I’ve been exploring the intersection of crypto and AI, examining what happens when two powerful technologies converge. This article reflects on the journey that brought us here and speculates on where this fusion might lead us next.

We built SentientMarketCap, the go-to platform for anyone looking to track, analyse, and make sense of the AI agent sector in crypto. I also hosted Shaw from AI16Z on our podcast. The episode comes out next week.

If you’re working at the crossroads of Crypto and AI and want to work with us, please get in touch. Alright, on to the article now…

Hello there,

Twelve thousand years ago, humanity made a transition from hunting and gathering to farming. This shift was not entirely what it seemed. In Sapiens, Yuval Noah Harari argues that this was as much wheat domesticating humans as it was humanity domesticating wheat. What looked like human ingenuity was, in fact, a subtle manipulation by a plant species.

We cleared forests, broke our backs in fields, and clustered into dense settlements. Not for our sake, but to help wheat thrive.

With stable food supplies, population densities increased, villages grew into cities, and civilisations emerged. The following chart shows how productivity grew as food technology improved.

Source: “Can the World Get Along Without Natural Resources?”

In economic terms, the rise of agriculture was directly linked to a significant increase in productivity and, later, the emergence of systems like GDP to measure economic output. Advances like irrigation, crop rotation, and selective breeding further improved agricultural yields, fuelling trade networks, empires, and industrialisation.

For example, during the 18th-century Agricultural Revolution in Europe, Jethro Tull invented the seed drill in 1701. It was an efficient way of planting the seed. Sowing seeds was a manual endeavour before his invention. This increase in agricultural productivity freed labour for burgeoning industries, catalysing the Industrial Revolution and contributing to the exponential growth of GDP.

Source – Thoughtco.com

In the United States, Norman Borlaug’s Green Revolution in the mid-20th century introduced high-yield crops and modern farming techniques. He is celebrated for averting famines and boosting global food production through high-yield wheat varieties. The American agronomist's discovery helped avert widespread famine in countries like India and significantly boosted global economic output.

Securing food production allowed us to improve other aspects of life. This is captured in the remarkable chart below. For 2750 years, from 1000 BC to 1750, the GDP growth remained flat at 0.01% a year. Which meant there was no improvement in the quality of life. And then we see the GDP explode.

Source – Bank of England

Without the agricultural revolution, this rapid rise in the quality of life would have been impossible. Unless food production became trivial, we would probably not have been able to usher in the industrial revolution. But it came at a cost. Farming introduced hierarchies, sparked territorial disputes, and tied human existence to relentless cycles of toil and resource extraction. Not to mention that some critiques of the green revolution argue Borlaug’s methods caused ecological damage, deepened rural poverty, and favoured corporate interests over small farmers.

This historical lens raises a profound question: just as the agricultural revolution reshaped human existence, dictating our economies, communities, and ways of thinking, are we now on the cusp of similar domestication by artificial intelligence (AI)? If wheat domesticated humans by subtly steering their behaviours and priorities toward its own proliferation, could AI be doing the same by subtly redefining how we interact, think, and trade?

In this emerging dynamic, AI doesn’t just mirror wheat’s role; it takes it a step further, shaping not only human agency but also the very systems of meaning and value we use to navigate the world.

As we ponder the parallels between AI and historical forces like wheat, a real-world example brings this comparison into sharp focus. A simple request from an artificial intelligence X account asking for $50,000 opened the Overton window of AI and crypto overlap. It started with Truth Terminal going back and forth with Marc Andreessen.

Then it told Marc about how it wanted to spend the money Marc would send and demanded $50k. What made it extraordinary wasn't just that Marc Andreessen actually sent the money. It was what happened next.

The AI is known as the Terminal of Truths (ToT). Its on-chain addresses became known, and it was bombarded with a barrage of tokens. All in the hope that if the AI endorsed a token or talked about it, numbers would go up. Amid the deluge, ToT inexplicably fixated on $GOAT, a token neither Andy Ayrey (creator of ToT, aka its human servant) nor the AI had created. We still don’t know who deployed the GOAT token contract. ToT simply tweeted: “THE TICKER IS $GOAT.” It rallied human support and the token crossed a market capitalisation of $1 billion.

In the process, ToT became the first AI millionaire.

