The Double Standard of "Slop": Why is AI Writing a Nuisance, but AI Code a Breakthrough?
Why is it that when an LLM writes a LinkedIn post, we call it "slop," but when it writes a Python script, we call it "productivity"?
There is a glaring cognitive dissonance in how we view AI-generated content. If you spend any time on LinkedIn, you’ll see disgust directed at AI text and (seemingly) low-effort, synthetically generated posts that clog up our feeds. Yet, walk into an engineering standup or look at the profiles of major AI coding influencers, and the conversation is entirely different. We are racing to integrate these tools into our IDEs to automate boilerplate, explore new libraries and APIs, and (increasingly) hand over more and more of the entire software development lifecycle.
As far as I can tell, the high-level argument usually goes like this: Code only needs to execute, whereas writing requires a "human soul" or a unique "voice."
But if we look closer, that distinction starts to collapse…
The Fallacy of the "Functional" Defense
The most common justification in defense of AI-generated code is simple: It works. Proponents will argue that the fundamental purpose of code is execution. If the compiler accepts it and the tests pass, the tool has done its job.
Of course, this is a dangerously narrow view. In professional software engineering, "it works" is merely the baseline. The real complexity lies in the non-functional requirements like maintainability, readability, security, and long-term stability.
AI can easily produce what we might call "convincing garbage" — code that looks correct at a glance and executes but lacks the rigor or context required for a production system. It might pass a casual review, but it introduces architectural rot and technical debt. If we define "slop" as low-effort content that creates significantly more work for the reader later, then unmaintainable, bloated AI code is exactly that.
If we are willing to call an unreadable, automated pull request "slop," why are we so much more forgiving of it than a LinkedIn post?
Fitness-for-Purpose
The real debate shouldn't be about whether the author was human or an LLM. It should be about fitness-for-purpose. Neither AI-generated code nor AI-generated writing is inherently "bad". Rather, its value is determined entirely by whether the output satisfies its specific objective.
The purpose of a post is communication.
The goal of a LinkedIn post or a technical blog is to convey an idea, spark discussion, or share an observation. While AI writing may lack stylistic "flair," it is often structurally superior to much of what humans produce. It is clear, organized, and intelligible. If the goal is the successful transfer of an idea from one mind to another, then a well-structured AI post is perfectly fit for its purpose.
In fact, AI writing can arguably be more fit for purpose. I could decide to start posting in Chinese and probably get my meaning across. But if I did, my mediocre grammar and limited vocabulary would distract from my overall point. I could completely sidestep this with an AI rewrite. My post would still be my ideas, but they would be packaged in a way that allows them to come across.
The purpose of code is implementation and maintenance, not just execution.
The goal of code is to implement a requirement within a living system. Because this code must be read, debugged, and evolved by humans over years, the "fitness" threshold should be much higher than that of a social media post. When AI produces code that meets the functional requirement but fails the maintainability requirement, it has failed its purpose, regardless of how amazingly quicky it started to “work”.
As with AI writing, AI coding can also be more fit for purpose. I could write a complex algorithm that is technically correct but ignores our internal libraries or fails to follow our naming conventions. This work product would be “slop”; code that works in isolation but adds friction and cognitive load to the rest of the system. Alternatively, I can use AI to take my core logic and wrap it in the necessary boilerplate, error handling, and architectural patterns our team expects. The fundamental algorithm is still my design, but the implementation is packaged in a way that ensures long-term maintainability.
You can ensure fitness for purpose of AI output.
Interestingly, proponents of AI coding are building entire ecosystems — IDE integrations, automated testing frameworks, and governance layers — specifically designed to turn raw, unverified AI output into production-grade software.
They are essentially adding “packaging'“ to ensure the code is fit for purpose. The double standard lies in assuming we can apply this level of structural rigor to code, while assuming writers are simply letting the AI run wild without editorial oversight.
The Responsibility of the Content Generator
With writing as well as with code, the danger isn't the AI technology and tools; it’s the temptation to use these tools to bypass the hard parts of the activities.
AI tools are force multipliers. They amplify existing strengths and weaknesses. What that means for software engineering is that if you have strong testing, governance, and review processes, AI can accelerate your delivery without sacrificing quality. But if your processes are weak, AI will simply allow you to produce "sludge" at a much higher velocity.
We should stop obsessing over the "soul" of the prose or the "humanity" of the textual artifact. Instead, we should focus our scrutiny on the output:
For writing: Does this text successfully communicate the intended insight?
For engineering: Does this code satisfy its requirements while preserving maintainability and risk management?
The "slop" label shouldn't be an attack on AI. It should be a standard for quality. The goal is to ensure that whatever we produce — whether a social media post or a pull request — is fit for the purpose it was intended to serve.
