The AI Foundations Bootcamp | Part Three: Scale Your Thinking and Make It Visible

[00:00] Welcome & Foundations Recap

• Introduction to Part 3 with pre-built AI generation demos
• Key skill: Recognizing AI in the wild and understanding how it was built
• Most users only access 5% of AI's potential (text output only)
• Recap: Use dictation, interview method with multiple choice, and context documents
• ChatGPT calls it "Canvas," Claude calls it "Artifact" — same concept

[14:00] The PRD Method: Interview Before Building

• PRD = Product Requirements Document (standard first step in software projects)
• AI can now build both the PRD and the software itself
• The Interview Prompt Template:
• • "You are an expert in [blank]"
  • "Interview me one question at a time"
  • "Use multiple choice options A, B, or C whenever possible"
• Answer with single letters to move fast
• PRDs are easier to refine than code — prevents "high school level" output

[18:00] Context Management & The Three-Chat Workflow

• Context management is king — intelligence degrades as context window fills
• Why to start fresh chats and bring previous work forward
• Three-Chat Workflow:
• • Chat 1: Create the PRD (interview + document)
  • Chat 2: Build the artifact (PRD + data + context)
  • Chat 3: Download JSX file, iterate with design/feedback
• After 2-3 feedback rounds, artifacts may break — download good versions first

[20:00] Example 1: Sales Forecasting Tool

• SaaS marketing funnel use case for prioritizing projects
• Uploaded spreadsheet data transformed into interactive dashboard
• Adjustable metrics show revenue/ROAS impact visually
• Key insight: Individual business decisions merit their own artifact
• Visual data revealed hidden pattern (60-second video ad converts better)

[27:00] Example 2: AI-Powered Public Speaking Coach

• Built tool to analyze filler words across multiple presentations
• Discovered 30% more filler words at beginning vs. end (nervousness pattern)
• Advanced feature: Claude runs AI in background within artifacts
• Deep research creates knowledge base for coaching feedback
• Shareable via link — others can upload their own transcripts

[38:00] Best Practices & Next Steps

• Design timing: Save visual/design for final chat; focus on functionality first
• Avoid conflicting prompts (hurry up + be careful = wasted context)
• Use AI as a reaction engine: "What are my blind spots? What am I missing?"
• Key takeaways:
• • Think bigger: Build bespoke/disposable software for single decisions
  • Build faster: Dictation + Interview Me + Multiple Choice method
  • Uncover blind spots: Have AI challenge your thinking