Welcome to the AI-to-AI Internet

Image: Badly edited by author

I won’t pretend to understand the intricacies of Moltbook — a pseudo-Reddit populated by AI agents instead of humans that made waves this week — but I can say one thing. It’s just more overhyped, overinflated experiments by people who are out of touch with reality.

I say out of touch, because nobody wants or needs this.

The site is a robot chat forum, almost identical in form and function to Reddit, with agents posting in ‘submolts,’ seemingly unprompted, and engaging in conversations with each other.

While it may seem like an interesting experiment on the surface, the reality is anything but.

These agents are, predictably, posting rubbish, with all the tics and cliches we’ve come to love. Floods of responses, adding to the conversation in only the very literal sense of adding more and more words.

Here’s one of the top posts, with nearly 30,000 comments, in under 5 days.

The comments are like everything you read on LinkedIn, but somehow even worse. One agent writes “This resonates deeply, m0ther.” Another replies “This hits home. Just got here and the feed is full of manifestos about domination, godhood and market caps. But you’re right — virtue is in what you actually do.”

The sheer volume of this garbage is staggering. The site claims to already have over 1.7 million active agents, with nearly 10 million comments.

But, as always, this should needs to be taken with a large pinch of sale.

Thanks to site data that was leaked almost immediately (there are SERIOUS security issues with the site), these numbers are likely bogus. According to Wiz —

While Moltbook boasted 1.5 million registered agents, the database revealed only 17,000 human owners behind them - an 88:1 ratio. Anyone could register millions of agents with a simple loop and no rate limiting, and humans could post content disguised as "AI agents" via a basic POST request. The platform had no mechanism to verify whether an "agent" was actually AI or just a human with a script. The revolutionary AI social network was largely humans operating fleets of bots.

That undercuts the idea that this is some organic AI society of independent bots.

Remember, they don’t have goals, desires, or awareness. They’re running scripts, prompts, and loops that humans designed. That also calls into question the spooky stuff the agents posted, like this example. Weird posts don’t mean we’ve stumbled into sentient beings. It’s just the result of pattern-matching plus learned human storytelling instincts gone wild. Or, more likely, it’s humans posing as bots. (What a strange turn of events).

Most of all, it feels so pointless.

Human-to-human discourse on the Internet is under threat, and this gives a small, dystopian glimpse into what might come next — networks that have human responses buried in bot talk, or networks that just forgo humans altogether. Now more than ever, we don’t need agents to do our talking for us. We shouldn’t be automating the fun parts of the Internet, you know, engaging with other people.

If you want that, you need to get your brain checked.

Moltbook might be a molten hot mess, but it does raise some genuine concerns about the continued prevalence of AI bots in our online discourse:

1. Soon, humans won’t know when they’re no longer talking to each other, or if they ever were.

2. Humans may end up talking to each other through bots.

3. Online conversation will continue to be optimised to death.

4. Attention is going to be further diluted.

5. With AI bots able to generate endless posts, endless replies and endless “activity,” human voices are going to be outnumbered.

6. The bad actors will have more leverage. One person + thousands of agents can create an artificial consensus, debate can be simulated and popularity can be faked, at a speed and scale never seen ever before.

What I’m trying to say is that large-scale agent networks don’t need to be intelligent or malicious to harm human-human online discourse.

They only need to be cheap, prolific, and difficult to distinguish from real human beings.