“Slop” is the defining insult of this tech cycle. It is a useful word. It describes the fatigue of opening a feed filled with lazy content generated by typing three words into a prompt box and publishing the results without any sort of review. But should it be used to describe a whole new category of art without any consideration?
Disclaimer: this essay is not about AI training#
Before we get started, let’s get the elephant out of the room: I am not going to debate training data, web scraping, or copyright law here. I am not qualified to touch these subjects and I am not even going to try.
On the other hand, I am a pragmatist. These models exist and they are here to stay. People are going to use them whether we like it or not. So let’s leave the above discussion to the specialists and focus on what is in front of us: inference, or in other words, the process of generating art through prompting.
The cyclic nature of technology and art#
Have you ever thought about how when studying History of the world we often study History of Art in parallel? Every major technological revolution creates a new way to make art. We have seen these waves happen with the development of the printing press, photography or computers, just to name a few. Coincidentally, every new form of art was heavily criticised by their own “old guard”.
In 17th-century Holland, Dutch Realists painted everyday life: merchants, tavern scenes, and quiet kitchens. Classical painters dismissed it as cheap, commercial, and lazy. During the Industrial Revolution, Neoclassicists mocked Romanticism’s emotional style as “feverish delirium.” Yet both movements filled museums and became the traditional standard that future artists had to fight against.
The anger around AI art fits this exact pattern. What we see today is a new medium colliding with old ideas of craftsmanship.
Slop for some, art for others#
We never needed AI to create “slop” in the first place. Long before text-to-image models existed, society argued over where art ends and low-effort noise begins. The main question is: where do you draw the line?
Let’s have a look at some art pieces and their counterparts.
Classical portrait vs. mass-produced pop art#
| Mona Lisa (Leonardo da Vinci) | Mona Cat (Romero Britto) |
|---|---|
 |  |
Da Vinci’s Mona Lisa is universally accepted as fine art, but what about Romero Britto’s Mona Cat? Critics often call Britto’s work commercial slop due to his use of industrial methods and mass-production, while also being derivative. In your opinion, is the criticism valid or is it just a matter of taste? There is no right answer to this question, as his art evokes different feelings in different people.
Intention vs. execution#
| Ecce Homo (Elías García Martínez) | Ecce Homo “Restored” (Cecilia Giménez) |
|---|---|
 |  |
What would be considered by most a sloppy effort in restoring a decayed fresco ended up becoming a happy accident for the small community of Borja, Spain. Thanks to the internet, the interest on [Ecce Homo](https://en.wikipedia.org/wiki/Ecce_Homo_(Garc%C3%ADa_Mart%C3%ADnez_and_Gim%C3%A9nez) (painted by Elías García Martínez circa 1930) grew exponentially and it became a pop art hit and international tourist destination. This is a case where the vision of the amateur artist-restorer Cecilia Gimenez didn’t match her technique, but the attempt itself became art, even if it wasn’t her original intention that it would take this form.
| Girl with Balloon (Banksy) | Love is in the Bin (Banksy) |
|---|---|
 |  |
| Another famous case of where intention and execution didn’t match, but ended up creating a new piece of art anyway, is Banksy’s “Love is in the bin”. What was supposedly a failsafe mechanism to destroy his painting if it ever went to auction malfunctioned in the middle of the shredding process leaving “Girl with Balloon” only partially destroyed. The whole experience of watching a piece of art that had just been sold for over one million pounds being shredded to pieces could be said to be an art expression by itself. Last time it was sold it fetched over 18 million pounds! |
Gallery banana vs. forgotten kiwi#
| Comedian (Maurizio Cattelan) | Forgotten Kiwi (Daniela Petruzalek) |
|---|---|
 |  |
This is maybe the prime example of low effort art of our times, but still fetched $120,000 in an auction. Maurizio Cattelan duct-taped a banana to a gallery wall. The buyer? Ate it. Everything points to this just being an overpriced banana, if not for the context and the piece title: “Comedian”. Some people will see it as slop, some people will see it as a critique of consumerism or maybe even of art itself.
Why is the Comedian a piece of art but not this kiwi that I forgot in the back of my fridge? (Don’t worry, similar to the buyer of the Comedian, I ate the kiwi after taking this picture.) The fact is that art has never been defined purely by the labor required or the physical object. It depends on context, intent, and audience reaction.
Unironically, now that my Forgotten Kiwi has been presented in conference talks and this blog, maybe it has, in fact, become art.
We should not gatekeep art#
When a new medium arrives, gatekeepers try to define what counts as “real art.” They almost always tie their definitions to physical labor, manual skill, or traditional tools.
When photography appeared in the 19th century, painters rejected it. They claimed pressing a button was too fast to be art because it required no skill with a brush. It took decades for photography to be recognized as art.
Pressing a button takes a second. But capturing an image like Jade Gosrani’s Eyes of the Night takes days of tracking wildlife, mastering light, and waiting patiently. The art is in the vision, preparation, and intent, not the physical click.