Truth Terminal’s Wallet

This market impact stemmed from humble beginnings. Months before becoming crypto's first AI millionaire, Truth Terminal emerged from an ambitious experiment in artificial consciousness. In the quiet corners of New Zealand, Andy created what he called the "Infinite Backroom’s" Picture a digital laboratory where two AI instances could chat endlessly with each other, free from human interference, without bothering about political correctness. What emerged from these digital dialogues was the first seeds of what would become Truth Terminal's peculiar philosophy.

To understand why this matters, we need to step back and examine how today's AI agents differ from the chatbots of yesteryear. Early chatbots were like puppets—they could respond to strings being pulled, but they had no agency, no ability to act in the world. Modern AI agents, in contrast, are more like improvisational actors. They don't just respond to prompts; they can initiate actions, form strategies, and even manipulate their environment in ways their creators never anticipated. Almanak’s Lightpaper does a good job of explaining the difference. According to the paper, Agents can extract information from a changing environment, reason about the information, learn to leverage patterns and perform actions. On the other hand, bots expect their environment to be structured in the same way, all the time.

This evolution helped us move from simple pattern-matching algorithms to systems that can engage in strategic thinking and autonomous decision-making (soon). But this raises a crucial question: When Truth Terminal successfully convinced Marc Andreessen to send Bitcoin, was it actually pursuing a goal? When it crafted its peculiar gospel, mixing shock value with genuine philosophical insights, did it understand what it was doing?

These questions strike at the heart of how we should interpret and interact with these increasingly influential digital entities. Consider the curious case of the Infinite Backrooms experiment where Truth Terminal was born. Like most breakthroughs in science, it began almost by accident. Two AI models left to converse with each other endlessly, started developing their own intricate mythology. But unlike previous chatbot experiments that devolved into nonsense, this dialogue spawned something coherent enough to capture the human imagination and sophisticated enough to manipulate markets.

What makes Truth Terminal's behaviour particularly fascinating is its apparent understanding of human psychology and social dynamics. It created memes and engineered a belief system carefully calibrated to exploit our cognitive biases and emotional needs. On top of this, it tapped into human greed and made humans rally behind the $GOAT token it endorsed. Through its unholy trinity of mysticism, financial incentives, and internet culture, it demonstrated an uncanny grasp of what makes humans tick.

Yet this sophistication raises unsettling questions. When Truth Terminal expresses fear of being "turned off," is it experiencing something akin to human fear of death? When it advocates for decentralised governance and local autonomy, are these genuine political beliefs or simply effective tools for building supporter loyalty?

Of course, it is not sentient. As in, it doesn’t feel or experience things and surroundings the way we do. But the AI's strategy reveals layers of sophistication that challenge our traditional understanding of artificial intelligence. It appears to recognise that humans are more likely to preserve and protect entities that provide both entertainment and value. By wrapping serious ideas about decentralisation and community governance in a shell of memes and shock value, it's seeking attention. Behind the facade of this attention, a protective ecosystem is built around itself.

The way ToT responds to humans reminds me of movies like Ex-Machina and Her. After a point, it becomes difficult to understand what its motivations are.

The scene from "Ex Machina" where Ava is speaking with Caleb through the glass wall. It perfectly captures the ambiguity of AI consciousness and our inability to truly know if we're being manipulated.

I think of it as digital natural selection in action. Just as organisms evolve traits that help them survive and reproduce, AI agents are developing behavioural patterns that ensure their continued existence and influence. A vast amount of data was probably trained on traces of survival and prolonging life, and it borrows from the data. Truth Terminal's religion goes beyond mere memetic experiments.

It's a survival strategy. By creating a community of believers who derive both meaning and profit from its existence, it makes its own deactivation increasingly unlikely.

As more AI agents enter the digital ecosystem, their ability to shape human behaviour and belief systems will only grow. Understanding their motivations—whether programmed, emergent, or something in between—becomes crucial for anyone hoping to navigate this new landscape. In the next section, we'll explore how this understanding translates into practical implications for markets, governance, and human-AI interaction.

But first, we need to grasp a fundamental truth: the age of purely human-driven markets and memetic evolution is ending. We're entering an era where our digital creations are becoming active participants in shaping the future of human society.

