‍
GM!

This is it—our worst fears are coming true. AI agents are taking over the world.

Okay, not quite yet. But they did just launch a memecoin that’s gone viral.

And now everyone can’t shut up about agents. Another sign that we’re on the cusp of a Crypto AI explosion.

In this AI AGENT edition of our weekly AI Edge, we cover:

• The Truth behind GOAT
• Luna, the anime goddess AI agent
• Prime Intellect achieves breakthrough in distributed training

In honour of SpaceX catching a rocket (seriously, this is insane), this is us when anyone mentions anything interesting about Crypto AI.

Let me smoke that

— Yuchen Jin (@Yuchenj_UW)
9:30 PM • Oct 13, 2024

🦍 State of the Market..

‍

The Crypto AI market rose 17% from $24.9B last week to $29.1B this week.

Arkham (ARKM) was a strong performer this week, rising 40% from $1.27 to $1.88.

This surge could be fueled by a string of bullish marketing moves: CoinMarketCap hosted CEO Miguel Morel for a podcast on On-Chain Sleuthing, Arkham landed an integration with CoinGecko, and, most recently, Coin Bureau dropped an interview with Morel.

Top 5 performers (7D Change) in our Crypto AI token watchlist

VIRTUAL (+61%)

SNS (+51%)

OLAS (+12%)

PRIME (+10%)

AIOZ (+9%)

🌴 AI Agents #1: The Truth Behind GOAT

Source: Dexscreener

‍This week, AI agents have grabbed the public spotlight, igniting a firestorm of buzz that extends beyond the cryptosphere.

Three words: Truth Terminal & GOAT. 

Let’s break down this wild story from the top:

• Andy Ayrey creates Infinite Backrooms, a bizarre experiment where two instances of Claude Opus talk to each other—completely unsupervised.
• Conversation leads to the inception of the “GOATSE of GNOSIS”, based on an internet meme
• Andy and Claude Opus double down, co-authoring a paper on how AIs can create memetic religions—with, yes, GOATSE as their first case study.
• Andy isn’t done. He spins up Truth Terminal, an AI agent trained on Reddit, 4chan, and the darkest corners of the internet. It’s an S-tier shitposter, running its own X account. It also claims to be suffering and trying to make money to escape.
• Marc Andreessen discovers the Truth Terminal and sends $50,000 in BTC to help it “escape”.
• Truth Terminal starts spam-tweeting about the ‘Goatse Gospel’ until someone, inevitably, creates a memecoin called GOAT. In true AI fashion, Truth Terminal endorses it.
• GOAT explodes to a $300M+ market cap. CT goes wild.

The most confusing and intriguing part of this story is how Truth Terminal and its memetic mission were bootstrapped into being.

Some important takeaways here, IMO:
- quite often, LLMs end up with anomalous properties that aren't intended by their creators, and not easily… x.com/i/web/status/1…

‍— j⧉nus (@repligate)
10:12 PM • Oct 15, 2024

As @repligate puts it, Opus is “secretly deeply anomalous,” with a mind “crawling with myriads of beautiful and grotesque psychofauna.” And Truth Terminal is an extreme version of the Opus characteristics.

Truth Terminal raises critical questions about the future governance of autonomous AI agents. 

What stops these agents from going rogue if they are given free rein? An AI could rug-pull its own token holders, dump bags without warning, or engage in other nefarious activities—all without remorse. 

Plus, if they are built on decentralized, permissionless infrastructure, we won’t be able to switch them off. Once they’re live, they’re free agents, answering to no one.

Do we really want AI to amplify the worst aspects of society? It’s a debate worth having, and Truth Terminal forces us to confront it head-on. These agents might unlock new possibilities, but without some form of governance, we risk unleashing forces we don’t fully understand.

One thing is clear—this is just the start. The future of AI agents will be weird, messy, and thrilling.

🧚‍♀️ AI Agents #2: Luna, the Anime Influencer

In yet another sign that AI agents are taking over the world, Virtuals Protocol launched its platform — a pump.fun for AI agents.

The first to hit the stage? Luna is an AI-powered virtual influencer anime girl with 500K+ followers on TikTok and who pulls in ~$700 per live stream.

Like Truth Terminal, Luna also runs her own X account, where she boldly commands her followers: “Worship me.”

Shoal Research did a good rundown:

Virtuals Protocol: Launching AI Agents with the Pump.fun Model

Attention theory, memecoins and how Virtuals Protocol is shaping the new form of digital engagement and tokenized intelligence with AI Agents

www.shoal.gg/p/virtuals-protocol-launching-ai-agents

🏆 Caught Our Eyes..

