Learning path map
Detailed content
🌅 Welcome: imagine and build
The introduction: what Claude Code is, the course mindset, and how to study.
Describing it in Portuguese is enough to build software.
The barrier is down; those who understand first get ahead.
Natural language → code · transferable skill.
Treat your ideas as executable projects.
The biggest obstacle is thinking “this isn’t for me.”
Imagination as specification · you’re the director.
A terminal agent that writes, tests, and fixes code.
Prevents false expectations and shows why it outperforms no-code.
Terminal · coding agent · abstraction layer.
Learn by building real things, with a project as the assessment.
Knowledge that becomes a project sticks; loose knowledge evaporates.
Learn by doing · portfolio · no filler.
Three progressive tracks + the Resource Library.
Seeing the map reduces anxiety and helps you choose your pace.
Fundamentals · Build · Operate & Profit.
Sequence, focus, and practice before moving forward.
The small tricks are where the gains come from.
Sequence · focus · living course.
🛡️ Fundamentals and Setup
Install, understand model×environment, permissions, context, and slash commands.
The model is the intelligence; the environment (IDE/terminal) is the armor.
Separating the two avoids confusion when configuring them.
Model · environment · IDE.
Download, authenticate, and open Claude Code.
Proper setup prevents 80% of early problems.
Download · sign in · first command.
Where commands, files, and chat live.
Knowing the interface speeds everything up later.
Prompt · files · history.
How much the agent can do on its own.
Balances speed and security.
Approve · auto-accept · risk.
The AI’s working memory limit per session.
Conversations that are too long make answers worse (context rot).
Tokens · context rot · clear the session.
Commands with “/” and a mode that plans before acting.
Save time and give you control over what gets done.
/clear · /context · plan mode.
🧠 Memory System
Short-, medium-, and long-term memory—and CLAUDE.md as the project manual.
Short (session), medium (project), and long (knowledge base).
Each level solves a different "forgetting" problem.
Short · medium · long.
What fits in the current session—and degrades when it fills up.
Knowing when to clean up maintains quality.
Session · clear · start over.
Separate projects into dedicated folders (5–8 buckets).
Organized context = more accurate answers.
Folders · scope · 1 manual per folder.
A file that explains the project to the AI (≈200 words).
It’s the average memory that isn’t lost between sessions.
Project manual · short · 1 per folder.
A knowledge base that accumulates what matters.
You stop repeating the same context every time.
Knowledge base · subfolders · accumulation.
A routine for saving conclusions in the right folder.
Turns conversation into reusable knowledge.
Save · correct subfolder · reuse.
⚡ Power features
Slash commands, skills, MCP, and subagents: from solo to turbocharged team.
Quick commands that trigger common actions.
Save typing and standardize tasks.
/command · shortcut · repeatable.
How context usage affects cost and quality.
Managing context prevents worse answers and extra costs.
Tokens · /context · clear.
Instruction packages that give AI a ready-made skill.
Reuses expertise without rewriting everything.
Skill · install · use.
A standard for connecting AI to external tools.
Opens the agent to the world (apps, data, APIs).
Protocol · connector · integration.
Several agents working on a task in parallel.
They finish faster and cover more ground.
Parallel · delegate · orchestrate.
One agent produces; another reviews and points out flaws.
Quality improves a lot with adversarial review.
Create · review · iterate.