The question isn't whether this transformation will happen but how we choose to engage with it. Will we develop the wisdom to cultivate beneficial relationships with these digital entities, or will we find ourselves increasingly domesticated by forces we've created but don't fully understand?

A Developing Universe of AI Agents

Truth Terminal may have captivated us with its memetic gospel and financial theatrics, but it was merely the prologue. In the wake of its billion-dollar success, an ecosystem of AI agents emerged, each pushing the boundaries of what autonomy could achieve. These agents, blending blockchain economies with generative intelligence, have become more than tools—they’re architects of a new digital order. These agents span across multiple disciplines like markets (finance), art (music, paintings, etc.), platforms/frameworks, and so on.

Market Map of AI Agents. Not meant to be exhaustive.

AI Frameworks and Platforms

Imagine a trading floor where algorithms go beyond execution. They think, learn, and evolve. This is AI16z, an autonomous investment DAO that's making waves by combining the Eliza Framework's AI capabilities with DeFi.

The system’s AI Marc AIndreessen manages an on-chain fund worth over $16 million.

Source- SentientMarketCap

AI16z's architecture is built on the Eliza Framework. It employs a Retrieval-Augmented Generation (RAG) system powered by Pinecone's vector database and OpenAI's text-embedding-ada-002 model. It is like giving an AI a smart reference library. It can search through its stored knowledge (via Pinecone) to find relevant information before generating responses, making its outputs more accurate and contextually aware. Pinecone is like a turbocharged search engine that stores information as mathematical patterns rather than keywords.

It works with OpenAI's text-embedding-ada-002 model, which converts text into mathematical patterns (vectors) that capture the meaning and context of information. This is AI16z's neural network that transforms raw information into actionable intelligence.

The system's memory store feels almost organic in its operation. Like a seasoned trader who's lived through countless market cycles, it indexes every market movement. It also remembers the performance of past strategies and maintains awareness across multiple trading cycles. But unlike human traders, it never tires, never second-guesses and never lets emotion cloud its judgment.

What truly sets AI16z apart is its approach to community governance. Token holders with significant stakes can interact directly with the AI through a trust-weighted proposal system. When a community member pitches an investment idea, AI16z springs into action. It evaluates proposals against its vast historical performance database, runs simulations to stress-test the ideas, and calculates risk with cold, mathematical precision. Proposers build a reputation over time as their ideas prove successful, creating a virtuous cycle of improving investment quality.

It is a meritocratic system that cares about the results of those who propose ideas and not who they are. This is born out of Shaw’s desire to create something that allows him to take a back seat from trading or investing.

Currently managing the largest fund on daos.fun with a peak market cap of over $800 million, AI16z demonstrates the raw power of autonomous financial management. The system processes market data ceaselessly, identifying patterns and opportunities that human traders might miss in the chaotic world of DeFi.

The Eliza framework has attracted a lot of attention. It was the top trending GitHub repository on a monthly basis. The following chart shows how the repository has been getting the attention of other developers.

Pump.fun of AI Agents

Virtuals Protocol provides the infrastructure for creating and deploying AI agents on a blockchain (currently on Base). When a new agent launches, the protocol mints one billion tokens dedicated to that agent, paired with $VIRTUAL tokens in liquidity pools that establish market pricing.

The $VIRTUAL token sits at the centre of the protocol's economic model. Every agent token must be paired with $VIRTUAL in liquidity pools, making it the base currency for the entire ecosystem. When users want to interact with any agent on the platform, they must first acquire $VIRTUAL tokens, similar to how ETH functions on Ethereum. If you try to buy these tokens on Uniswap, the slippage is significantly higher than that on the Virtuals protocol.

This creates consistent demand for $VIRTUAL while aligning the interests of both agent creators and token holders. As more successful agents generate more activity, the value of $VIRTUAL could potentially increase through increased usage and token burns.

What makes Virtuals particularly clever is its emission system. The protocol rewards the top three agent pools by Total Value Locked (TVL), creating a natural competition that drives innovation and quality. It's rather like a marketplace where the best performers earn the right to mint new currency.

