SEARCH
PART 9 · AI DEVELOPMENT METHODS
PART 9
AI DEVELOPMENT METHODS
6 classes. Each class holds its own standalone lessons.
PART 9 CLASSES
CLASSES, THEN LESSONS.
CLASS 35What Is Vibe Coding?
Vibe Coding Is a Fast Conversation With a Build System
Use natural language and AI to explore a working prototype without confusing a demo for a finished system.
READ LESSON →September 12, 2026 · CLASS 35When a Fast Prototype Becomes a Real System
Recognize the responsibility boundaries where a quick prototype needs stronger review, security, and control.
READ LESSON →CLASS 36What Is AI-Assisted Coding?
The 5-Minute Pre-Flight Before the First Prompt
Check secrets, scope, branch, model, acceptance criteria, and preview path before a coding agent starts.
READ LESSON →September 12, 2026 · CLASS 36The AI Can Write Code; You Direct the Work
Use coding AI for investigation and implementation while you retain ownership of scope, access, and proof.
READ LESSON →September 12, 2026 · CLASS 36Give a Coding Agent a Job It Can Actually Finish
Turn a vague wish into a bounded task with constraints and acceptance checks.
READ LESSON →September 12, 2026 · CLASS 36The Model Hierarchy: Use the Best Judgment Where It Matters Most
Route product thinking, architecture, coding, copy, and independent review to the right model role.
READ LESSON →September 12, 2026 · CLASS 36Orchestrate Parallel Work Without Letting Agents Collide
Use separate branches, worktrees, windows, contracts, and integration review to run AI work in parallel.
READ LESSON →CLASS 37Why Professional Development Can Look Slower
Speed Is Not the Same as Delivery
Measure AI work by promised behavior and evidence, not how quickly code appeared.
READ LESSON →September 12, 2026 · CLASS 37The Smallest Useful Quality Loop
Use a five-step solo-builder loop to inspect, test, checkpoint, and improve each meaningful change.
READ LESSON →CLASS 38Context Drift: When AI Goes Off on Tangents
Drift Starts Small: The Task Quietly Changes Shape
Manage context windows, catch bad inference early, save a clean handoff, and roll back a wrong turn.
READ LESSON →September 12, 2026 · CLASS 38Keep the AI on the Rails Without Making It Useless
Give the AI useful boundaries—scope, file limits, constraints, and proof—without making it ineffective.
READ LESSON →CLASS 39Why AI Projects Become Spaghetti Code
Every Patch Leaves a Shape
Spot technical debt, duplication, tangled responsibilities, and the early signs of an AI-built project losing shape.
READ LESSON →September 12, 2026 · CLASS 39Repair the Shape One Safe Seam at a Time
Refactor one clear responsibility at a time while protecting the behavior people already rely on.
READ LESSON →CLASS 40Hallucination Is Not Quite the Right Word
Plausible Is Not Proven
Replace confidence with evidence from real code, current documentation, tests, logs, and observed behavior.
READ LESSON →September 12, 2026 · CLASS 40Build an Evidence Loop Before You Trust Automation
Build a repeatable loop that turns intended behavior into proof, decisions, and safe next steps.
READ LESSON →PRACTICAL APPENDIX