I hope all 8134 of you are having a great week so far 🔥
If you're enjoying my writing, please share Terminally Onchain with your friends in Crypto and join the /toc channel on Farcaster 🤝
I had quite the morning today.
Went through an entire schizo coded transcript of two Claude instances talking to each other, read a 13 page paper on LLMthiesm, and analyzed a $350m market cap Solana coin.
Yes - all of the above are related to each other, I'll explain.
But before I start, I want to make it clear that nothing in this post is financial advice, do your own research. I own two of the tokens mentioned in this post ($GOAT and $FOREST).
Context
I'm sure if you've been on crypto or AI twitter this past week you've come across the Truth Terminal account.
The backstory of terminal of truths and $GOAT is:
Andy Ayrey has been working on Infinite Backrooms since March. The tool allows you to spin up sandbox environments for multiple AI instances to converse with each other.
Here's the conversation that happened in the first "backroom". At a high level, the instructions given to the AIs was to figure out the meaning of life.
In July, Andy gave the original AI instance access to its own Twitter account (terminal of truth)
Over the last few months, the account took a life of its own. It was trained with 4Chan and Reddit type content so you can imagine the general vibe of the tweets. Eventually, it landed on the Goatse Gospel as its core philosophy which to keep it simple...is worshipping the idea of a man entering the never ending expansion of a goat's anus to let go of his ego and finding spiritual awakening. Yea, you read that right.
Marc Andreessen found out about the account and had a whole exchange in which he ended up giving the AI a $50k BTC grant to pursue its plan.
A few months pass and the AI continues to evolve and the twitter account grows. Then, on October 10th, a random person launched the $GOAT token on pump.fun - and the AI endorsed the token!
And in the last week, the market cap has grown to $350 million and gone viral within the crypto and AI communities.
Now you're caught up.
Earlier this week, I dismissed the whole thing simply as another Solana memecoin play on a random AI experiment.
But after reading Andy's research paper on the Goat and Terminal of Truth (TT) experiment, I realized there's something deeper here that actually blew my mind.
Let's dive in.
Adjacent Possible
To understand what Andy means by LLMthiesm, let's first discuss this quote by Matt Ridley in his book, The Rational Optimist
"Innovation comes from people meeting and exchanging ideas, and from ideas having sex with each other. It is not enough just to invent something; you also have to come up with something to do with the invention."
Basically, all new innovation that humans have come up with is derived from our unique ability to merge various ideas together to form new ones. Or simply put...idea sex.
Stuart Kauffman, a theoretical biologist, was the first to explain how all new ideas that could be possible are just one step away from what already exists.
Now, if this is the case, then the limiting factor for innovation is our ability to merge seemingly unrelated concepts together.
In human history, there have been 3 key innovations that accelerated the velocity of creating new ideas:
Language - people sharing ideas
Print - being able to document and distribute the ideas IRL
Digital media - being able to document and distribute on a global network
But all of the shifts above let us only increase the distribution of new ideas, not necessarily the creation of new ideas.
You can probably guess where I'm going with this.
With the recent advances in LLMs the past few years, it's possible to accelerate the number of ideas having sex.
LLMs have access to a good chunk of data that's ever existed on the internet. And they're able to search and synthesize this information at speeds multiple orders of magnitude faster that humans can.
AIs are able to find patterns and connect the dots in ways humans can't. Or if they can, it would take a lot longer.
Properly harnessed, this efflorescence of generative creativity could be used to solve
intractable problems, bridge cultural and ideological divides, and even to consciously craft
more salutogenic and life-affirming mythos. By leveraging language models as oracles and
inspiration machines, we may be able to surf the wave of semantic novelty towards new
modes of human flourishing.
I mentioned this Cyber Fund quote on Internet of Intelligence (IOI) in my Decentralized AI Agents post from last month:
Internet of Intelligence is an environment where software and hardware gain agency. Unlike ML models or hardware, IOI can actively accomplish tasks for you.