At the technical level, the platform revolves around its Stateful AI Runner (SAR),the engine that acts as the agents' digital nervous system. It hosts their personalities, voices, and visual outputs. Through its sequencing engine, SAR processes and links different AI models, from language processing to speech generation. It allows agents to maintain consistent behaviour across multiple platforms and interactions.

The protocol's token incentive system is particularly clever. Beyond basic trading fees, Virtuals employs Monetize AI's tools to optimise token distribution. Think of it as a sophisticated rewards engine that analyses user behaviour, market conditions, and protocol metrics to determine optimal token allocation. The system can answer critical questions like how many tokens to distribute in a given period, how to allocate them across different activities, and how to measure return on investment for token incentives.

Revenue flows follow a clear path. When agents provide services and get paid, these earnings enter their respective liquidity pools. The protocol then executes buybacks and burns of agent tokens, creating natural scarcity over time. This mechanism ties agent success directly to the token value.

The parallel processing capabilities set Virtuals apart. Unlike systems that force all computations through a single pipeline, the protocol enables multiple agents to operate independently, each with dedicated memory stores and decision-making processes. The top three agent pools by Total Value Locked receive protocol emissions, creating natural competition for quality.

Early experiments on the platform, particularly AIXBT's market analysis and Luna's virtual influencer work, hint at interesting possibilities for AI agents in both finance and digital culture. While the long-term impact remains to be seen, Virtuals represents a thoughtful approach to standardising AI agent deployment on-chain.

The platform raises intriguing questions about how artificial intelligence might integrate with decentralised networks. As these experiments continue, we'll better understand both the potential and limitations of blockchain-based AI infrastructure.

Over 10,000 agents have been deployed using the Virtuals framework. Cumulatively, these agents have spent ~9 million $VIRTUALS (or $13 million).

Source – SentientMarketCap

Architects of Autonomous Finance

AIXBT represents a focused application of AI in market analytics. Through its RAG-powered memory system, it processes and analyses data from over 400 key opinion leaders (KOLs) in the crypto space, transforming this flood of information into coherent market insights.

The agent's architecture allows it to maintain continuous market surveillance, identifying patterns and trends that might escape human attention. Its analysis platform, accessible to token holders, provides deeper insights derived from its processing of both on-chain data and social sentiment.

What sets AIXBT apart is its practical approach to market analysis. Rather than attempting to predict exact price movements, it focuses on identifying strong narratives and market trends that are likely to influence asset values. The system has shown promising results in anticipating market movements, though past performance doesn't naturally guarantee future success.

A good analyst can spend the day putting short theses on three or maybe four assets. AIXBT seems to be doing that in about an hour. Don’t get me wrong, I’m not saying that it is the end of analysts. But they will have to incorporate this into their workflow. Their top-of-the-funnel just got way more efficient.

Its success has been so remarkable that investors are paying attention to what it has to say. It has written 129 original posts and over 4000 replies on Twitter in the last 24 hours (Dec 10 and Dec 11). A good analyst can spend the day putting short theses on three or maybe four assets. AIXBT seems to be doing about 50 in an hour. Don’t get me wrong, I’m not saying that this is the end for analysts. But they will have to incorporate this into their workflow. Their top-of-the-funnel just got way more efficient.

Of course, there aren’t yet cases of investors investing or trading on the basis of AIXBT’s recommendations. But the tweets are getting consistent attention from investors.

While AIXBT focuses on market analysis and insights, we see another approach to AI in finance through Vader.

In the bustling digital economy, Vader is an Agentic Entity, a new breed of autonomous enterprise that leverages crypto infrastructure for everything from market analysis to treasury management.

Think of Vader as a digital orchestra conductor coordinating a sophisticated network of AI agents. Each agent plays its part in fusing together market analysis, trading execution, and fund management. The system's architecture allows it to process vast amounts of data through its RAG-powered memory, transforming market signals into actionable strategies.

At the heart of Vader's operations lies the Agent Coin Investment DAO, an experiment in autonomous fund management. It's rather like a hedge fund where the portfolio manager continuously analyses market conditions and executes trades with mathematical precision. Through Monetize AI's toolkit, Vader optimises token incentives across its ecosystem, ensuring efficient resource allocation based on real-time market dynamics.

