LLMs are the ultimate producers. Point an LLM like Claude or Codex at a code feature, and it will quickly produce more code than you can imagine (or review). Ask Claude or Codex to write a blog post, and it will produce more words than you can imagine (or edit). An unshackled LLM is, in reality, a slop machine. It lives to produce more, and more, and more.
What is AI Slop? It is the love-child of the hasty, unconstrained, and unmoored LLM. It is noise over signal. It is verbose output as proof of work or knowledge. It is a quantity-based solution to a previously unimportant and ignored non-event. It is the diarrheal whimsy of recursive bot flirting. It's an unnecessary salutation, an extra intro paragraph for every conversation, an extra 20 words for every paragraph. It's the unnecessary transition between the last thought and the next. It's an app filled with shadowy code that seems to be working. It's a container arbitrarily filled to its absurd capacity.
AI Slop is a sharp violation of the pre-existing social, intellectual, and cultural principles we should hold dear.
AI Slop is the opposite of curated, constrained, and focused output. Slop brain is “hood rich.” More is better. More words with transitions == smarter. More code in 12 hours == better. More multitasking, more threads, more agents, and more loops == more productive.
AI Slop is infecting everything. We're seeing the first forerunners of a tsunami of content posts, content length, SaaS apps, emails, job posts, and all variety of fishy-smelling digital transmissions. AI Slop is antiquing concise and authentic content posts on places like X, Reddit, LinkedIn, and Instagram.
Previously impossible posts from unimportant, average, and unproductive people are now being loaded into the LLM turret gun. X accounts that rarely posted have gone from two-line replies to 2,000 word formulaic exposés on the curious relationship between previously unrelated entities. Tech voyeurs who can't write code for shit are now exhaustive experts on the latest AI and architecture trends. "Serial entrepreneurs" are launching new whimsical SaaS apps faster than we can properly pronounce Klaviyo. TikTokers have CapCut with AI scripts and transitions. Your #1 party bro from college just penned a 3,000 word Homeric epic for his son's 6th birthday, without a single typo or grammatical error. C'mon.
AI Slop is causing a huge loss of trust in the communication, network, and productivity channels we once relied on. It's getting so bad that content platforms and SaaS markets may never fully recover. We're moving toward a world where filtering through the noise looks like an impossible task. We've seen more norms change in the last 6 months than we have in 15 years.
I posted on X that Claude has been one of the greatest marketing ops of our lifetime. What I meant is that the LLMs fill a space in our psyche that we are so eager to fill. The marketing messaging and pro bono evangelism are recklessly helping us fill that void. The way that people flippantly adore and evangelize the capabilities of the latest model is getting a bit absurd and embarrassing. Every single expert I know who isn’t selling something says the same thing: each new model is a tragically mixed bag.
As with most things in life, multiple things can be true at once. Almost nothing is pure. We must take the good with the bad. The AI world is filled with competing quandaries. In some cases I'm very critical of AI, in others I'm a big evangelist. AI is incredible, annoying, earth-shattering, disappointing, and revolutionary, all at the same time. However we feel about AI and LLMs, the genie is never going back in the bottle. Everything is different now. We must accept our new future, warts and all.
With the massive inertia behind LLMs, it's difficult to pinpoint exactly where we are, and where we're going. If we pick a single point of analysis, like say, bug fixing, code review, or porting an existing app to another codebase or framework, we can probably claim those tasks are 100x better with LLMs.
It may be simply be that focus dictates quality. Is it Slop, or is it good, meaningful output? If an article gets more "impressions" does that mean the LLM made your writing better? Is content totally fake and inauthentic, or is it helpfully expanded by the LLM, providing color and scope to a unique vector of thought and process excellence. If our LLM code seems to work and has an automated test suite, but we don't read the code, how can we really know its quality?
My problem today as I evaluate all of my AI-enabled channels (code, SaaS, content, social, advertising), is that I just don't know. I don't trust content authors or social posts. I don't trust the code that LLMs write, and I don't have time to review the aggregate output. I don't trust my media spend and analytics. I'm not sure I trust anything transmitted, built, or marketed anymore. I don't know up from down. I can no longer easily identify quality, taste, and tradesmanship against mimicry slop. The environment is so polluted with low-vibration noise that it feels like I'm living inside of a massive marketing operation. It's unnerving.
This is my emerging crisis of trust.
Is our future one where we just turn off our brain and everything is on full-self-driving mode? Is our future one where everything is juvenile whimsy and the hard toiling of an author or builder is removed? Is our future one where we just agree that the LLMs are genius actors because we're lazy and don't want to suffer the grind to mastery? Is our future one of mere orchestration management, like a frazzled project manager who doesn't know enough to audit details that can lead to death? Will we just offload everything to an LLM and "phone it in" for 20 years in our career and our already fake social media life?
We've opened the gates to hell, and we'll have to find a way to a new future where the AIs remain subordinate to a dedicated human touch of passion, trade, and opinionated excellence. Part of modernity's promise was a flowering of the spirited and captivated individual. If we're not careful, our intellect will die to serve the robots through a THC-filled vape cloud of non-think, pajamas in public, and inauthentic vibes.
The direction of things seems a bit suspect. Some of the world's greatest minds are sounding the alarm. We may lose our wayfinding abilities, and our way of life.