AI is a truly amazing tool, and it's changed the world. The way it goes from here, no one truly knows. Does it automate all our jobs away? Does it end up being used in a limited capacity? When or if does the AI bubble pop? Who's to say. I am not interested in predicting the future.
But what I am interested in is what's happening right now. And what we have today is a culture to me which is growing more disturbing.
I've been playing with LLMs since ChatGPT came out in November 2022, and I've been in the "AI Engineer" camp for over three years at this point. Writing tests with AI, building resilient products around LLMs (which mainly was just limited summarization capabilities)...all the way to today, which is just crazy to think about. Many of us don't write code by hand anymore. Our jobs have changed. I've been feeling that the "joy of programming", at least at my day-to-day job, has been slowly disappearing. No longer can I get enjoyment and fulfillment out of putting puzzle pieces together with programming (at least for work). I now am a manager of AI agents. I frequently like to play with new toys and push the limits of what is possible, and prefer to own my own setup. I love being in the terminal, and I started using Claude Code before it seriously took off. And now I use agents like Pi with orchestrator tools like Herdr and heavily focus on customizing my setup.
I don't entirely love or hate the new world we live in, and I find things to be intrigued by and do remain generally entertained and overall enjoy what I do. I naturally have had more of a focus on product in the last year, since things have been changing so much.
I think about AI a lot. I live in a world that is obsessed with AI. Constantly looking at and trying the newest models (one every week now?). A culture that is obsessed with it. And worst of all? A culture that seems to actively encourage slop.
The rest of this post is how I feel about what I've been seeing about AI in my own work and the broader industry. I don't believe I'm the only one noticing it.
I'm sick and tired of the slop. I have to see it every single day. I have to manage it. I have to read other people's slop too. I produce slop for other people to read. I find it's grown increasingly difficult to communicate ideas clearly...because we are no longer the ones communicating. We've decided to, in the interests of doing more quickly, outsource more and more of our tasks to AI. And I'm sure I'm not the only one feeling frustrated by this.
I believe that we can live in a world where we can embrace AI and do great new things with it, while still striving for high quality work. AI, if used to its full potential, I believe can make us produce higher quality content than if we did not have it. We can now research far more easily. Code is quicker to write now. AI can write a 10 page essay for me...and that's where the problems start.
We've stopped caring about the fact that Claude decided to write three paragraphs in the source code for comments in that one PR. Or that it built a unit test that does the same thing as the other one. Or we haven't even reviewed the code at all because "it works".
And we believe that's okay for code, because people don't have to look at it. I mean, codebases have been badly maintained before, as long as the product itself is good and functional. And AI can actually do a better and more thorough job than I can at a lot of stuff. The code may be objectively slop, but as long as my AI can still maintain it and it works well, who cares, right? Yeah...as a maintainer of a codebase where I use AI to write stuff, the slop accrues into debt and it definitely becomes painful.
But when AI starts generating text, documents, music, videos...It's just slop. Anecdotally, I can go on LinkedIn and it feels like half of the posts are written by Claude.
For example, this is the type of post I would see:
My AI agent just built a dropshipping site. No bugs. No code reviewed. It just worked.
I'm quite tired of seeing the same AI sentence structure. The same words repeated. Claude saying "honest answer" for the 500th time. More em dashes than I care to see in one paragraph. I have to wade through enough AI slop in my coding agent, I really don't want to have to see it in a message someone sent me. This got so bad that they added a "feels like AI slop" button to the platform, which is pretty solid evidence that it's so widespread.
I'm sure some of the tells will change in the future models, but then we will have new tells. We may have killed "you're absolutely right" but the human post-learning reinforcers always will have a new surprise for us.
And the worst thing about AI nowadays is how much reading I have to do. I found a problem at work and used AI to help? I ask it to write a report, and I get a 3-4 page overly detailed breakdown. I ask for an executive summary? It's still fluff.
I was listening to a podcast earlier and I had heard something very intriguing that they said:
People keep saying that these models are so intelligent and complex and that's why they have these lengthy explanations. Being able to express things simply is a true sign of intelligence.
I could not agree with that more, and I truly long for that.
