't need to become a programmer. The goal is to guide AI clearly, not memorize syntax.
doesn’t become obsolete
less anxiety
not to memorize
durable skill
🆚 AntiGravity vs. Cursor & other IDEs
AntiGravity isn’t alone: Cursor e Windsurf are very similar AI IDEs. This course uses AntiGravity for simplicity, but the principles apply to all.
| Resource | AntiGravity | Other IDEs |
|---|---|---|
| Browser control | native | often via a plugin |
| Several parallel agents | native | varies |
| Self-correction | embedded | varies |
| Usage principles | the same ones | the same ones |
The lesson: you learn a concept (directing agents, planning ahead, reviewing), not a specific button. Switching tools later is easy.
not the button
browser + agents
easy switch
one course, one app
🚫 When NOT to use AntiGravity
A good professional knows when no use the powerful tool. Not every problem needs to become an app.
✓ Use AntiGravity when
- ✓You need your own interface
- ✓The logic is complex or changes
- ✓You want full control and functionality
✗ Prefer no-code when
- ✗It’s a simple, stable automation
- ✗The steps never change
- ✗Speed matters more than customization
⚠️ Common pitfall
Wanting to turn EVERYTHING into an app out of enthusiasm. Result: you spend hours on what a no-code workflow could solve in minutes. Always ask: "Does this change? Does it need its own screen?".
→ no-code
→ app
time and energy
does it change? Is there a screen?
🧭 How to use this course
Each module follows the same recipe so you can learn methodically and get straight to work.
Read the topics
Concepts with examples and diagrams. Mark each section as read — your progress is saved in the browser.
Do the example and the exercise
Every module has a practical example and a hands-on exercise. That’s where the knowledge sticks.
Use the ready-made prompts
Copy and adapt. Track 5 brings the entire library of skills and prompts together in one place.
◷ Track your journey
Use the button "Journey" at the top to see your progress, review flagged questions, and your notes. Select any paragraph to highlight or annotate it.
mark as read
example + exercise
ready-made prompts
journey
🔬 Practical example
Imagine you want to notify your customers when a product’s price drops. Where does this fit in the 3-layer stack?
- 1.Just “checked the price → sent an email”? Fixed workflow → visual layer (no-code).
- 2.Want a polished dashboard with history, charts, and login? Becomes an app → local layer (AntiGravity).
The same idea can live in different layers. What matters is its complexity and whether you need a custom screen.
🏋️ Exercise
List 5 of your ideas (tasks you’d like to automate or apps you’d like to have). For each one, answer:
- É deterministic (fixed) or agentic (changes)?
- Which layer which part of the pie does it fit best (visual, cloud, local)?
- Need own screen? If not, maybe it doesn’t even need to become an app.
Save this list: we'll build one of these ideas in Track 2.
⚡ Ready-made prompts
Copy, paste into your AI assistant, and adapt to your use case.
Tenho a seguinte ideia: "[descreva a ideia]". Me ajude a decidir: 1) É um problema determinístico (fixo) ou agentic (muda)? 2) Em qual camada se encaixa: ferramenta visual (Make/n8n), vibe coding na nuvem, ou ambiente local (AntiGravity)? 3) Precisa de interface própria? Justifique cada resposta em 1 frase e recomende a ferramenta.
Explique, em linguagem simples e com uma analogia, a diferença entre automação determinística (Make/n8n) e construção agentic (um IDE com agentes de IA como o AntiGravity). Use o exemplo de "[seu contexto: ex. uma loja online]".
✅ Module summary
Next module:
1.2 — No-code automation vs. AI apps