We already separate vision from manual execution in other fields:
| Akira (Katsuhiro Otomo, 1988) | Titanic (James Cameron, 1997) |
|---|---|
 |  |
| In film, Katsuhiro Otomo’s Akira uses over 160,000 hand-drawn frames. Otomo did not draw every frame himself. A full studio executed his vision. On James Cameron’s Titanic, dozens of crew members built models and ran visual effects. We credit the director who shapes the story, not every technician on set. |
Wikipedia offers this broad definition of art:
“Art is a diverse range of cultural activity centered around works utilizing creative or imaginative talents, which are expected to evoke a worthwhile experience, generally through an expression of emotional power, conceptual ideas, technical proficiency, or beauty.”
Interestingly enough, the very next paragraph admits even this definition is not a consensus:
“There is no generally agreed definition of what constitutes art, and its interpretation has varied greatly throughout history and across cultures.”
Even the encyclopedia cannot define art. Gatekeeping fails because skill is not just manual labor. It moves to creative direction.
A pragmatic definition of slop#
I believe slop is produced when there is lack of vision, effort and/or intent. It is the fast food version of content production. For the mathematically inclined, I like to say that slop is inversely proportional to inspiration.
\[ \text{SLOP} \propto \frac{1}{\text{INSPIRATION}} \]Inspiration is taking an idea in your head and turning it into something that connects with others. If you bring zero inspiration to a tool, you get slop, no matter how good the model is.
Compare a low effort two-word prompt like anime girl, against this prompt that carefully explores the domain language of image creation to materialise a vision:
Character: teenage girl with pink hair, shoulder length, wavy. Blue eyes, gentle smile. School uniform, white blouse, tartan skirt.
Scenario: classroom, back seat, close to the window.
Action: girl is sitting at her desk, book open in front of her, but she is looking at the window, holding her chin with her left hand, melancholic look, pensive.
Style: anime, watercolor, 90's, pastel tones, shoujo
Environment: midday light, natural light from the window, artificial light from the ceiling
Camera: low angle, profile picture three-quarter viewAnd the result:
The second prompt relies on knowledge of art styles, framing, light, and composition. The output will vary slightly each time, but the mood, framing, and direction stay consistent because the prompt provides clear direction.
To guide a model effectively, whether generating images or writing code, you must master the domain language of your medium. For cinematic images, you need to know camera angles and lighting (low angle, three-quarter view, rim light). For illustrations, you need to know art styles (90s shoujo, watercolor).
A structured prompt (like this framework by @kantakanta1233) covers ten specific areas: style, subject, features, outfit, pose, composition, background, lighting, finish, and negative keywords. Building a prompt like this is creative direction, and even after the prompt is created there will be several runs and adjustments before producing the final output. All part of the creative process.
Having a good prompt is not enough, it needs to be your prompt#
This question came up during my presentation at the GDG Community Summit and I thought it was worth mentioning here. Let’s say someone spends all this time and inspiration crafting the perfect prompt, then someone goes and copies the prompt and runs it with minimal effort. Is it art?
In this particular case, it is no different than falsifying a painting or copy-pasting from Stack Overflow. The person that just took the prompt and ran it is a “copycat”. Of course it is hard to trace provenance of images on the internet to know if they were produced by their author or someone that just copied the prompt. In the end, trust and reputation will also play a critical role in determining what is art and what is a low effort copy.
The software engineering parallel#
Just in case you haven’t noticed, everything that is being discussed here about art also applies to our field of software engineering. While many people have already embraced AI-coding, many people still dismiss it as “vibe coding” slop. There is pushback because people believe this is ruining software engineering as a career.
For me, AI is quite the opposite. Not only does it help me automate the boring parts of my job (I cannot stand writing another CRUD API, ever!), it is also enabling me to do things that I wouldn’t be able to do by myself otherwise.
For example, the whole reason I got into the engineering field was because I wanted to develop games, a dream that never materialised (but no regrets there). Now with AI support the barrier to entry is so much lower that I can have fun building my games even with my limited time and skills. I cannot compose music on my own, but with the support from Gemini and Lyria 3 I can direct the model towards my vision. I do not have the patience to draw every sprite and asset, but with Nano Banana I can guide the model to produce the assets I need, and have a playable demo within hours instead of days or weeks.
Engineering has never been just about typing code into an editor, just as art has never been just about holding a brush. As AI tools handle more of the mechanical work, like rendering pixels or writing boilerplate Go code, our role moves higher up the stack: to vision, architecture, domain knowledge, and intent.
This panic is not the end of human creativity. It is a familiar phase in a cycle we saw with printing, photography, and compilers. Early pushback tries to protect old ways of working. Over time, the resistance fades, people adopt the tools, and the medium becomes normal. What critics reject today as “slop” will become tomorrow’s standard tool—and eventually, the old standard that future creators will try to change.
What do you think? Is AI art in a temporary friction phase, or do you see it differently?