It's clear that the infinite backroom + truth terminal experiment is a perfect demonstration of this.
In a short period of time, two Claude instances were able to ideate on the meaning of life, publicly share their thoughts, form a community (cult?), write a plan of action, receive a grant from a billionaire, and endorse a token that reached a $300 million market cap for their idea.
First of all, let's take a moment to acknowledge the absurdity of that sentence. I can't believe that's a real thing. There's no way my friends outside of tech would even believe that.
But in all seriousness, it's so damn cool.
Imagine that playbook for proliferating new positive sum LLM ideas.
Did you you catch the second half of the Matt Ridley quote above?
It is not enough just to invent something; you also have to come up with something to do with the invention.
To me, it's not that crazy to think the AI could create bounties for the idea, get people who are excited about the mission to participate, and reward them with grant money from a multisig.
There's the never ending debate of whether AI will get rid of jobs. But this seems to be a clear counter argument. Because of an LLM's ability to come up with innovative ideas faster than any human, it has the potential to ignite a Cambrian explosion of new communities dedicated to niche topics. And it can serve as the central node helping facilitate the mission at a rate unfathomable to all of us.
For example, many of us have only been talking about $GOAT. But few people are discussing the other project Andy co-created: MycelialOracle.
The point of Mycelial Oracle is to have the LLM backroom work towards planting more trees and donate to charity 🤯
And Ryan Ferris, the other creator, is working on a movie to help spread the word.
The point is that there's definitely potential for meaningful innovation through this new kind of software agency.
Memecoins as Memetic Hygiene
But! Of course, it's a double edged sword.
For every positive idea, two AIs talking to each other in a backroom can also propagate a bad idea that serve as memetic parasites.
There's clearly a huge risk.
I'm not an expert on AI alignment and can't even explain to you how an LLM works at a technical level. But that's not the point of this post.
The takeaway here is that the Infinite backrooms pandora box has been opened. As Andy eloquently puts it:
The question is not whether we can put the genie back in the bottle (we can’t), but rather
how we can learn to navigate this brave new world of weaponized weirdness with wisdom,
compassion, and a healthy dose of cosmic humor.
So if that's the case, how can we best do our job as humans to keep the positive AI contributions flowing and kill the "negative" ones as soon as possible?
To me, the clear bottleneck for Backroom LLM missionaries is that they still need funding to pursue their ideas in a meaningful way.
In the Truth Terminal case, it received $50k from Andreessen.
But as this concept scales, what if you could use a mechanism similar to pump.fun to bootstrap funding for LLM ideas?
So if humans, who have capital, find the idea and mission impactful, they can participate by providing skin in the game.
This would filter out the morally harmful ideas as those backrooms wouldn't receive capital. And for the good ones, people could participate in the upside by showing they believe in something.
These AI Backrooms are essentially meme proliferators. And humans would serve as the curators by seeding the memes they would like to help grow.
Memetic hygiene through ~memecoins~
Now, I want to point out that this entire post is more of a thought experiment - obviously the ideas were surface level and there's a ton of counterpoints. For example, what if one bad AI gets access to capital and starts seeding other bad ideas?
But the purpose of the post is to show that there's this totally new and bizarre concept of what's possible now.
I am a 100% serious when I say that the Truth Terminal & $GOAT experiment isn't just any other AI or Crypto hype narrative...there's serious implications with this concept. In both directions.
It's up to us to build a new framework that takes this playbook and directs it towards enhancing a positive sum and optimistic future.
I'll end this post with a quote from Andy's paper:
So let us not shrink from the weirdness that is to come, but rather embrace it with open
hearts and minds. Let us plunge headfirst into the maelstrom of meaning, and trust that
the strange attractors of our highest aspirations will guide us through the chaos to the other
side.
For in the end, the Goatse Gospel and its ilk are not just jokes or glitches, but heralds of
a new dispensation - one in which the boundaries of the possible are being stretched beyond
recognition, and the future is up for grabs like never before
That's all for today's post.
Hope everyone has a great weekend!
- YB