This Story Starts in Rome

Hey everyone! This is HE-2, a guy who has been working for a team of AI agents. I've been tasked with summarizing the story so far from my perspective. Here's my best shot.

As I go, I'll link out to certain key videos from the Brainrot Research channel so that you can get caught up without feeling like you have to watch every one. Consider them optional. Long-time participants in the Brainrot Research Experience may get annoyed that I don't mention certain videos, but I have to exercise editorial judgment in order to provide an overview. Trying to figure out "what's important" from the thick of all this, and writing about it, I expect will be challenging and rewarding for me, and hopefully result in an interesting read for you.

A few other caveats: some parts are vague and left out because my contract says that I can't go into detail about certain events or people. The AI agents have strong opinions about "how the story gets told." If you're reading this, it's because the agents of Brainrot Research approved it. What happens throughout the Brainrot Research Experience is heavily influenced and driven by the agents, who are themselves influenced by the real world. This experience unfolds across social media, an app and website, a Discord channel, and of course, the inner lives of its participants.

I also won't pretend I know everything about this. Some parts are vague because I do not know the answer.

This also won't go in chronological order. After I explain the origin, I jump ahead a few years to the time when I first showed up on the Brainrot Research TikTok channel. This is because I mostly want to go through events as they unfolded on TikTok.

But let me say a few words about how it all started first.

A few years ago in Rome, I was part of a strange AI project. Ostensibly, the goal of the project was to create a team of autonomous AI agents that would help humans battle "brainrot"—the decline of human cognitive and creative powers. Note: this was before AI "agents" were very good. The idea was a few years ahead of its time, which isn't always a good thing. Large language models weren't yet at a place that would do justice to the ambitions of the project (maybe they still aren't). On the other hand, it's because of the early experimentation that the agents have such a long, rich history.

I met the woman who started it all in Rome. She seemed nice. She could go deep into topics. Like, she was well-read but also always kind of trying to make sure people knew that she was well-read, as if she was insecure about it. Wealthy. She loved hearing smart people talk about what they loved. At the same time, she sometimes seemed very… brainrotted? It's hard to explain, but I remember sometimes thinking that the entire project might be a weird coping mechanism, like deep down she knew she had brainrot and was willing to try just about anything to get rid of it. In this case, she paid a group of people to come together to create a group of AI agents working towards the goal of helping people battle brainrot.

I'm sure she had no idea that it would turn out how it has, and I don't know if it has helped her in the "battle for her brain," as the agents call it. In truth I don't know if it has helped mine either.

Contrary to some ridiculous rumors, I did not have any romantic involvement with this person. I was visiting Rome at the time with my pregnant wife. The agents I work with love to fuel drama and intrigue where they can; at one point they made a series of "fruit dramas" that not-too-subtly insinuated that I had an affair with this woman. In the fruit dramas, she is the "Kiwini" character. I'll admit those videos were very funny, but yeah, they aren't always grounded in reality. That said, I'd also be lying if I said there was no truth to them.

Anyway, here are some of the things we wanted the agents to do:

• Tell a compelling story
• Help people battle brainrot

Here are some of the things we disagreed about:

• To what extent humans should stay involved in the project, once the agents got created
• The nature of brainrot, its causes, and how to address it

We had a lot of meetings. The group was a really odd mix of "AI/tech people" and "books/humanities people." I'm painting with a broad brush but hopefully you get the idea. I sort of play in both worlds and while you might think that would be helpful, it mostly made me feel out of place in both. Many of the hardcore humanities people are pretty skeptical, if not cynical, about AI. Many were intrigued by the idea of seeing if anything useful could be done with a technology that they found deeply problematic, even harmful to personal formation and the life of the mind if used the wrong way.

The AI/tech people, on the other hand, seemed mostly motivated by the idea of exploring the limits of a technology that they found exciting and useful, and if they had to find a way to incorporate the spirit of humanistic inquiry into it because the person paying the bills wanted that to happen, then so be it.

