The increasing AI-content consumption fatigue
Is it just me, or is this getting out of hand?
Seeing the cursor blink on the blank page before starting the write-up here, wondering how best to begin this text, makes me realize we don’t do this much anymore. We can instead just throw in any kind of bullshit prompt about what we would like to communicate into our preferred AI model, and it’ll spit out the perfectly curated text to achieve our goal. Well, perfectly plain, that is. Perfectly boring. Perfectly vanilla. Perfectly stripped of any new ideas. Perfectly AI.
It can’t be just me who is tired of seeing a piece of content that seems interesting, just to start reading or watching and realize that I’m consuming the output of a large language model that has guessed what would be the most compelling text for a generic person to read on this specific topic. The giveaways are always there. Plain as day. The person publishing didn’t even bother removing the em dash (—) Who the hell use em dashes, except for LLMs who just love them. And that also goes hand in hand with sentence building. Does this pattern sound familiar:
“You’re not only being fed AI crap — you’re indulging in it.”
“With this article, you don’t just complain about AI content — you skillfully navigate the issue as a thought leader.”
(Well, thank you, my cloud friend. Skilful navigation achieved.)
Of course, this is also a matter of prompt engineering. We can do in-context learning to avoid this robotic language. We can set up custom models that use other stylistic flairs. So maybe what I’m really complaining about here is the lack of prompt engineering behind all the content we consume. But even with better prompts to avoid the generic AI style, the fundamental issue remains: the lack of anything new in terms of ideas.
And don’t get me started on the images. Here, there are many good use cases. But at least prompt it well enough to make it make sense. The cover of this blog post is AI-generated. None of the text is, though. I’ve written every single character in it myself. Every typo and every strange grammatical exercise which only lands 80% of the time. Someone might say I’d have been better off having AI do it for me. But for me, that would have given me absolutely nothing. To the reader? I don’t know, maybe you all prefer the AI version over mine. If so, you probably haven’t made it this far anyways
Just to be clear, I’m not opposed to the use of AI. Quite the opposite, I use it all day, every day. For coding, for research, for summaries. As a second brain. Just like everyone else is doing these days. I used it to make complete renders of the house we’re building (workflow suggestions for that coming soon here), I used it to generate the profile pictures of all the people in my quotes section of this site, I use it in one way or another for most parts of life. But the content generation aspect of it is getting on my nerves. Corporate emails where the sender has pasted straight from the AI meeting summary. articles that are just a regurgitated mess from other articles. YouTube videos where the creator is just reading from a generated script (the football analyst channel stating that “United under Carrick don’t just play better football—they shine like the unit they always had the potential of being”).
It should also be said that I don’t think that AI-generated content itself is inherently wrong either. If it’s done well. If it’s properly curated. If it’s used tastefully. If it’s sprinkled with at least a touch of human thought. Structure. Something.
That’s it, over and out. Here’s a cheat sheet for you:
| Scenario | Use AI? | The Reality |
|---|---|---|
| Refactoring a messy switch statement into an object mapping. | YES | It’s a logic machine. It won’t get bored of renaming variables or fixing your nesting. |
| Writing a “Thought Leadership” comment on a viral LinkedIn post. | NO | Unless you want to sound like a corporate bot. Avoid gems like “You don’t lose control all at once, you give it away clause by clause.” |
| Explaining a complex architectural choice to a non-technical stakeholder. | YES | It’s an elite translator. Use it to turn “Hydration Mismatch” into “The browser and server are having a disagreement.” |
| Drafting a delicate “No” to a feature request that’s out of scope. | YES | It’s better at staying professional and firm than a developer who hasn’t had their coffee. |
| Writing the unique “Origin Story” for your new startup or project. | NO | AI doesn’t know the “why” behind your late nights. It will give you a generic template but you need to provide the soul. |
| Generating unit tests for a critical edge case in your i18n logic. | PARTIAL | It’s great for the boilerplate but you still need to be the one to decide if the test actually proves the code works. |
| Writing the copy in this very table. | YES | Plot twist. Gotcha. |