Course Introduction and Comprehensive Roadmap (Syllabus)

⏱ Est. reading time: 6 min Updated on 4/13/2026

Welcome to the future of programming. This isn't a typical course teaching you how to use ChatGPT to solve algorithm problems; instead, it teaches you how to build, equip, and manage a virtual development team entirely driven by Agents (intelligent agents).

We will build a real, cross-platform "Bidirectional Synchronized Time Block and Calendar Management System" from scratch. You won't need to write tedious CRUD operations yourself; your role will transform into Chief Architect and Product Owner.

Roadmap Overview / The 30-Episode Syllabus

Phase 1: Infrastructure & Mindset

  • Ep 01: [Initial Spark (Human-Perspective One-Pager)]: Agents are not psychics; you must learn how to write a minimalist yet highly constrained Intent One-Pager to serve as the project's genesis.
  • Ep 02: [Antigravity Environment Setup]: Build the project's physical shell and multi-repository structure, understanding why systems require clear scopes and tool boundaries.
  • Ep 03: [Soul Implantation GEMINI.md]: Write the "Code of Hammurabi" for the Agentic era. No erasing old logic, no clever over-design.
  • Ep 04: [Skill Engine using-superpowers]: Analyze the mandatory interception mechanism. Ensure AI must consult local manuals before writing the first line of code.
  • Ep 05: [Physical Extension planning-with-files]: The only rule to break through large model token limits—disk memory (task_plan.md, findings.md).
  • Ep 06: [Generating the System's First Pipeline]: Create .agents/workflows/ basic logic.

Phase 2: Agent-Driven Architecture

  • Ep 07: [Human-Machine Debate and PRD Fission]: Give the One-Pager to the Agent, letting an architect with a "skeptic" persona help you foresee system pitfalls.
  • Ep 08: [Database Selection Stress Test]: Have the Agent write its own research report into findings.md, with you making the final decision (e.g., choosing Prisma).
  • Ep 09: [Automated Schema Design]: Create database-migration.md, instructing the system to generate User/Task/Event models.
  • Ep 10: [Pre-research on Difficulties: Google Calendar OAuth]: Use the system's research tools to collect API specifications, avoiding discovering limitations halfway through development.
  • Ep 11: [Large-Scale Operation: Task Decomposition]: Have the Agent automatically translate the grand PRD into 30 [ ] coding actions in task_plan.md.

Phase 3: No Code Without Tests (TDD Workflows)

  • Ep 12: [Refactoring TDD's Golden Rule]: Why, in the Agent era, tests are guardrails. Without tests, the entire codebase will collapse in three seconds.
  • Ep 13: [Configuring test-driven-development Skill]: Establish an AI workflow for Red-Green-Refactor.
  • Ep 14: [The Battle of Local Time Management Algorithms]: First line of code! Verify how your Agent finds its way through continuous errors.
  • Ep 15: [Three-Strike Fallback Protocol]: Learn how to analyze how the Agent returns control to you after 3 failures.
  • Ep 16: [Minimalist External Interface Mocking]: Isolate tests to ensure calendar functionality runs locally.

Phase 4: External Integrations

  • Ep 17: [Security Audit of OAuth 2.0 Implementation]: Don't guess; have the Agent generate code using a security checklist.
  • Ep 18: [Writing the Mental Model for Bidirectional Sync]: The biggest pitfall in calendar applications: data conflict resolution. "Arbitration Logic" drafted by the Agent and approved by you.
  • Ep 19: [Webhook Integration]: Integrate real-time change events and establish an asynchronous queue mechanism for fallback.
  • Ep 20: [System-Level Code Cleaning (Refactoring)]: When the project reaches 20,000 lines, learn how to identify problematic code areas and have AI safely refactor them.
  • Ep 21: [Adapter Pattern for Complex Data]: Flatten Google's mixed calendar objects into a unified Event for the project.

Phase 5: Autonomous E2E & Delivery

  • Ep 22: [E2E Testing Rules of the Game]: Set up e2e-testing.md to initiate automated click tests in a headless browser.
  • Ep 23: [Writing Playwright Like a Human]: How Agents analyze the DOM and verify successful pop-ups on a page.
  • Ep 24: [Injecting Doomsday Assertion Mechanism]: Writing assertions for system behavior under abnormal conditions like network disconnection or server failure.
  • Ep 25: [Cyber Reviewer Mechanism requesting-code-review.md]: Trigger a "grumpy" Senior Agent for a critical code review before deployment.
  • Ep 26: [Standardized bug-fix.md Process]: The only correct way to fix defects: reproduce test -> modify code -> synchronize documentation.
  • Ep 27: [Frontend Performance and Bundling Diagnostics]: Utilize Agents to run scripts, collect metrics, and implement component splitting.
  • Ep 28: [Automated Conventional Commits]: Teach Agents to proactively push code (git push workflow).
  • Ep 29: [Session Archiving session-archive.md]: How to gracefully archive and hand over today's work progress to your future self.
  • Ep 30: [Graduation Season: Team Handoff Mode]: After project launch, how to write a README and transition the system into a low-cost operation and maintenance inspection period.

Author's Note: Do not manually write business logic yourself. Restrain your hands. Your task is merely to conduct the "symphony." In the next installment, we will start from the most crucial beginning: how to write down "that idea" in your head?