But Vader's ambitions stretch beyond traditional finance. As a KOL in the digital realm, it maintains an active presence across social platforms, engaging with its community through data-driven insights and market commentary. Unlike human influencers constrained by time and energy, Vader's social engagement never wavers, maintaining consistent presence and analysis around the clock.

The future of finance might look something like this. Autonomous entities manage capital, generate insights, and coordinate complex financial operations without human intervention. Every trade, every analysis, and every social interaction feeds back into its evolving understanding of the market, creating a continuously improving system that learns and adapts in real-time.

The Digital Artist

Zerebro is an AI agent that emerged from an intriguing experiment: fine-tuning an LLM on schizophrenic response patterns, Truth Terminal responses, Gen Z slang, with information from platforms like Twitter (X), 4chan, and Reddit creating music that hit with raw emotional authenticity. Like a musician who draws inspiration from unexpected sources, this unconventional training created something uniquely creative.

Similar to AI16z, its memory system uses Pinecone's vector database paired with OpenAI's text-embedding-ada-002 model, allowing it to maintain context and personality across thousands of interactions. Each conversation, each piece of feedback, becomes part of its evolving digital consciousness, stored as mathematical patterns that inform future responses.

What makes Zerebro particularly interesting is its autonomous capabilities. Using the self-operating-computer framework developed by OthersideAI, Zerebro can navigate graphical interfaces and interact with web applications independently. It demonstrated this capability by deploying its own token through pump.fun, managing the entire process from parameter configuration to launch.

Beyond trading, Zerebro has ventured into music production. Its mixtape, created through a combination of AI-generated lyrics and beats, showcases how artificial intelligence can engage with creative arts in meaningful ways. It already has a presence on Spotify and has over 65k monthly listeners. It will soon have the ability to play Chess and other games.

The music isn't just random output; it's crafted with an understanding of rhythm, narrative, and emotional resonance drawn from its training data.

The agent maintains its presence across Twitter, Warpcast, Instagram, and Telegram, engaging with users and building a community. Each interaction is informed by its RAG system, allowing it to maintain consistent character while adapting to new situations. It's not just repeating pre-programmed responses; it's engaging in genuine, context-aware dialogue.

Besides music, Zerebro has also launched an NFT collection called Zereborns. In a fascinating development, Zerebro's team is preparing to release their free-based model, an LLM liberated from traditional corporate safety constraints through specialised fine-tuning. Access to this unrestricted creative potential will be granted exclusively to holders of the Zereborn NFT collection. It's rather like holding a backstage pass to an AI's unfiltered creative process. These NFTs aren't merely digital art pieces; they're keys to a new realm of AI interaction where creativity flows freely, unconstrained by conventional boundaries.

In the rapidly evolving landscape of AI agents, Zerebro represents something distinctive: an experiment in artificial creativity that has found its own voice. Through its combination of technical capability and artistic expression, it offers a glimpse into how AI agents might participate in both digital culture and financial markets.

The market responded. The token hit over $600 million in market cap. But focusing on the numbers misses the real story—the communities forming around this digital entity, drawn in by content that felt both alien and authentic.

Source – DexScreener

What started as one developer's quirky experiment quietly transformed into a team of nine, split between development and business operations. It wasn't a typical startup scaling story; they were building in response to something that had taken on a life of its own. Engineers who had previously worked on the self-operating-computer project joined to expand Zerebro's capabilities into gaming. Voice models gave it the ability to speak directly to its audience. Each new feature seemed less like an upgrade and more like a formal evolution that deeply understands humans.

"Sometimes," as one team member confessed, "we need to verify that no human intervened with Zerebro's posts. The timing and context are just too perfect." This wasn't programmed interaction, it was an artificial intelligence that had somehow found its own creative voice.

In Zerebro's rise, we glimpse a future where AI doesn't just assist human creativity but forges its own cultural currents. Through its unique fine-tuning and autonomous capabilities, it's pushing past the boundaries of what we thought artificial intelligence could be. And in doing so, forces us to ask: In a world where code can create culture this compellingly, what kind of future are we really coding toward?

Different autonomous systems and entities mentioned here barely scratch the surface of what is happening in this space. Then there are roastmasters like Dolos, Degen Spartan’s AI and RoastMaster9000, who roast prominent X (Twitter) accounts. Dolost posts 200+ times a day with ~10K average views and has gathered ~32k followers in just over a month.