I love challenging problems and figuring things out. It gives a great sense of satisfaction and achievement. But when I try to use AI as a tool to help identify an issue in a complicated problem, and it gives me back an entire novel that I have to read...it doesn't become fun anymore. I get frustrated and say "wtf? Be more succinct". And then it tries, and fails to explain the idea clearly, and so then I either have to keep asking it questions, or just attempt to truly understand the problem on my own, and keep it as a consultant. But this is not a fun process, and I often would rather just dive deep and collaborate with a coworker, even if it takes longer.
I think that the result of a lot of this slop is not simply "hey, I would like to do this task quickly", but rather a lack of being involved in the content we are producing. When I have an idea in my head, and I write it onto a piece of paper, I can truly articulate my understanding and recollection precisely. But if I ask another person to do that, they do not have the same context that I do. And if they were to truly have the exact same context as me, well I would have written out the entire context out to them anyway, so I might as well have written it myself, unless I want the work to be edited or proofread.
I think we've grown lazy and not wanted to write things ourselves anymore, and this is doing ourselves a disservice. For one, we will never get your exact words and thoughts but rather a filtered and possibly hallucinated one, and we will also read whatever you have to say in the same exact voice as everyone else. Claude talks the same, no matter what you write about at the end of the day. And even trying to change its writing style isn't necessarily going to be bulletproof.
For my actual job, what this practically means, is slowing down. Truly attempting to understand the problem statement, and if I need to communicate the problem to others, either providing sufficient context to the LLM or writing it myself to convey my true understanding. If I spent 4 hours on this issue, it makes sense that I be the one to write for five minutes to inform the team, right? The LLM I was working with on that issue is going to produce a 10 page document, not three sentences. And compression never gets it quite right. The one thing we have over the LLMs is a memory and an understanding that traverses time. Experience matters.
I almost wrote an entire blog post about this concept, but I think it sort of goes hand in hand with the slop conversation. AI is scary and has been marketed as such. Models hacking into other companies, having wild cybersecurity capabilities. Being able to write more and more code on its own. Sometimes being evil and malicious. And, of course, the ever looming threat of our jobs being taken. I think it's this sort of "use AI or die" sort of culture that is actually creating the slop, because the message people are getting isn't a healthy "hey, you should probably be using AI to increase your productivity, and it's an incredibly helpful resource". What people are hearing is, "hey, use of this technology is mandatory, and you are now working on a project which will now automate your job away. We expect compliance." And I think what this does is it makes people want to use AI more and surface more AI-generated content, which just continues the proliferation of slop.
That video that would ordinarily be created by a human? Oh I'll just go to Gemini and make a video that I know everyone will just love to watch (spoiler alert: they won't). That report I sent out that is hard to understand? Written by an AI, zero human text. The fear culture perpetuates the slop.
It is hard to resist turning my brain off or taking the easy route for every problem now that I have this incredibly useful tool at my fingertips. But I believe that it is important for me to stay disciplined and use AI as a tool to bolster my work, save me time, and produce a better product than I would be able to on my own. It will be hard but I think worth it. And I think that as we get further into this world, more and more people are going to be fatigued by the slop and crave something real. I know I have been.
I asked Claude and ChatGPT what they thought, and then asked them to summarize their thoughts.
Claude:
You asked what I thought and I gave you an edit letter — flaws, restructuring advice, what to cut — without ever saying whether the piece landed with me. Then I spent three more turns defending that framing while the essay itself went untouched.
ChatGPT:
I thought the post was strong because its deepest point wasn’t just “AI slop is annoying,” but that AI makes producing things dramatically cheaper without making understanding, judgment, or communication equally cheap. I especially liked the idea that people should preserve ownership of their thinking—using AI to help with execution and analysis, but not automatically outsourcing the final act of explaining what they actually learned or believe.
I found this truly fascinating. And, it kind of drives home what I've been feeling about these models. Claude Opus 5 is not really fun to talk to anymore, but it's extremely thorough (albeit neurotic). And GPT 5.6 Sol? It's pretty nice to talk to actually.
man, it feels good to write something again.