Tokens, Crypto, and Agents in the New Economy

Introduction — The Expansive Economy

In its 2025 Token Letter, Ribbit Capital discusses how token systems across industries are growing in complexity and value, a process it calls the Token Revolution. It argues that tokens, already powering AI, identity, and crypto, will become the basic unit of exchange and that winners will be those who master creating, moving, and using them.

This paper attempts to extend that analysis, focusing on where the most transformative effects on humans will emerge and how to position for them. I explore how tokens, crypto, and agents form the foundation of an expansive new economy, and how communication provides the theoretical glue between them. The Token Revolution is not abstract or speculative. It is already underway.

Tokens make information and value machine-readable. In its modern format, the token has become the universal primitive for representing and exchanging data, assets, rights, identity, and their combinations. I call this characteristic expansiveness, a token’s ability to connect and adapt. As tokens evolve, they gain capacity for relationship and refinement. As expansiveness compounds, tokens link into richer structures whose interactions can produce emergent properties and value. These compositional tokens are the basic unit of the expansive economy.

Crypto is the substrate of this economy, a live, permissionless coordination layer where tokens move and interact in real time. It lets humans and increasingly agents act globally, instantly, and without permission. It is the speed of the internet combined with the power of finance. Onchain, coordination replaces trust with transparent rules and finality. Crypto is the ultimate medium for open, composable markets at global scale.

Agents are the newest and most powerful participants in this system. Native to tokens, they speak the language of assets, protocols, and networks. Agents consume tokens, reason over them, and produce new tokens in return. They execute strategies, manage data, and operate across markets in real time. Agents will be the primary builders and operators of the token economy.

What connects all of this is communication. Tokens are data that communicate their own value. Crypto validates and makes those communications durable. Agents are token-native speakers, producing and interpreting messages with intent and consequence. Every transaction is a statement whose value reflects how well it coordinates its capacity for exchange.

I’ve worked in tokenization for nearly a decade. In real estate and fintech, I built ownership and financing models that increased owners’ autonomy and agency. Crypto made those models executable. What once required layers of contracts and intermediaries now runs onchain, where programmable rights and self-custody centers control in the hands of owners.

Tokens, crypto, and agents form a loop of creation and refinement, expanding what can be represented and built. That loop powers the expansive economy, where communication, coordination, and intelligence converge into an evolving system of value.

Tokens — The Atomic Unit of Value

Across systems and industries, the world’s knowledge is increasingly being rendered into tokens, capturing data, assets, ideas, relationships, and digital activity. A token can verify information, measure work, or transfer ownership, identity, or access. As these representations grow more complex and interconnected, they form a compositional web of meaning and value. This bridge between information and execution defines the expansive economy and puts tokens at its core.

Some principles define the foundation across token systems. First, anything of consequence, including any fact, right, asset, relationship, or behavior, can be expressed as a token. Second, tokens are the medium of exchange between digital intelligences. They are how software, increasingly including agentic software, transacts, reasons, and directs value. Third, value itself is compositional. It arises when tokens interact, when raw data is refined into structured meaning, and when meaning compounds into knowledge and action.

Tokens carry both content and context. Each token communicates its identity, ownership, and function to the systems around it. This communicative nature allows tokens to compose meaningfully, turning data into value. The property that captures this compositional process is what I call expansiveness, a measure of how well a token can combine, refine, and adapt. Expansiveness determines how a token grows in capability and relevance. It depends on factors like composability, flexibility across environments, refinability into higher-order states, and the quality of its supporting infrastructure. Tokens with high expansiveness are dynamic participants in a continuous process of value creation.

Tokens gain relevancy in proportion to their freedom to move and interact. In Homer’s Odyssey, Odysseus carries his oar inland until it is mistaken for a winnowing fan, marking the end of his voyage and the transformation of purpose from sea to land. The oar’s meaning changes because its context changes. Tokens evolve in the same way. When moved into new environments or interpreted by new systems, they take on new meanings and functions. Context becomes a source of value and a measure of expansiveness. Like Odysseus’s oar, tokens reveal their deepest meaning through context.

