🌱 New here?
You don’t need to know how to program. Every technical word (AI, prompt, token, node, workflow…) is explained in a “What is X?” box the first time it appears. If you come across a term without a definition, it was defined earlier — use the index on the right to go back.
Before the seven habits, take a look at the shift that ties them all together. In the old mode, you gave step-by-step instructions to a predictable system. In vibe coding, you say result and an agent reasons through the approach — and the result may vary from one run to the next.
Illustrative diagram — two runs of the same request can reach the same goal by different paths and in different formats. That's expected, not a defect.
🗣️ What vibe coding is
What it is
Vibe coding it’s building software and automations by talking with an AI in natural language: you describe the result what you want, while AI handles the technical implementation. Instead of learning how each component works, you focus on the problem — AI handles tool selection, configuration, and debugging.
What is it?AI (artificial intelligence) — here, a model that understands and generates text. You talk to it in Portuguese; it responds and takes action. Automation — a repetitive task that starts running on its own (e.g., “every time an email arrives, log it in a spreadsheet”).
Example of the kind of request that works: "check my email every 10 minutes; if a new message arrives, notify me by direct message." AI produces a plan, asks clarifying questions (channel? direct message?), and then builds. You describe the destination; it plots the route.
Why learn
Natural language conversation guides both the build and iteration. When something needs to change, you don’t reopen a configuration editor: you just ask for the change. The AI does the “heavy lifting.”
- •Speed: from concept to working prototype in minutes, not hours.
- •Low technical overhead: without memorizing how each tool works.
- •Focus on the problem: the effort goes into the result, not the tool.
Key concepts
Natural language drives the build.
Just ask to make a change.
AI chooses and configures everything.
Fundamentals → technique → practice.
🎯 Describe the outcome, not the steps
What it is
The central principle of this mindset: say what should happen, not how do. “Identify technical questions and draft a response” is an outcome; “use an IF node with a text rule, then a code node…” is a technical step. You describe the destination; the AI chooses the road.
What is it?Node (node) — a little box that does one thing in an automation flow (read email, filter, send). Prompt — the message you write to AI. Here, the ideal prompt describes the result.
Why learn
When you dictate steps, you limit the AI to your (probably incomplete) knowledge of the tools. When you describe the outcome, it draws on its full range of knowledge to get there—and often finds a better path than you would have imagined.
✓ Describe the RESULT
- ✓"Check email every 10 min; if anything new arrives, let me know."
- ✓"Have it check every hour." (change in natural language)
- ✓Answer the AI’s clarifying questions.
✗ Don’t dictate the technical STEPS
- ✗"Add a trigger node, connect it to an IF, then to a code node…"
- ✗Micromanage each parameter setting.
- ✗Expecting the AI to guess without you answering its questions.
Key concepts
Say the “what.”
Format, name, where to save it.
Let the AI choose the "how".
Answer hers and approve.
🎲 Deterministic vs. nondeterministic
What it is
Traditional automation is deterministic: the same input always produces the same output—predictable, "boring and beautiful." Generative AI is non-deterministic: it reasons about alternatives along the way, so the same instruction can produce different names, structure, and layout each time.
What is it?Deterministic = a predictable, repeatable result (2 + 2 always equals 4). Nondeterministic = it can vary between runs, even when it meets the goal (like two cooks making the same dish in different ways).
Illustrative diagram — two valid variations don't mean there's an error; they mean the AI chose different paths to the same goal.
Why learn
If you don’t know this, you’ll compare your screen with someone else’s, see different names and layouts, and think something broke. Accepting variation changes how you evaluate things: you judge by result, not by how it looks step by step.
🎯 Practical tip
Don't compare “screen by screen” with tutorials or colleagues. If your version has different names but does the same thing correctly, it's right. Always evaluate the final behavior.
Key concepts
Traditional = always the same.
AI = can change each time.
Variation isn't a defect.
Evaluate the result, not the screen.
🔎 Trust but verify
What it is
Trust but verify ("trust, but verify") is the habit of always testing what AI created and checking the output against your expectations — even when no errors appear. The golden rule: green ≠ correct. A workflow can run without errors and still produce the wrong answer.
What is it?"Green" — when the tool reports the run as successful (with no error message). This only tells you that ran, not that the content is correct. Hallucination — when AI makes up information it doesn't have.
Why learn
AI makes mistakes — sometimes confidently. If you only look at the "green" status and don't check the content, the error carries forward. A good agent, by the way, refuses make up what it doesn't know; testing is what reveals whether it's doing that.
✓ Verify for real
- ✓Test with real scenarios (and edge cases).
- ✓Check numbers and facts against the source (30 days ≠ one year).
- ✓Expect the agent to say "I don't know" when data is missing.
✗ Don’t blindly trust the green light
- ✗Assume that “no errors” = “correct.”
- ✗Skip testing because "it seemed to work."
- ✗Accepting an invented answer without checking its source.
Key concepts
Running doesn’t mean getting it right.
Use real scenarios.
Check the numbers and sources.
"I don't know" is desirable.
🪜 Iterate in layers
What it is
Iterate in layers is to improve in steps (stair-step): one change at a time, testing after each step before moving on to the next. Instead of asking for ten improvements at once, you add one, validate it, and only then add the next.
What is it?Iterate — repeat a cycle of small improvements. Edge case — an unusual situation that can break the flow (e.g., an email without a signature, an empty field).
Why learn
Isolated changes are easy to test, revert, and diagnose. If you change five things and something breaks, you won’t know which one caused it. One layer at a time keeps the error easy to locate — and always confirm that the “happy path” still works before and after each change.
🎯 Practical tip
Before making any changes, make a backup (duplicate or export) and establish a tested baseline. That way, if something goes wrong, you can roll back without losing what already worked—and avoid blaming the new change for a problem that was already there.
Key concepts
Stack changes.
Validate at every step.
Back up before changing.
Confirm that it didn’t break.
🎓 AI as a learning tool
What it is
AI isn't just the one who builds — it's also the best teacher you have right there. Whenever it chooses a node, a setting, or an approach, ask “why”. This question turns every build into a mini-lesson and helps the knowledge stick with you.
What is it?"Set it and forget it" — the temptation to accept everything the AI does without understanding it. The opposite is what builds intuition: watching the build and asking why it made those decisions.
Why learn
Watching the AI's reasoning early has two benefits: you learn how the system works and can interrupt quickly if it goes off track — saving time and tokens. When you understand why, you can give better feedback and depend less on it.
Make the knowledge stick.
See the reasoning.
Stop if it goes off track.
Each build is a lesson.
👔 Who does what
What it is
The mental model that ties the previous six together: you’re the manager; the AI is the developer. You define the goal, provide context, approve the plan, and review the result. The AI, like an experienced developer, chooses tools, writes, debugs, and self-corrects. Thinking this way avoids micromanaging.
Illustrative diagram — the two-way relationship: you delegate outcomes and give feedback; the AI executes and returns them for review.
Why learn
Treating AI as a specialist (rather than as a tool that needs detailed instructions) brings out its best. And when making corrections, prefer feedback ("I like the structure, but the tone is too formal"), dry correction ("this is wrong") — feedback helps AI adjust its reasoning, not just patch the output.
Key concepts
Defines and approves.
AI executes.
Treat it as a specialist.
Better than a blunt correction.
Setting the mindset: which phrase is the right way vibe coding before asking?
📌 Module Summary
Next Module:
1.2 — The Ecosystem: How the Pieces Fit Together