LEARNING ATLAS / CLAUDE WORLD Path review · July 20, 2026

Orient first. Go deep with purpose.

Claude Code has grown from a chat interface into a system of tools, memory, agents, workflows, and long-running execution. This page routes you to the right next step instead of giving you another pile of links.

Find my route Open the architecture course
goal-based routes
5
architecture sessions
32
practical guides
24
ecosystem maps
2

COURSE SIGNAL

Your architecture course

No course progress yet. Start with the Agent Loop, then return here to continue.

Start Session 01

THE LEARNING CONTRACT

A route is useful only when it changes what you can do.

Every path follows the same loop: locate the gap, build a mental model, make something, and verify the result.

  1. 01 Locate

    Name the capability gap before choosing a tool.

  2. 02 Model

    Understand the system boundary and tradeoffs.

  3. 03 Practice

    Apply the idea to one concrete artifact.

  4. 04 Verify

    Use observable evidence, tests, and stop rules.

CHOOSE BY OUTCOME

What are you trying to do next?

Open one route. Each sequence starts with the strongest current entry point and distinguishes verified lessons from maps and editorial analysis.

  1. 01 START First reliable result I am new and want Claude Code to complete one small task well. 45–60 min

    Start with the agent loop, tool boundaries, and structured tasks—not a giant stack of plugins.

    You will be able to Ship one small, reviewed change with a clear verification step.
    Begin at the Agent Loop
    1. 01 LESSON PUBLISHED The Agent Loop S01 · 18 min read
    2. 02 LESSON VERIFIED Tool System & Permissions S02 · 15 min read
    3. 03 LESSON VERIFIED Planning with Task Tools S03 · 15 min read
  2. 02 ORIENT Understand the ecosystem I want to know why the tools evolved this way and where each product is heading. 35–50 min

    Move from the historical capability map to the current product landscape, then read the dated comparison.

    You will be able to Place a feature or product on the larger agent-computer map.
    Open the evolution map
    1. 01 MAP / PRACTICE Evolution map Open the map
    2. 02 MAP / PRACTICE Tool landscape Compare tools
    3. 03 EDITORIAL AI Coding Models & Agent Tools: The July 2026 State of Play 14 min read
  3. 03 BUILD Build repeatable workflows I want to turn recurring work into a dependable system. 60–90 min

    Begin with an outcome guide, then learn how hooks and workflows turn intent into repeatable execution.

    You will be able to Design a workflow with explicit inputs, checks, and a useful artifact.
    Choose a practical guide
    1. 01 MAP / PRACTICE Guides Ship a task
    2. 02 LESSON VERIFIED Hooks System S15 · 15 min read
    3. 03 LESSON VERIFIED The Workflow Tool: Scripted Multi-Agent Orchestration S27 · 16 min read
  4. 04 ORCHESTRATE Orchestrate multiple agents I need parallel work without duplicated effort, context leaks, or merge collisions. 70–100 min

    Learn the task graph first, treat Agent Teams as experimental, and isolate write scopes with worktrees.

    You will be able to Split work into a dependency-aware graph with ownership and stop conditions.
    Start with the task graph
    1. 01 LESSON VERIFIED Task Graph & Dependencies S07 · 18 min read
    2. 02 LESSON EXPERIMENTAL Agent Teams & Communication S09 · 12 min read
    3. 03 LESSON VERIFIED Worktree Isolation S12 · 12 min read
  5. 05 SHIP Ship with boundaries I want agents to operate on real projects without turning autonomy into risk. 75–110 min

    Study permissions, quality gates, evidence boundaries, and the current tool belt before widening autonomy.

    You will be able to Define permissions, verification, rollback, and evidence boundaries for production work.
    Start with permissions
    1. 01 LESSON VERIFIED Permission Models & Security S18 · 22 min read
    2. 02 LESSON VERIFIED Orchestration Quality Patterns S28 · 15 min read
    3. 03 LESSON VERIFIED The New Tool Belt: Artifacts, Tasks & Typed Outputs S32 · 16 min read

ONE ECOSYSTEM / FIVE QUESTIONS

Send each question to the right page.

Learn chooses the route. The maps explain change and coverage. Tutorials explain the machinery. Guides produce outcomes.

  1. 01 CURRENT Learn

    Find the right starting point and build a mental model.

    Choose a path
  2. 02 ↗ Evolution map

    See what the agent toolchain has completed and what a full computer still needs.

    Open the map
  3. 03 ↗ Tool landscape

    Compare Claude Code, Codex, and Grok by capability and direction.

    Compare tools
  4. 04 ↗ Architecture course

    Understand how Claude Code works beneath the interface.

    Study the system
  5. 05 ↗ Guides

    Apply the tools to concrete writing, analysis, coding, and business work.

    Ship a task

FIELD KIT

Keep the reference shelf small.

These are supporting tools, not another curriculum.

DOC Official Claude Code docs

Current product behavior, setup, workflows, and reference.

Open official docs LAB Claude World examples

Practical patterns, templates, and case studies in one repository.

Browse examples KIT Director Mode Lite

An open-source practice kit for agentic development workflows.

Open product docs LOG Curated field notes

Product changes and ecosystem shifts, delivered as a compact digest.

Read the newsletter

READY WHEN YOU ARE

Pick one route. Finish one artifact.

Depth comes from a completed loop, not from opening more tabs.

Open the architecture course Browse practical guides