Skip to content

Introduction

This is a practical guide to reliable AI-assisted engineering.

It helps you:

  • start safely
  • choose a workflow and stack that fit how you work
  • run AI-assisted workflows that stay verifiable
  • keep learning without outsourcing your judgment

This is not a giant market guide. Vendor details, benchmark snapshots, and product claims change quickly. The center of this primer is the part that ages slowly: workflow, verification, context control, and judgment.

  • Developers new to AI coding who want a clear path from zero to useful
  • Practicing engineers who already use AI but want better workflows and fewer avoidable mistakes
  • Tech leads evaluating how these tools should fit into team practice

If you’re new to AI coding: Start with Agentic Engineering, then Choose a Workflow and Stack, Setup Checklist, and Your First Session.

If you’re already using AI tools: Jump to Workflow Archetypes, Scenario - Fix a Bug, or Context Engineering. That is where the real leverage is.

If you’re evaluating for a team: Start with Governance and Rollout, Security Risks, and Workflow and Stack Criteria. Use the reference material only after the workflow and policy constraints are clear.

  1. Start Safely so your environment, feedback loops, and security baseline are not working against you.
  2. Choose a Workflow and Stack if you still need a setup that fits how you work.
  3. Work Reliably because this is the real center of gravity: bugs, features, refactors, verification, and recovery.
  4. Control Context once you are ready to make the tools more consistent and less noisy.
  5. Learn Without Dependency so speed gains do not quietly become skill loss.
  6. Use Reference only when you need volatile detail such as vendor pages, privacy comparisons, or benchmark snapshots.
  • verification matters more than confidence
  • workflow fit matters more than feature volume
  • smaller accurate context beats giant prompt dumps
  • a smaller accurate guide is better than a broader stale one
  • if you cannot review or explain the code, you should not ship it

Use the reference layer for:

  • vendor-specific tool pages
  • privacy and retention comparisons
  • benchmark snapshots
  • template downloads
  • research backstops

Those pages are useful. They are just not the center of the curriculum.

Found something outdated? See an error? This guide is open source.

Edit this page on GitHub or open an issue.