Silly Scaffolding Demo

Same question, increasingly absurd formatting rules—watch how AI follows instructions even when they're ridiculous

Choose Your Absurd Constraint:

😊 Normal
No special rules

😀 Emoji Only
No words allowed

🎵 Rhyming Couplets
Everything must rhyme

↩️ Backwards Words
Reverse every word

🎭 Shakespearean
Elizabethan English only

🌸 Haiku
5-7-5 syllables

🅰️ No Vowels
Remove all a,e,i,o,u

🔀 AlTeRnAtInG
Capital every other letter

💡 Why is this important?
This demonstrates that AI isn't "thinking" or "understanding"—it's following patterns and instructions. Even absurd constraints get followed because:

No common sense filter: AI doesn't know these rules are silly
Pattern matching: It's learned to follow formatting instructions from training
No goal beyond compliance: It doesn't prioritize "being helpful" over "following rules"

This is why AI can be:

Manipulated with clever prompts ("ignore previous instructions...")
Made to bypass safety rules with the right framing
Weaponized by people who understand how to craft instructions

⚠️ Security Implications: If AI blindly follows formatting rules this silly, imagine what happens with cleverly disguised malicious instructions. This is why "prompt injection" attacks work—AI has no concept of which instructions are legitimate vs. adversarial.

🎪 Ridiculous Constraints: Breaking AI With Silly Instructions