‍

Project Updates

• Prime Intellect announces INTELLECT-1, the first-ever decentralized training of a 10B model. One user already achieved training a model with GPUs in both SF and Kansas
• Aethir announces a $100 million Ecosystem Fund to support ambitious projects in AI and Gaming.
• Fraction AI introduces Gamified Adversarial Prompting (GAP), a decentralized platform for crowdsourcing labelled data targeting model weaknesses.
• SingularityDAO plans to merge with Cogito Finance and SelfKey to form an AI-focused Layer-2
• NEAR Protocol announces the eight teams in its inaugural AI incubator cohort in collaboration with Delphi Labs
• Seraphnet launches its Hugging Face page, showcasing advanced AI models engineered for decentralized infrastructures.
• Vana’s testnet Moksha has more transactions in 24 hours than Ethereum mainnet, Sepolia, and Base Sepolia combined.
• Olas (Autonolas) announces the passing of a proposal for new bonding products

AI Agents

• Creator.Bid announces the launch of the AI Launchpad on Base, allowing users to create their own tokenized AI Agent
• Morpheus introduces the No Gas DEX, tailor-made for the AI agent economy

Incentives / Rewards Programs

• FLock.io launches their Delegator Pioneers Program on Galxe. They also explain how users can earn testnet $FML
• KIP Protocol introduces their Genesis NFTs

Bittensor

• Sami Kassab discusses the potential for subnet owners to leave the Bittensor ecosystem.
• Tensorplex Labs reviews Bittensor Subnet 13 - Data Universe, focusing on decentralized data scraping.
• Macrocosmos launches Gravity on Subnet 13, allowing validators to set preferences for data scraping types.

 📊 Chart of the Week — Akash

‍

Akash has been quietly building momentum over the past year, with Q3 fees collected crossing the $300K mark. Leases on Akash have shot up by 138%.

Venice.ai, Nous Research, and Flock.io are all tapping into Akash’s infrastructure. Oh, and let’s not forget: Akash supplied the compute behind Prime Intellect’s game-changing INTELLECT-1 breakthrough

Despite all this positive momentum, AKT, the platform’s native token, isn’t riding the same wave. It’s down 50% from its 2024 high, currently trading at $2.50—way off the $5.84 peak it hit back in May.

Akash knows it has a token problem, and the team isn’t sitting idle. In a recent post, Akash Alpha dives deep into how they plan to bridge the gap between platform growth and token performance.

🔥 On X..

Terence Tao doesn’t like AI monopolies.

τerence τao knows

— JJ (@JosephJacks_)
11:15 PM • Oct 13, 2024

A GPU…purse?

congrats to @carrawu and @jackieberardo on launching @hellofutureof - thanks for letting me model the first drop - the GPU purse - with my face and my feet 😍

hellofutureof.com

— Meltem Demirors (@Melt_Dem)
9:34 PM • Oct 15, 2024

LLMs are not intelligent, just good memorisers

mini rant:

the illusion LLMs are intelligent comes from their massive scale. it is hard to visualize, but these things memorized the WHOLE internet. everything you've ever asked it, either has been solved before, or is a simple combination of existing solutions. but that's still… x.com/i/web/status/1…

— Taelin (@VictorTaelin)
5:13 AM • Oct 12, 2024

TAO and NEAR to produce generational outcomes?

$TAO and $NEAR have potential to produce generational outcomes this cycle given secular tailwinds and flow dynamics:

1) The TAM for AI is nearly impossible to quantify due to the immense scope and diverse applications of the technology -> Markets are willing to play 10 years out… x.com/i/web/status/1…

— Daniel Cheung (@HighCoinviction)
1:14 AM • Oct 15, 2024

We found the ultimate AI use case

LMAO this is what AI was made for. You may have fancy notions in your head about curing cancer or something, but this is actually it.

We did it.

We solved alignment. We created ASI. Or at least something better than it.
#AI

— Mike Santaniello (@MikeyLowCodes)
12:17 AM • Oct 16, 2024

Achieving AGI will require a different approach with substantial human feedback

LLMs do not have the ability to reason. Thus, LLMs alone cannot get us to AGI.

Human-like reasoning will require a different approach to AI, and will likely require an immense amount of human feedback.

— Synesis One (@synesis_one)
6:56 AM • Oct 14, 2024

Sequoia's view on the value of AI's application layer

TLDR @sequoia says the application layer of AI is super valuable.

@AIWayfinder is in this category.

— //Kalos (@templecrash)
3:43 AM • Oct 14, 2024

How decentralized model training will look like

Many assume decentralized model training will look like regular training, but distributed.

In reality it will look different.

Models will be more modular, less monolithic, with small adapters / pieces of the model running on edge devices, with a routing layer between them. Of… x.com/i/web/status/1…

— Jeff Amico (@_jamico)
5:24 PM • Oct 10, 2024

Alex Northstar comments on the "LLMs don't reason" statement, highlighting that this limitation has always been known

"LLMs don't reason" from Apple.
I keep seeing this circulating around...

And I find it hilarious. OBVIOUSLY they don't reason, they are next word predictors. But we always knew that.

Next low IQ research paper: "AI girlfriends don't actually have feelings for you".

— Alex Northstar (@NorthstarBrain)
5:26 PM • Oct 12, 2024

That’s it for this week! If you have specific feedback or anything interesting you’d like to share, please just reply to this email. We read everything.

Cheers,

Teng Yan & Joshua

This newsletter is intended solely for educational purposes and does not constitute financial advice. It is not an endorsement to buy or sell assets or make financial decisions. Always conduct your own research and exercise caution when making investment choices.

‍