OPEN_SOURCE ↗
X · X// 3h agoTUTORIAL
ChatGPT Images 2.0 prompt goes viral
A Threads post is spreading a deliberately awful redraw prompt that pushes OpenAI’s image model to recreate attached images as clumsy, MS Paint-style scribbles. The joke works because stronger instruction following makes even “bad art” feel surprisingly controlled.
// ANALYSIS
The meme is really a stress test for control: if a model can preserve structure while intentionally degrading style, it’s showing real edit fidelity, not just pretty outputs.
- –Viral prompts like this are becoming the new demo format for image models because they are easy to share and instantly show capability
- –The prompt spotlights one of the biggest differentiators in this category: whether the model can obey weird, specific constraints without collapsing the composition
- –For designers, that means faster iteration and more playful mockups; for product teams, it means social virality may outpace practical reliability
- –OpenAI’s own launch materials emphasize better precision, text, and layout control, which is exactly why a “draw it badly” prompt lands so hard
- –If the model keeps improving, expect more adversarial and joke prompts to surface as the best public proof of quality
// TAGS
chatgpt-images-2-0image-genprompt-engineeringmultimodal
DISCOVERED
3h ago
2026-04-30
PUBLISHED
3h ago
2026-04-30
RELEVANCE
8/ 10
AUTHOR
arrakis_ai