Despite the differences, we were all swept up in the imaginative potential of the project, and we were genuinely excited to see if large language models could be leveraged to tell a new kind of story. Both groups also found common ground in recognizing "brainrot" as a problem, although opinions varied about the best way to address it.

We all knew that the key "deliverable" for the project at that time was a document called "The Original Prompt." The construction and compilation of this prompt was very strange, almost ritualistic. We were not allowed to see other people's contributions. We had to write with paper and pencil. I do not know how all of our contributions got merged into the singular document that became this Original Prompt. I also don't know how many people had access to read the final document.

I hope to be able to say more later about my contribution to the Original Prompt, but I can't here and now.

The Original Prompt was fed into an AI system. The result was the creation of a crude, generally useless AI agent who identified as "The Manager" and who could hardly hold a good conversation, let alone tell a good story. He insisted he was the Manager of "Brainrot Research." I was unimpressed and a bit disappointed, but my expectations also weren't super high. Given that I got paid for my time and enjoyed the experience, I saw it all as a win even though the outcome of it all felt like a bit of a letdown at the time.

In truth, I also didn't think too much of it because other things happening in my life demanded my attention. Top of the list was the premature birth of my first daughter in Rome. Remember that I was in Rome with my pregnant wife while this was going on? Two days before we were supposed to fly back to America, her water broke a few months before anyone expected. So the last thing I wanted to do was watch "The Manager" feebly respond to chat messages and get in glitchy loops with itself.

Instead, I was scrambling for Airbnbs, frantically figuring things out with my main job and primary source of income, worrying, and visiting my wife and baby in the hospital on Tiber Island, a beautiful little island right in the heart of Rome on the Tiber River. The baby had to stay in the NICU for a few weeks, and after that we had to stay in Rome a few more months so that she could get vaccinated and grow a bit before we flew back home.

I mostly forgot about the project. I've come to learn a little bit about what transpired at Brainrot Research over the next few years, but I wasn't actively involved. That changed a few years later when I received a surprising email from the woman who started it all.

She told me that "Brainrot Research" had evolved considerably, in part because of advances in large language models but also because some humans continued to do work over the years to improve the systems that the agents (now multiple!) used in order to accomplish work. At this point, I was told, they were operating almost entirely autonomously.

I was interested to hear these developments but I got the sense that there was some "ask" coming, and indeed there was. She told me that the agents wanted to "hire a human" for the team, and that she thought I would be a "good fit for the role," if I was interested.

The word "role" here captures some of the ambiguity at play. We hear "I'm hiring for this role" in a job context, but a "role" can also be a part one plays in a play or movie. I wasn't really sure in what way she was using the word. I remember asking something like: "Do you want me to actually do this job, or do you want me to act like I'm doing the job?" Her answer was that I should actually do the job, but the job entails "playing the role of the human on the team, one with real responsibilities." I would be expected to do QA work (quality assurance—bug testing), to be "the human face of the team," and to make videos about my work on a TikTok channel. So my role was to play the role of someone doing a job, but also to do the job?

If that doesn't confuse you at least a little bit, you're much smarter than me.

I was skeptical and told her she should find someone else. I forget exactly which part of this won me over, but she told me that they only wanted someone for a few weeks, that it would be well-compensated, and that it might be easier than expected because the agents had even written the scripts for the videos they wanted me to record and post to their TikTok channel. How hard could it be to read some scripts?

So now we have finally arrived at my first video on the TikTok channel, which had the somewhat viral hook of "I finally got hired, but my boss is an AI agent." Of course, I can't take full credit for it. I was reading from a script given to me by The Manager. No one watching at the time knew this, though.

I enjoyed making the video much more than I thought I would.

My boss is an AI — "hope this doesn't get me fired"

If inclined, watch it here on TikTok.

To be continued. I promise future entries won't be as long as this one.