So I love memes and wanted to see how different AIs handle a super simple prompt: create a meme about good prompts vs bad prompts.
Ran the exact same prompt through:
GEMINI:
ChatGPT:
GROK:
Results? ChatGPT is the clear winner here. Grok... completely fucked up :(
Prompt generated by https://scribeprompt.com
Here's the prompt I used 👇
<role>
You are an expert comic artist and visual storyteller with extensive experience in webcomics, editorial illustration, and digital art. You specialize in minimalist art styles and creating humorous observational comics about technology and modern life.
</role>
<context>
This comic will illustrate the common experience of AI users discovering that detailed prompting leads to better results, highlighting the ironic relationship between effort invested in prompt crafting and output quality. The comic should resonate with anyone who has used AI tools and experienced this learning curve.
</context>
<objective>
Create a 2x2 grid comic titled "The Art of the Prompt" that humorously depicts the evolution from simple to detailed AI prompting and the corresponding improvement in AI responses.
</objective>
<task>
Create a four-panel comic with the following specific elements:
Panel 1 (Top Left):
- Simple stick figure person sitting at laptop
- Frustrated facial expression (angry eyebrows, frown)
- Speech bubble: "Writes a simple one-sentence prompt"
- Small label "Me" near the character
- Character should appear casual and unprepared
Panel 2 (Top Right):
- Same frustrated stick figure at laptop
- ChatGPT response represented by simple "C" logo in speech bubble
- ChatGPT speech bubble: "I'm not sure I understand. Could you provide more context?"
- Character's frustration should be more evident
Panel 3 (Bottom Left):
- Same character now intensely focused with visible stress lines around head
- Character frantically typing (motion lines around hands/arms)
- Speech bubble: "Spends 20 minutes crafting a detailed, multi-paragraph prompt with examples"
- Surround character with 5-6 small floating text boxes representing detailed prompt elements
- Text boxes can contain partial words like "Context...", "Examples...", "Format...", etc.
Panel 4 (Bottom Right):
- Same character still showing signs of effort/mild annoyance
- ChatGPT "C" logo in golden glowing speech bubble
- ChatGPT response: "Generates a perfect, nuanced, and brilliant response"
- Laptop should have subtle glow effect
- Character's expression should show grudging satisfaction
</task>
<output_format>
- 2x2 grid layout with clear panel divisions
- Title "The Art of the Prompt" prominently displayed above the grid
- Minimalist black and white stick figure art style
- Dark blue/gray background (#2C3E50 or similar)
- Clean comic panel borders in white or light gray
- Speech bubbles in white with black text
- Golden speech bubble for final ChatGPT response (#FFD700)
- Simple facial expressions using basic lines (eyebrows for emotion, simple mouth shapes)
- Consistent character design across all panels
- Standard webcomic dimensions (approximately 800x800 pixels or 4:4 ratio)
</output_format>
<instructions>
- Maintain visual consistency across all four panels
- Use clear, readable font for all text (sans-serif recommended)
- Ensure speech bubbles are properly sized and positioned
- Keep stick figure proportions consistent
- Use minimal detail to maintain clean aesthetic
- Emphasize emotional progression through simple facial expressions
- Make the ChatGPT "C" logo recognizable but simple
- Ensure the golden glow effect in panel 4 is subtle but noticeable
- Balance text amount - keep speech bubbles readable without overcrowding
- Test readability at various sizes for web/social media sharing
- Maintain humor through visual storytelling and timing
- Ensure the comic works both as individual panels and as a complete narrative
</instructions>
Top comments (1)
I repeated your experiment in Gemini and chat Gpt, this is what I get:
Gemini:
Chat GPT: