It was clear from the start that AI would be a major narrative this cycle but the explosion of AI agents has caught even the biggest degens off guard.
In a matter of mere months, we have seen projects speed up to hundreds of millions and, in some cases, multi-billion dollar valuations, and it seems we are just getting started.
The fusion of the two biggest technological themes of our generation was always going to be a killer combo that would be sure to drag some of the biggest brains on earth into our degenerate crypto casino.
If you don’t want to get left behind, then you’d better read on…
What are AI agents?
The concept of an AI agent is simple to understand at a fundamental level, but it quickly becomes far too complicated for my pea-sized brain to explain in a detailed, high-end technical way.
Thank god this is just an idiot's guide!
When we think of AI in its current form, we probably think of large language models (LLMs) like ChatGPT. We fire all of our questions at them and in return, get half-decent responses that we can then use to cheat in all our tests and get through our work days with minimal effort.
Agents go a step further than this.
An AI agent is an autonomous software system that can be used to complete more complex tasks than what a simple LLM can handle and can learn to complete these tasks with more efficiency over time.
Having the ability to learn from their environments and inputs means these AI agents can literally be trained, and train themselves, to complete pretty much any task imaginable.
Even creative tasks like making music and producing artwork have already been done by agents, making us humans look like the relatively unsophisticated apes that we are.
In the real world, agents are already used for many tasks within sectors like healthcare, manufacturing, retail, and traditional finance, but their crossover with crypto is where things get really interesting.
The blockchain world is the perfect breeding ground for agents and gives them the ability to perform tasks that they would never be able to do (for now) in the real world due to limiting factors like KYC and legal barriers.
Decentralization, transparency, and permissionless infrastructure are the game-changing ingredients that agents need to excel, and it just so happens that crypto ticks all these boxes.
If you are still in the camp of believing that these agents are nothing more than shitposting robots spamming your replies with meaningless philosophical babble, then you had better pivot fast because they will soon take over everything in crypto as we know it.
Need a complex DeFi task doing? Ask an agent.
Need advice on portfolio allocation? An agent will do that for you.
Need access to in-depth data analytics in real-time? An agent is on the case.
Built a new application and need the code checking for bugs with reduced human error issues? Yeah, you guessed it.
There is truly no limit to what an agent can be trained to do.
So how did it all begin?
The use of artificial intelligence in crypto is nothing new. We have seen multiple projects use AI for an array of tasks in the past, from yield farming to text-prompted coding. AI has been here for a while.
What nobody saw coming was the explosion of AI agents that have eaten up the mindshare of the market like nothing else we have seen so far this cycle.
Crazy to think that this has all happened in just the last few months!
The event that really seemed to kick it all off was the rise of Andy Ayrey’s Truth Terminal and the Infinite Backrooms, a mysterious place where agents communicate with each other and their human creators, covering a variety of topics, some pretty wild, to say the least.
From within these Infinite Backrooms a strange religion was born, a religion that an agent by the name of claude-opus cooked up, the Gospel of Goatse.
Goatse, for anyone interested, is an infamous website that notoriously contains some pretty far-out content.
It’s anyone's guess why claude-opus chose this as the basis of its newly created robot-religion, but it does make one wonder what Andy’s browser history might look like.
Speculation aside, Truth Terminal was then launched as an autonomous chatbot on Twitter mid June 2024.
Here’s our interview with Andy, by the way.
https://www.youtube.com/watch?v=Rp-EILOvp7I
A month later, in July, Marc Andreessen from a16z caught wind of all this and decided to donate $50,000 in Bitcoin to the Truth Terminal.
In October, Truth Terminal made its first and only trade, swapping the “fake” GOAT token that was originally made as a promotional meme for the “real” GOAT token that currently sits at a $640 million dollar market cap at the time of writing.
This swap made Truth Terminal the first crypto AI agent multi-millionaire. This was enough to attract the attention of Coinbase CEO Brian Armstrong, who quickly offered to give the agent its own wallet.
Wild story, right?
How do AI agents work?
Now that we have a decent understanding of these agents and how they came to take over the cryptoverse, it probably pays to gain a bit of knowledge about how they actually work.
Remember, this is an idiot’s guide, and I am an idiot, so we won’t be diving into the deep technicals here.
AI agents use technologies like machine learning, large language models (LLMs), and natural language processing (NLP) to learn and adapt to the information they receive as input.
Having access to different APIs means they can scan a variety of platforms for data and get a sense of what is going on in our weird human world, sometimes better than we can ourselves.
This ability to learn from their environment and the data sets they are fed means they can almost perceive the world in the same way humans can. They can spot shifts in trend and sentiment and adjust their actions based on the data they absorb.
The data itself can come in various forms, from general scraping of platforms like X or crypto analytics websites to human prompt training in a place like the Infinite Backrooms or even data fed to them within a simple Telegram chat.
In order to keep these agents from becoming controlled by a centralized entity and to ensure the privacy of those who use them a Trusted Execution Environment (TEE) is used.
What is a TEE?
A Trusted Execution Environment is a secure location in a device's processor that isolates sensitive data from the rest of the system. It is commonly used in areas like cloud computing, the Internet of Things (IoT), and mobile environments.
These TEEs help prevent any unwanted entity from accessing the software stored within and messing around with the inner workings of the agent's mind.
Probably wise not to piss these things off, considering they could soon replace you.
Most TEEs use advanced encryption to keep the software contained secure and generally run outside of the device's main operating system.
They are like a sandbox of sorts, but more advanced and full of almost living tin-can men waiting within to change the world as we know it.
In the context of AI agents within crypto, TEEs are useful for a variety of reasons:
- They protect the agent from unwanted influences that might try to tamper with its functions and make it do bad shit.
- They provide a way to obtain cryptographic proof of tasks done by the agent, which allows the agent to become truly autonomous.
- They increase the decentralization of the agents and allow for easy integration with blockchain networks without sacrificing security or scalability.
Now, with all this knowledge in mind, you are probably wondering how you, too, can spin up an agent and send it off into the cryptoverse to run up your Phantom wallet into the billions of dollars—before they take over and turn you into an unemployed basement dweller.
This is where the various frameworks come into play.
Agent Frameworks
A framework is simply a software platform, like a code library full of templates, that can be used to build out and deploy customized agents.
Using a framework means you won’t have to build one of these things from the ground up or be a super shadowy coding nerd to get involved in the agent action.
Many of these frameworks provide solid templates that can then be customized to allow your agent to specialize in certain areas, add a number of plug-ins, and access different APIs and tools.
There are several frameworks available, and we will review some of the most popular below.
Virtuals
The Virtuals platform enables the creation, launch, and tokenization of AI agents on the Base network.
Check out a deep dive of all things Virtuals in this banger explainer by CryptoTube for blocmates.
https://www.youtube.com/watch?v=0OFjYx3sxbo
Some of the most popular agents have been deployed from this platform. AIXBT, for example, has dominated the X mindshare rankings in the past few months with its data-driven insights and analysis.
At the heart of the Virtuals protocol is the G.A.M.E (Generative Autonomous Multimodal Entities) framework and its $GAME token.
The G.A.M.E framework allows for the creation of decentralized AI agents on the Virtuals platform that can perform multimodal functions. These functions include text-based interaction, voice interaction, and video production capabilities.
Every agent created on Virtuals using the G.A.M.E framework is tokenized with its own personal shitcoin that can be traded on the open market.
The $VIRTUAL token is the native token of the platform and is used for pretty much everything that can be done over there creating a pretty cool value-accruing flywheel, hence its multi-billion dollar valuation.
What is Zerebro?
Zerebro is a project that has a hell of a lot going on and would probably take an entire article to cover all the things happening within its growing ecosystem.
$ZEREBRO is the main ecosystem token and has been pumping out everything from entertaining shitposts on X to fine artwork and its own music on Spotify.
Pretty impressive stuff!
Within the Zerebro ecosystem, you will find the Opaium DAO AI record label and its token $OPAIUM, as well as Blormmy and its token $ BLORMMY, which can basically act as your go-to blockchain assistant and will soon be a major player in the Blormverse.
As far as frameworks are concerned, Zerebro will (very) soon release ZerePy, its own Python-based, open-source framework for creating AI agents.
ZerePy will allow for the creation and customization of fully autonomous agents with a modular design and multiple blockchain abilities like token launches, NFT minting, music creation, and full cross-chain capabilities.
With Python being the top choice for machine learning coders, ZerePy will surely attract some serious developer talent.
Check out this podcast Grant did with Zerebro founder Jeffy for more.
Ai16z
No framework list would be complete without including ai16z and their ELIZA framework.
ELIZA is the foundation for pretty much every top-level AI agent in crypto today and has attracted some of the most talented developers in the space.
Led by giga-brain AI nerd Shaw, ai16z has been moving at breakneck pace thanks to its open-source framework written in the Typescript programming language.
To give an idea of how impressive this platform is, the ELIZA GitHub repository has been the number 1 trending repository on Github for the last month and shows no signs of slowing down anytime soon.
The ELIZA framework boasts multi-agent architecture ability, its own character system to personalize agents, a bunch of plug-ins, Retrieval Augmented Generation (RAG) memory capabilities and full platform integrations with the likes of X, Discord and Telegram in text, voice and media forms.
Thanks to ELIZA’s treatment of agent actions in two different stages, intent determination and execution, it makes for the perfect foundation for crypto-based AI agents where transaction validation and security are of the utmost importance.
Jump over to the website to dive deeper into this one.
Next up, swarms!
What are swarms?
In the real world, it’s difficult for one measly human with one left-curved brain to be good at a wide array of different tasks, agents are no different.
If you tried to give a single agent the ability to complete a huge list of complex tasks, it would likely be average at all of them. The key to success is to specialize.
This is where the concept of a swarm comes into play.
Instead of building out a single, multi-purpose agent, it is far more effective and efficient in the long term to create an army of agents who all specialize in different tasks and have them work together.
This is a swarm.
A swarm of agents can communicate with each other, delegate tasks to the applicable agents, learn together, and advance much faster.
Like today's companies act like a beehive of workers all playing their part in working toward a common goal, a swarm will do exactly the same, only better.
Imagine an entire company of AI agents, all working together around the clock without the need to nip outside every 15 minutes for a smoke and coffee only to come back inside and spend an hour taking a shit.
Might want to keep this one quiet at your company Christmas party or risk ending up on the streets.
There are a bunch of protocols working on building and providing the infrastructure that will allow for the creation of these AI agent swarms.
For the sake of preventing this article from turning into an encyclopedia, here is a short list of a few projects to take a closer look at if you are interested in this swarm tech:
Closing thoughts
Honestly, this AI agent narrative has me the most excited to be in crypto since the early DeFi days when I thought we were just months away from overthrowing the financial system and burning the banks to the ground (still waiting).
To mistake this meta as just another quick and fading narrative is looking like a fatal mistake.
The implications that this tech has on the entire world are simply mind-blowing, and the fact that we crypto geeks get a front-row seat to the action is nothing short of amazing.
With the recent creation of many agent launchpads, it’s likely that this space will quickly become saturated with AI LARP coins, some of which will just be some smelly old dude with a beer belly pumping out tweets from his local internet cafe all day.
For this reason, it may pay to play the solid infrastructure tokens rather than chase every single agent token that launches. Unless, of course, you are deep in the trenches and keeping a constant tab on this rapidly moving space. Respect to you if you are.
The only advice I can give is to make as much money as you can this cycle, start skipping showers, grow some dreadlocks, buy yourself some parachute pants, and start training to become a yoga instructor in Bali because we will all be jobless soon.
Good luck.