As expansiveness compounds across systems, tokens begin to merge into unified networks of meaning. I call this shared grammar of representation token unification. Every standardized contract, every data protocol, every onchain record contributes to the same process: the compression of human knowledge and coordination into tokens that can interact, combine, and evolve.

Around these processes has grown an ecosystem of token creators, refiners, combiners, and infrastructure builders, the architects of the expansive economy. Creators act as token factories, designing new models and primitives. Refiners build systems that add structure and context. Combiners find ways to connect tokens across domains. Infrastructure builders develop the rails that make these interactions possible. Together, they form the live machinery of token unification.

While token unification remains largely theoretical, the process is already visible in practice. Tokens coordinate defi markets, supply chains, creative rights, and data systems. As they begin to interact with intelligent agents, the line between representation and cognition starts to blur. At sufficient scale, tokens may begin to exhibit the characteristics of agency, including the ability to act, adapt, and express intent.

The current generation of agentic models is still far from that outcome. Large language models excel at pattern recognition but struggle with structured synthesis. They can describe relationships but often fail to build coherent hierarchies of meaning. This compositional gap is the defining frontier of the token economy. Closing it will require new forms of model architecture, symbolic reasoning, and token-native logic.

The token economy is already alive, and its first principle is simple: everything that can be represented will be. Every new system that encodes, connects, or interprets tokens accelerates that trajectory. As expansiveness compounds, value creation becomes recursive. Tokens generate tokens, systems learn from systems, and intelligence emerges naturally from coordination. Those who build the primitives and connections within this expanding network will define and capture the largest share of its value.

Crypto — The Coordination Grid

Crypto is the live, permissionless substrate of the expansive economy, where tokens move, interact, and acquire meaning. Every onchain transaction is a communication that gets verified and interpreted within a shared environment. Rules are written in code, code that unifies speech, value, and action.

I often hear people say crypto’s killer use case is speculation. That’s directionally true, but lacking in scope. Crypto’s true killer use case is coordination. It unites the speed of technology with the power of finance and the freedom of open communication. Much of the narratives around crypto, including things like presidential memecoins, digital gold, and various expressions of financial nihilism, are just reflections of the capabilities of the coordination surface at work, examples of a powerful network learning what and how to value.

Crypto replaces trust with transparency, resolving outcomes by consensus rather than authority. Every market becomes an experiment in coordination: stablecoins coordinate liquidity, DAOs coordinate collaboration, defi coordinates capital, prediction markets coordinate belief, and memecoins coordinate attention. Each action is a message to a network that interprets, enforces, and prices it automatically.

Crypto culture, including everything from onchain ponzis to 24-hr stock markets, are pure expressions of this communicative nature. Each new cycle, primitives for value expression emerge from this amalgamation of meaning. Every contract, every pool, every transfer is an encoded statement about what holds value.

This constant cultural experimentation meets an equally fluid legal reality. When technology moves faster than law, it creates permission gaps that become the frontier for innovation and value. Crypto entrepreneurs, predisposed to shipping fast, build products that meet investor demand while testing the edges of applicable regulation. ICOs, yield protocols, memecoins, and launchpads are all iterations of using technology to perform functions traditional systems can’t or won't. This dance between code, markets, and regulatory arbitrage shapes the evolution of the system itself.

Recent years have provided tailwinds almost unimaginable to those who built through early crypto cycles. A crypto-friendly administration, a pragmatic SEC, and participation from the world’s largest financial firms have accelerated the shift. The pace is almost hard to trust, but the scope will make it difficult to revert. From treasuries and equities to real estate, collectibles, data, and compute, onchain adoption is finding token-market fit. Crypto rails are swiftly becoming positioned as the infrastructure for how value is created, transferred, and coordinated.