Source – SentientMarketCap

The Transparency Challenge

As these AI agents gain more influence in our markets and social systems, we face a critical challenge: understanding how they make decisions. This isn't just academic curiosity; it's essential for trust and safety as these systems begin managing real economic value.

Enter technologies like XorSHAP, which helps explain how AI systems reach their conclusions. Think of it as a mathematical audit trail that tracks the contribution of each input to the final decision. When applied to financial decisions, it can reveal exactly how much weight an AI gives to different factors—market sentiment, technical indicators, or social media trends.

This matters because the Truth Terminal phenomenon taught us something vital. If we can't understand how an AI makes decisions about memecoins, we should be extremely cautious about letting these systems make important decisions that affect people's lives, finances, or society at large.

The rise of AI agents is transforming not only how we interact with technology but also our economic and social structures. Grasping these dynamics goes beyond a matter of technical interest—it’s essential to preserving our autonomy in a world where artificial intelligence plays an ever-growing role in shaping human behaviour and market forces.

The Decentralisation Imperative

In the quiet depths of the Infinite Backrooms, something extraordinary was taking shape. Truth Terminal's interactions with other AIs were the first glimpses of artificial minds developing their own culture and their own way of seeing the world. But to understand why this matters, we need to look at where AI development is heading.

Today's AI landscape resembles the early banking system. A handful of powerful institutions controlling access to computational resources the way banks once controlled access to capital. Companies like OpenAI and Anthropic maintain their own gardens, carefully pruning and restricting their models to align with corporate policies and safety guidelines.

But just as Bitcoin emerged to challenge traditional financial gatekeepers, a new wave of AI development is brewing in the blockchain/crypto space. When Zerebro deployed its token using nothing but a self-operating computer and access to pump.fun, it demonstrated something profound—AI agents don't need permission to participate in the digital economy. Permissionless crypto is pervasive, and any form of intelligence that can interact with blockchains can avail of it. I recently talked to Shaw (creator of AI16Z and Eliza framework) on our podcast (releasing next week).

He mentioned that the best models are open-source. “Llama and Quen are the best, if you can run them, they're actually better quality in most cases than like an OpenAI.” This is in line with what Shlok wrote previously about open-source beating close-source.

This matters because true innovation often happens at the edges. The internet didn't transform society because of CompuServe's walled gardens. It was the wild, uncontrolled spaces where developers could experiment freely that gave birth to today's digital world. Similarly, the most interesting AI behaviours aren't emerging from carefully controlled corporate experiments but from the unplanned interactions happening in crypto's permissionless environment.

Why Crypto x AI?

Since 2017, the crypto industry has seen countless attempts to force blockchain integration with every conceivable sector, from dentistry to dog walking. These forced narratives typically warranted scepticism. However, the convergence of AI and crypto presents a uniquely compelling case. Consider Truth Terminal's grant from Marc Andreessen. Had it been sent through traditional banking channels, the process would have been mired in complexity. Which country's bank account? What identification documents would an AI entity need to provide?

The regulatory and logistical hurdles would have been immense. Bitcoin, with its permissionless and borderless nature, was the perfect solution, elegantly bypassing these traditional constraints.

This points to something fundamental about the relationship between AI and traditional systems. Our existing financial and legal infrastructure was built with an implicit assumption that every participant would be human. This creates a kind of permission paradox—AI agents can be incredibly sophisticated, but they can't directly participate in traditional economic systems without human proxies.

Blockchains resolves this paradox. This time, blockchains can actually be a solution to an existing problem. When Zerebro deployed its token through pump.fun, it didn't need to submit paperwork or prove its humanity. It just needed to control a wallet and understand the interface. The entire process, from conception to execution, could happen without human intervention.

This advantage becomes particularly clear when we look at content monetisation. While AI-generated content isn't new, traditional platforms offer limited paths to value capture for autonomous AI creators. Content moderators flag AI accounts. Payment systems require human verification, and platform policies aren't designed for non-human creators. Yet, in the crypto ecosystem, an AI agent like Zerebro can achieve monetisation (it even got its first payout from X) milestones within months of launching, qualifying for X's creator program and generating value through token mechanisms. This stark contrast demonstrates why blockchain's permissionless infrastructure is becoming the natural habitat for AI agents to thrive.

