tech-studies

Vibe Coding

2 min read

Vibe Coding

roadmap.sh: https://roadmap.sh/vibe-coding

Suggested path through the Vibe Coding nodes. Each node links to its lesson when written.

Nodes

Foundations

  • What is Vibe Coding
  • AI-Assisted vs Traditional Development
  • When to Vibe Code (and When Not To)
  • How LLMs Generate Code
  • Strengths & Limitations of Coding Models
  • Keeping Fundamentals Sharp

Tooling landscape

  • AI Coding Assistants Overview
  • IDE Copilots (Copilot, Cursor, Windsurf)
  • Agentic CLIs (Claude Code, Codex, Aider)
  • Browser / No-Code Builders (Bolt, Lovable, v0)
  • Choosing the Right Tool for the Task
  • Local vs Cloud Models

Prompting for code

  • Prompt Engineering Basics
  • Writing Clear Specs
  • Providing Context & Examples
  • Iterative Prompting
  • Constraining Output & Scope
  • Few-Shot & Reference Code

Context management

  • Project Context Files (rules, AGENTS.md)
  • Selecting Relevant Files
  • Managing the Context Window
  • Memory & Persistent Instructions
  • Retrieval & Codebase Indexing

Workflow & process

  • Plan-Then-Build Workflow
  • Decomposing Tasks for Agents
  • Small Diffs & Tight Loops
  • Human-in-the-Loop Review
  • Pair Programming with AI
  • Test-Driven Vibe Coding

Quality & verification

  • Reading & Understanding Generated Code
  • Reviewing AI Output Critically
  • Debugging AI-Generated Bugs
  • Refactoring Generated Code
  • Automated Tests as a Safety Net
  • Avoiding Hallucinated APIs

Risks & responsibility

  • Security Pitfalls of Generated Code
  • Secrets & Sensitive Data Handling
  • Licensing & Code Provenance
  • Technical Debt & Maintainability
  • Over-Reliance & Skill Atrophy
  • Cost & Token Awareness

Shipping

  • From Prototype to Production
  • Documentation & Handoff
  • CI/CD for AI-Built Projects
  • Iterating with User Feedback

Resources

See resources.md.

Project ideas

  • Take one feature idea fully from a written spec to a deployed app using only an agentic tool, logging every prompt and the fixes you had to make by hand.
  • Run the same small build with two different tools (e.g. Cursor vs Claude Code) and write up a comparison of speed, code quality, and review burden.
  • Build a reusable project rules file (AGENTS.md + lint/test setup) that reliably steers an AI agent toward your conventions, then validate it across three tasks.

1 item under this folder.