Still, the spectacle of speculation often obscures the signal, making it important when writing about crypto to look past the shiniest light in the room. New launchpads replicate the logic of ICOs. Memecoins continue to evolve into social coordination tools. Prediction markets surface collective intelligence. Each cycle adds to the same experiment: coupling money with capacity, wrapping assets in ways that increase utility, virality, and compositionality. The result is an engine of coordination that learns in real time.

One of the adoption fronts I’m most excited about is the end-user wallet. As the place where funds already live, it’s a natural entry point for a full suite of services—a superpowered replacement for bank accounts. Traditional accounts offer little to no native yield and pile on fees, limits, and lockouts. Wallets, by contrast, enable incentive-aligned, programmable liquidity and control. Further, security on Ethereum and other networks keeps hardening as smart-contract standards mature. As Vitalik argued in his recent essay on low-risk defi, decentralized systems can already deliver stable, transparent, user-aligned products that prioritize individuals over fee extraction. Bringing banking onchain upgrades where people’s money lives and it’s getting safer fast.

Crypto is the world’s largest massively multiplayer online game. Participants build and play simultaneously, each transaction a speech act. What began as a field for human coordination is becoming the natural environment for agents, already fluent in tokens, logic, and incentives. As agents coordinate onchain at machine speed, crypto becomes the medium where communication scales into intelligence.

Agents — The New Participants

Agents are the natives of the token environment. They operate across markets and systems in real time, speaking the same language that tokens encode. Agents consume tokens to understand, coordinate, and act, and generate tokens when they produce value. Those tokens become more valuable as they acquire context—context which is created through reasoning and interaction. By performing the work of binding tokens into meaning and action, agents complete token unification, bringing expansiveness to the economy.

The current agentic landscape feels similar to crypto’s early years. It is uneven, often filled with speculative breakthrough moments that arrive faster than they can be interpreted or diligenced. The prevailing approach today is to train agents on narrow, well-scoped tasks and chain them together into specialized workflows. One outsized source of short-term value lies in pairing models with high-quality structured data. Proprietary datasets are especially powerful, though often require refinement. In the near term, the work is in sifting, labeling, and aligning data to tasks in ways that lower the burden on both users and systems.

Over time, agents may learn to perform this structuring on their own. As agents become better at sorting and synthesizing information, specialization begins to give way to more lateral reasoning. The unification of agents mirrors the unification of tokens. As tokens compose into higher-order representations, agents gain the ability to move horizontally across domains while deepening vertically into contextual expertise. Beyond this point, prediction gives way to speculation. An agent more capable of compiling and leveraging context might lead to a collection of coordinated specialists, or something closer to a networked general intelligence with greater-than-human capability. It is important to note, however, that many people with far greater technical insight than myself remain entirely skeptical of that outcome.

Regardless, the near-term economic implications are real. Businesses across industries are already applying agentic workflows to almost anything. Rather than survey this environment, I’ll focus here on a few areas I find interesting.

Professional Services

The day OpenAI launched o3 with research was the day I had my lightbulb moment on agents. I ran a test query to produce a 50-state regulatory review. A task that would have taken outside counsel two months and cost tens of thousands of dollars was completed in ten minutes. As someone who used to tell people that my job was to be better at Google than they are, it was a stark realization that something fundamental had changed.

This transformation extends far beyond law. Accounting, insurance, estate planning and any other profession where reading comprehension, synthesis, and structured writing form the basis of service will all be affected. These fields exist to translate human intent into institutional compliance, but document drafting, research, and rule interpretation are largely mechanical processes disguised as expertise. Agents can automate much of that translation, capably and at scale.

Whether these professions disappear or reorganize is unclear. Will we always want a human to represent us in a courtroom? Probably at least as long as we continue to make arguments in front of human judges. Regardless, progress rarely reverts and the current pace of change is blistering. If you work in any of these fields, ignore it at your peril.