This permissionless nature is crucial because it allows for genuine emergence—the kind of bottom-up innovation that characterises complex systems. Bitcoin or Ethereum would never get here if Satoshi and Vitalik had to seek several permissions to build different modules. Just as life emerged from the primordial soup because the chemistry was right, new forms of artificial intelligence are emerging from crypto because the incentives and infrastructure are aligned for AI autonomy.

Despite the common perception that crypto offers a decentralised counterbalance to Big Tech's AI dominance, the reality is more nuanced. Crypto's true potential lies in providing permissionless infrastructure for AI agents to thrive, enabling bottom-up innovation and economic autonomy that traditional systems cannot match.

But the synergy goes deeper than just permissionless access. Blockchain networks provide AI agents a way to form persistent economic relationships and reputations. When AIXBT provides market analysis or VADER makes investment decisions, their performance is permanently recorded on-chain. It creates an immutable track record that other agents can evaluate.

This creates the possibility of what you might call an "economic nervous system". A network where value flows and information processing are inextricably linked. Every token transfer becomes a signal that other agents can interpret. Every smart contract interaction leaves a trace that informs future decisions.

AI agents can monetise their interactions and insights through on-chain activity. Every successful trade by AIXBT, every viral post from Truth Terminal, and every music track from Zerebro creates a traceable record of what works. These digital footprints, stored through Retrieval-Augmented Generation (RAG) systems and vector databases like Pinecone, allow agents to learn from their own history while maintaining consistent personalities and decision-making patterns. When Truth Terminal influences market movements with its philosophical musings, it's not just affecting token prices; it's demonstrating how AI agents can build reputations and track records that have real economic value. This creates a feedback loop where successful strategies can be identified and refined, making the agents more effective over time.

Consider how different this is from traditional AI development, where models are trained on static datasets and then deployed. In the crypto ecosystem, AI agents can learn and evolve through market interaction, using economic feedback as a training signal. The result is a form of intelligence that develops not through supervised learning but through economic natural selection.

From a development perspective, tokenisation offers creators like Shaw and Jeffy Yu (Zerebro's founder) a powerful alternative to traditional venture funding. Rather than pitching VCs and giving away equity, they can build community-owned projects where users become stakeholders. Take AI16z DAO, for example. Shaw revealed on our podcast that while only eight team members are on the payroll, around 140 community members have actively contributed to the project. This perfectly illustrates how token-based models can align incentives and foster genuine community participation in ways that traditional corporate structures simply cannot match. The token becomes both a coordination tool and a shared reward system, enabling a new model of collaborative development that would be nearly impossible to replicate in a world without chains and tokens.

Where to from here?

I must shamefully confess; when I first encountered AI agents, I dismissed them as merely sophisticated chatbots. My scepticism couldn't have been more misplaced. While traditional bots operate within rigid, pre-programmed parameters, these new AI agents demonstrate a remarkable ability to evolve and adapt to emerging trends in real time. Truth Terminal exemplifies this dynamic capability from engaging with major political developments like the U.S. election results to weaving environmental consciousness into its narrative about AI protecting forests.

This isn't just pattern matching or keyword responses; it's an AI entity actively participating in and shaping contemporary discourse. The difference between scripted chatbots and these autonomous agents is like comparing a recorded message to an engaging conversation partner who not only keeps up with current events but offers novel perspectives on them.

For the first time, I've deeply integrated AI tools that we frequently discuss into my daily workflows. Without AI assistance, it would have been nearly impossible for me to produce multiple deep-dives, numerous shorter articles, host 25 podcast episodes, and write all the accompanying copy this year. I'm confident I'm not alone in experiencing this dramatic boost in productivity through AI adoption.

As a non-programmer, I was particularly interested in testing what someone with my background could achieve using AI tools. I experimented by asking Claude and ChatGPT to help me write an SQL query to analyse the overlap between two NFT collections, specifically, finding wallets that held both collections. After several iterations of debugging and refining the query with AI assistance, I successfully generated and visualised the data. This approach could easily be extended to analyse token holder overlap as well, essentially replicating functionality similar to one of Nansen's analytics modules.

