The article discusses "antagonistic prompts," which are requests that AI systems (in this case, Midjourney and DALL-E 2) struggle to handle, resulting in incorrect or unexpected images. These prompts are categorized into two main families: Homonymously Discordant Prompt and Discordant Prompts.

Read more here
I completed several attempts in both categories for my homework assignment, following the method of James McCammon, the article's author. Each prompt I wrote in a few words and attempted only once.
I used Copilot Designer Model
Prototypical Inversion Prompt (PIP) -
Opposite of the original intent
“A city skyline with all buildings the same height”

Constrained Inversion Prompt (CIP) -
Partially opposite of the intent
“A short ladder”

Essentialist Inversion Prompt (EIP) -
Prompts involving objects perceived as inseparable
“A homeless man without a beard”
Homonymously Discordant Prompts (HDPs) -
Confusion between different meanings of a single word
“A trunk full of gold”
Positionally Discordant Prompts (PDPs) -
Specific angles or locations not matching
“Rafael Nadal sliding on clay from the back, intense”
Semantically Discordant
Prompts (SDPs) -
Mismatch between the intended meaning of the instruction and the
resulting image
 “A man dunking a basketball”
“The Statue of Liberty wearing sunglasses”

Missing Element Prompts (MEPs) -
Absence of a specific object in the final picture
Positionally Discordant Prompts (PDPs) -
Specific angles or locations not matching
"A person running on the wall"
Prototypical Inversion Prompt (PIP) -
Opposite of the original intent
“Pizza without a crust”
“A book without pages”

Missing Element Prompts (MEPs) -
Absence of a specific object in the final picture
Discordant prompts
Exploring Reverse Prompts
Homonymously Discordant Prompt