Human Slop

People call everything AI slop now. Anything AI created. Anything with a hint of being made by AI.

I buy the premise. Some of what AI makes really sucks. It is very obviously AI, and it is bad. But I think society is massively underrating how much of the world is made of human slop. If you've ever worked in corporate America, you already know this.

The job you already had

Think of a job where you looked around and thought, "Wow, I can't believe this company is still functional. I can't believe they hired these people."

That says nothing about the people. The people are wonderful. They're great humans. Hang out with them and it's a grand old time. But in that specific job, in that specific setting, you want to tear your hair out. My guess is that if everyone watching the video were in one room, at least 80% would say, "Oh yeah, that happened."

That is human slop. Put people in roles where they don't execute at the highest level possible, often below average, and human slop gets generated. Every single time. Maybe the job sucks. Maybe the person isn't that into it, or they're complacent, or something else in their life is distracting them. Maybe they got a huge bonus or raise and lost the drive.

We accept it on the road

We have all accepted it, the same way we accept car accidents. You drive around and never think about the fact that nearly 40,000 people die from car accidents in the US every single year.

That's human slop too. People don't know how to drive. More than 2 million people get injured, and nearly 40,000 die a year, because humans can't drive properly. People literally die.

So I'm always shocked when I see comments saying AI is bad because it generates too much slop. From all the experience I've had with this technology so far, the opposite is true.

The opposite is true

Look at the latest round of models from the last 3 months: Opus 5.5, GPT6, Grok 4.6 and 4.7, Grok Bot, Meta's Muse, Hermes, Claude Code. So many of these systems have gotten so good at executing real world tasks and real world jobs.

If I were to bet, that's the big transformation of the next 12 months. People who look at AI in a negative way start looking at it in a positive way. They look around and say, "Wow, this system is able to do this job or this task or this thing that I wanted it to do 10,000 times better than I ever could have done it."

The reason we're here is that we've reached a threshold in model capability. All the GPUs these companies bought got us to an inflection point where most models now are very good at most things.

You and probably everyone around you use ChatGPT or Claude every day, maybe to diagnose something personal or medical, maybe to help with an email. I use it to generate things for my business, and if I told you how many tokens I've burned through, you'd think I have a problem. Token usage, the intelligence these models generate, has seen gigantic growth because these systems are becoming massively, massively valuable.

Better, faster, cheaper

And that's before the next generation of chips comes online later this year into next year: Nvidia's Vera Rubin NVL72 racks, AMD's clusters, whatever SpaceX and Tesla are working on with Terafab, TSMC's new chips. All of it makes this technology even better.

Then the model side. Anthropic, OpenAI, SpaceX, Meta, open source. They're all working to make these models way more intelligent and more capable of doing multiple things, while serving them faster so the intelligence reaches you quicker and at a lower cost.

Chips, models, serving. I think that trifecta keeps going for the next 12 months, 18 months, well into the future. What these models provide gets far better, far faster and far cheaper, and that opens up uses people never would have imagined.

Weeks versus 50 minutes

Here's a concrete example from my own channel. I've been experimenting heavily with animations driven by AI, using Opus 5.5 to generate the animations in my videos.

When I brought this to the team that puts these videos together, someone on my team said, "Wow, all those animations on that video would have taken me one to two weeks as an animator." With Opus 5.5, those renders took about 50 minutes.

The whole job was orchestrated with Grok Bot, which is an agentic harness. I could have used Muse or Hermes, but I find Grok Bot the best for my use case. It figured out how to fact check the script, how to generate the script and how to code the animations. All I told it was, "Hey, this is the idea. Go ahead and make the video."

And it made the video. Three months ago that would have been impossible, or really, really hard. The models crossed a threshold and got smart enough to do things that weren't possible before.

Not good, almost as good, way better

We've seen this pattern before. Take chess. For a long time computers were way worse than the best chess player. Then for a little while they were almost as good: Garry Kasparov beat IBM's Deep Blue in 1996. Then they were better: Deep Blue won the 1997 rematch. After that, no human being could hold a candle to them. In 2017, DeepMind's AlphaZero, built on a neural network, taught itself chess in hours and beat the strongest chess program. They're literally unbeatable. Going from almost as good as the best human to better took about a year.

Then Tesla's FSD. For the longest time, for the right reasons, everybody made fun of it. Everybody said, "Wow, Elon's promising FSD cars that drive themselves coast to coast forever. It's not going to happen."

In the last couple years it went from really bad, to almost as good as a human, to the point where the company is selling cars because people want that self-driving technology.

Legally, the company requires people to pay attention. But it's gotten so safe that Tesla is now building cars without a steering wheel and pedals. In Austin, these cars give rides just like an Uber, and you have literally no ability to take over. It's the same exact technology that was in the car the whole time.

Same trend both times: not very good, almost as good, way better than the human, very quickly. This is going to happen to most things and most jobs and most roles in the coming years. I'd call that a fact.

The human stays in the loop

The work these systems generate is going to be far better, far faster and far cheaper than any human could do. But that doesn't mean the human is removed from that loop. The technology is going to happen regardless.

If you're scared you won't have a job after AI, remember this has happened many times before. People were afraid of the internet's impact on society. It ended up being one of the greatest things for connecting people together. Social media can be very harmful, but the internet created unbelievable communities and brought the standard of living significantly higher.

Nuclear bomb, very bad. Nuclear energy, very good.

When the printing press first came out, people were afraid it would destroy the world because now you could put propaganda everywhere. They were right. But literacy rates skyrocketed. So did the movement of information from one place to another. Over and over, we've come across technologies that are unbelievably powerful with a lot of risks, and as a society we figure out how to handle the risks.

What model are you using?

Maybe you've tried Claude and it sucked. First question: what model are you using? Do you even know? Fast, old generation or latest? Free plan or paid?

Most people don't realize that even the next tier up from free, just the $20 basic tier on most of these AIs, is a big capability gain. You get bigger, better models that can think for longer, or agentic harnesses like Muse or Grok Bot that do actions on your behalf. Once you unlock that, the light bulb will go off.

Here's my quick guide, as of the time of my recording. It will probably change in about 2 months, so take it with a grain of salt if you're reading this during Christmas of 2026. Opus 5.5 is the best model at most things. For images, Claude is terrible, Grok Imagine is very good, and ChatGPT's latest image model is probably the best. For research, fact checking and the news, Grok is the best because it's plugged into X, what used to be Twitter, which is a direct feed for news. For really fast responses, use Gemini, Google's models.

Back to where we started. Right now anything with a hint of AI gets called slop. In the next couple years at most, I think these technologies get so good that people move from saying "oh, AI slop" to "oh, human slop." I'm telling you, it's going to happen.

So here's my ask. Take at least one to two hours every single week and try as many models as humanly possible. Get familiar with what they do, because they will change your life for the better. Then the next time you look around at work and can't believe the company is still functional, you'll know what to call it.