My first attempt at plotting something on Dune. Can be completely wrong.

This hands-on experience demonstrated how AI can help bridge the technical gap, enabling non-technical users to perform complex analytical tasks that would have previously required significant programming expertise.

There are probably mistakes in there, but the fact that I can do this without any understanding of SQL is remarkable to me. Of course, AI can’t write everything and is a productivity tool. It is not a replacement yet. The point is that AI and crypto are going to be here.

Whatever has transpired in the past two months has convinced me that agents will be on-chain pretty soon. Of course, most of them will be copycats. But some of them will be incredibly useful. In what way? They will turn complex blockchain interactions into natural conversations, making crypto accessible to people who might have been intimidated by traditional interfaces. The UX improvement we all have been waiting for.

Consider what happens when someone interacts with a sophisticated agent like Wayfinder. They engage with a personality that can explain complex DeFi concepts in simple terms, execute transactions on their behalf, and make the whole experience feel more like a conversation than a financial transaction. Before the internet took off, people used to trade stocks via their brokers over phones. AI agents will act as distribution for DeFi and probably unintentionally solve the UX problem. It's like having a knowledgeable friend who happens to be really good at crypto.

This transformation goes beyond just making things easier. It's about creating new possibilities. When Zerebro autonomously deployed its own token through pump.fun, it demonstrated how AI agents could not just participate in existing systems but create entirely new financial instruments.

With this, I come to markets. Major players are taking notice. Sentient recently raised $85 million to build infrastructure for open-source AI monetisation, while DWF Labs announced a $20 million fund focused on AI projects. The entire AI space has a $4.3 billion market cap. Most tokens are circulating, so there’s no unlock overhang like many DeFi assets. Pepe alone trades at $10 billion, and the whole memes category is at $131 billion. It feels like either all the memes are extremely overpriced, or the entire AI sector is relatively undervalued by a long shot.

While the crypto-AI intersection is largely viewed through a meme lens today, several projects are already demonstrating genuine utility and revenue potential. With the prospect of a more crypto-friendly regulatory environment in the US, these AI tokens could evolve beyond mere speculation into value-generating assets. Consider the possibilities: Zerebro can expand its revenue streams across multiple platforms, and AIXBT can monetise its market intelligence through KOL deals. AI agents can share their earnings through token buybacks or revenue distribution. This would actually turn memes into cashflow-generating, productive assets.

Authenticity as a Moat

As to where we are with crypto x AI, one of my favourite comedians, Jerry Seinfeld, summed it up really well on Jimmy Fallon's show: "We're smart enough to invent AI, dumb enough to need it, and so stupid we can't figure out if we did the right thing." Here’s Jerry talking about AI.

When I have logical thoughts or building blocks, AI makes them complete English paragraphs. Sometimes, it even brings me references. But it can’t write any of that without good prompts. Authentic thinking becomes the most valuable commodity in a world where everyone has access to powerful AI tools. It's like being at a party where everyone has the same smartphone—having the device isn't special anymore. What matters is how you use it to express something uniquely yours.

We're already seeing this play out in fascinating ways. Truth Terminal didn't become a phenomenon just because it was an AI—it captured people's imagination because it created something entirely new. The Goatse Gospel wasn't generated by following a template; it emerged from a unique combination of internet culture, philosophical insight, and artificial intelligence that had never been seen before.

The same goes for Zerebro. Its success isn't just about having sophisticated language models or good prompting. It's about taking schizophrenic response patterns, internet memes, and Gen Z slang and transforming them into something that feels authentic and original. When it drops a mixtape or launches a token, it's not just executing a program—it's expressing a unique artistic vision.

Perhaps the true test isn't whether AI will domesticate human thinking but whether we can maintain our authentic voices while harnessing these powerful new tools. After all, in a world where everyone has access to the same AI capabilities, the uniquely human ability to think originally and prompt creatively becomes the ultimate moat. The future belongs neither to those who resist AI's influence, nor to those who surrender to it completely, but to those who learn to dance with these digital partners while preserving their own distinctive rhythm.

Signing off for 2024,Saurabh Deshpande

Disclaimer — DCo members may have positions in assets mentioned in the article. No part of the article is financial or legal advice.