Learning ThroughDocumented Experiments
In-depth explorations of prompt engineering techniques, AI limitations, and real-world applications. Each case study includes interactive examples, bilingual videos, and downloadable resources.
What Are Case Studies?
Unlike traditional blog posts, these case studies are detailed experiments where I test prompt engineering techniques, document every step, analyze results critically, and share everything—including what didn't work.
Each case study may include the following and more:
- Bilingual video walkthroughs (EN & FR)
- Downloadable prompts and templates
- Before/after interactive comparisons
- Expandable deep-dive sections
- Live demos and deployed applications
- Detailed analysis and key findings
All Case Studies (5)
XML-Structured AI Agents: Optimizing for How Models Actually Work
I discovered XML structuring could optimize my 7 Context Engineers for token efficiency, minimize attention degradation, and enforce priority hierarchies. Here's the redesign.
Anthropic recommends XML tags for prompt clarity. Before implementing the 7 Context Engineers for 777-1, I restructured their definitions from prose to hierarchical XML. Here's why this matters for token efficiency, attention management, and context engineering.
Visual QA on Autopilot: Building a Self-Correcting AI Pipeline
I built a pipeline where Claude screenshots apps, finds visual bugs, and fixes them itself.
Too lazy to screenshot every UI bug? I built a workflow where Claude sees and fixes its own visual mistakes. Score improved from 63 to 77 in the first phase - then things got interesting.
Meet the Team: 7 Custom Subagents Built from 129 Code Reviews
Real names. Real personalities. Real job descriptions. Download the markdown files and use them yourself.
I analyzed 129 code reviews and extracted the 7 most common issues. Then I turned each one into a subagent with a name, a personality, and a detailed job description. Here they are.
From Trained Instance to Portable Prompt: Extracting Nancy's Blog Refiner
How a financial blogger captured 3 months of AI training and made it work across Claude, ChatGPT, Gemini, and Grok
I had the perfect blog refiner. Then I lost it. Here's how I extracted my trained AI's personality using prompt extraction and made it work everywhere.
Meta-Prompting in Action: What Happens When You Let Claude Code Redesign Your App
A brutally honest experiment in trusting AI without human oversight
I asked Claude to write detailed prompts for Claude Code, then executed them without any testing or modifications. Here's what worked brilliantly—and what broke completely.