Summary
Keywords
Full Transcript
In Part 7 of our Zero to Claude Certified Architect series, we break down Task 1.6: how to design task decomposition strategies for complex AI workflows. You'll learn exactly when to use prompt chaining (fixed sequential pipelines) versus dynamic decomposition (adaptive plans that evolve as the agent discovers new information). We walk through the canonical code review pipeline example, show the four-phase adaptive testing workflow, and explain the "lost in the middle" attention dilution effect — a concept the exam tests repeatedly. By the end of this video, you'll know how to choose the right decomposition approach for any scenario, how to split large reviews to avoid attention dilution, and how adaptive investigation plans work in practice. ✅ What you'll learn: - Prompt chaining: what it is, when to use it, and how it creates focused context windows - Dynamic decomposition: how plans emerge from discoveries and why the step count is unpredictable - Decision framework: 6-factor table comparing both approaches side by side - Attention dilution: the "lost in the middle" effect and how per-file passes solve it - BAD vs GOOD: visual comparison of one-prompt reviews vs per-file pass + integration pass - Adaptive investigation plans: how a plan revises itself after discovering a shared utility module - 5 anti-patterns tested on the exam (overly narrow subtasks, no integration pass, wrong approach selection, and more) - 6 exam takeaways to memorise before test day ⏱ Timestamps: 00:00 Quick Recap — Part 6 01:30 Two Fundamental Approaches 03:10 Prompt Chaining — How It Works 04:50 Code Review Pipeline (Code Pattern) 06:20 Dynamic Decomposition — How It Works 08:00 Adaptive Flow: Add Tests to Legacy Codebase 09:40 Decision Framework 11:10 Attention Dilution — Lost in the Middle 12:40 BAD vs GOOD Review Pattern 14:10 Adaptive Investigation Plans 15:40 5 Anti-Patterns the Exam Tests 18:00 6 Exam Takeaways 20:00 What's Next — Part 8 🔖 This video is Part 7 of the Zero to Claude Certified Architect series covering the full Foundations Exam curriculum. In Part 8, we tackle Task 1.7: Session State, Resumption, and Forking — how Claude stores named session state, how to resume a paused agent, and the critical exam decision between forking and resuming. #ClaudeCertified #AnthropicClaude #AIAgents #PromptEngineering #LLMEngineering #AIArchitect #ClaudeAI #AIWorkflows #MachineLearning #AIEducation #TaskDecomposition #PromptChaining