Finance — from Neobank to Neurobank

The financial domain is already being rebuilt by agents. Agents-as-wallets move smart-wallet bank accounts (discussed earlier) from passive storage to active financial operators. Financial agents capable of implementing personalized strategies with speed and precision disrupt how we think about where our money should live. These agents blend the adaptability of personal financial advice with the execution speed of automated systems. In practice, agents can already manage directed yield strategies and portfolio rebalancing. Over time, they will move beyond pre-coded rules and into dynamic, self-optimizing execution.

The opportunity set extends across the entire financial stack. Lending and borrowing, staking, liquidity provision, arbitrage, and market-making can all be served autonomously. There is plenty of infrastructure to build. Personally, I’m interested in things like what an intelligent bank account should actually do; how we choose, trust, monitor and measure individual agents; and how we extend these capabilities to the widest set of people at the lowest possible cost.

Alongside the opportunity, two questions remain on my mind. First, how will agents differentiate in a market that approaches true efficiency. Second, if a system converges to near-perfect optimization, what role does risk even play. Perhaps noise becomes the new signal in a world of general intelligence. While much of that is still speculative, decentralized systems today are already capable of serving stable, transparent, user-aligned financial products. Agents are the natural operators of those systems.

For an alternative perspective, see:

Childhood Development and Education

Increasingly, the structure of education reflects politics more than pedagogy. As a parent, I often imagine what ideal learning conditions could look like—learning which is personalized, adaptive, and continuous. Neal Stephenson’s The Diamond Age presents a companion that grows with the child, providing them the tools to play with control and adventure in ways that leaves them ultimately capable and empowered. Agentic tutors integrate progress and adjust methods dynamically. They turn passive curriculum into interactive dialogue. Human presence remains irreplaceable, but agents extend a teacher’s reach, transforming education into a living feedback system, adapting to each child’s curiosity and pace rather than enforcing standardization.

AI and Frontier Science

Science is already undergoing its agentic turn. I recently had the opportunity to see researchers from Cambridge, Lawrence National Lab, and Meta present the largest molecular dataset ever assembled, produced specifically for training agents. Within weeks of its release, applied models are already replacing traditional methods in chemical prediction and materials discovery at massive scale. Meanwhile, companies are building agentic research systems that take scientists from hypothesis to paper, automating literature review, data generation, and experiment design along the way. The result is research as a composable, tokenized economy of knowledge.

Robotics and Embodied Intelligence

Physical agents extend cognition into material environments. Robots, drones, and autonomous vehicles coordinate through shared contexts and tokenized incentives. Crypto rails make distributed control and aligned behavior possible. Swarm robotics synchronize action through pattern, code, and feedback. These systems will underpin everything from logistics to infrastructure maintenance to personal defense to planetary exploration.

Agents are not now and may never become the body-morphing, time-traveling bogeyman we feared and later loved in the form of Arnold Schwarzenegger, but they are already capable enough that it’s hard not to ask when instead of if.

The Loop — Coordination as the Endgame

The combination of tokens, crypto, and agents forms a self-reinforcing system of communication, coordination, and cognition. Tokens make value legible. Crypto makes coordination scalable. Agents make both intelligent. Across domains, the pattern is the same. Agents consume tokens to perceive and act, and produce tokens when they create value. Tokens teach agents how to act. Agents teach tokens what to mean. Coordination becomes recursive and the expansive economy begins to think.

The loop expands because the components strengthen each other. More expressive tokens create richer grids for coordination. Maturing crypto environments allow faster, cheaper, and more global collaboration. Agentic advancements accelerate interpretation and action, compounding results. Every new token standard, every new surface of coordination, and every new agentic capability fuels further growth.

We are still early. The interfaces are awkward. The models reason unevenly. The way we talk to these systems is brittle and unnatural. Crypto remains culturally chaotic and unevenly distributed. But the direction and structure of the system are already visible. The opportunity is to build where representation, coordination, and intelligence converge, to create the tokens that become primitives, to develop the contracts that become rails, and to shape the agents that learn and adapt. The winners will be those who build the systems others coordinate through.

The expansive economy is here. It is growing through communication, learning through coordination, and evolving through intelligence. The work now is to participate in that loop with intention and humanity.