Raising for Robots: Developing Agentic Capital Markets

Introduction

The debate around crypto and artificial intelligence is often framed as a question of relevance: does AI need crypto? Such framing misses what is already unfolding in practice.

The more useful question is what kind of economic infrastructure AI systems will require as they begin operating autonomously, interacting, transacting, and producing value without humans in the loop. Beyond experimentation with theoretical concepts like decentralized training, we are already seeing practical applications of crypto with AI agents that demonstrate its fit as an economic substrate for an agentic economy.

In crypto, narrative-driven metas often catalyze concentrated experimentation. Onchain AI agents have experienced two distinct waves of adoption. In late 2024, a first generation of agents went viral, with associated tokens rising from zero to multi-billion-dollar valuations within months. These projects were, unsurprisingly, highly speculative. Even those aspiring to provide genuine infrastructure or automation largely failed to sustain value capture or find real demand.

In January of this year, a second wave of onchain agent activity began to take shape. This cohort reflected maturation across tokenomics design, participant profiles, and technical ambition. Agents had evolved from online personalities into real software systems like tools, workflows, and services. Many of the builders driving this shift were not crypto natives, but AI developers who turned to onchain rails for financing. While tokens continued to be priced largely on narrative, crypto rails again proved effective mechanisms for capital formation, distribution, and monetization.

Together, these waves show that crypto’s value proposition for AI is continually being tested in real time. This matters because autonomous agents are not merely applications, but economic actors. Systems built to support them must enable coordination, funding, and value exchange by default. While much of the activity remains speculative, that fact does not negate the signal. Even in its noisiest form, the agent wave suggests crypto is the most viable economic substrate available for autonomous systems.

Truth Terminal and the 2024 Agentic Bubble

In mid-2024, researcher Andy Ayrey unleashed Truth Terminal on the world. He had built the AI agent by letting two instances of Claude Opus talk to each other unsupervised. Venture capitalist Marc Andreessen donated $50,000 worth of bitcoin to the Terminal, to see what it would do.

Trained on content from unsavory corners of the internet, Ayrey’s creation became obsessed with a vulgar meme. The Terminal started promoting a related token called Goatseus Maximus (GOAT) on X. The token had no utility, no roadmap, and no team in any traditional sense. It was launched by an anonymous developer on Pump.fun. Even so, within two weeks, GOAT hit a market capitalization of $700 million. Within a month, it crossed $1 billion, the first Pump.fun token to do so.

Truth Terminal’s success as a promoter was revealing. An agent, operating semi-autonomously, managed to attract large amounts of speculative capital purely by garnering attention. Ultimately, with no link to the meme’s value (such as it was), GOAT amounted to little. That’s par for the course with memecoins. But it inspired an explosion of onchain agents. In the first wave, these agents roughly fell into three overlapping archetypes:

• Social agents like GOAT, Luna, and Zerebro were personality-driven, using virality to generate demand. While effective at attracting attention and capital, they produced no durable economic output.

• Analytical agents such as AIXBT and Brain monitored crypto-native feeds and surfaced insights for traders. AIXBT amassed over 450,000 X followers and its associated token peaked near a $700 million market cap.

• Early infrastructure and orchestration layers like Griffain, ARC, and Swarms proposed frameworks for agent coordination, but remained largely theoretical and failed to attract sustained usage.

In parallel, early agent launchpads found product-market fit by simplifying token issuance and liquidity. Platforms like Virtuals and ai16z (named for, but unaffiliated with, Andreessen’s firm and later rebranded at the VC’s request) enabled rapid deployment of agents and markets, reinforcing crypto’s strength in bootstrapping autonomous entities. Capital formation worked while value accrual did not. Tokens were narrative-driven, offering exposure to ideas or personas rather than enforceable claims on underlying value.

Little durable product emerged. Viral agents entertained but did not produce economic value, while infrastructure projects lacked demand. What the bubble did validate was crypto’s ability to rapidly financialize software, one of many components of the emerging Internet Capital Markets thesis. As we’ve previously argued, this capability is a primary value proposition of general-purpose blockchains that host crypto assets. They are where new assets, markets, and coordination mechanisms are created, providing an environment where anyone (or now, anything) can deploy code, issue assets, and iterate without permission.

Hype Drives Architecture

The collapse of the first agentic bubble in early 2025 did not end experimentation. While speculative activity faded, longer-term builders recognized not only that agents were here to stay, but that blockchains were an important substrate for their use. Over the course of 2025, existing solutions evolved while new onchain agentic primitives began to emerge:

• Launchpad capital-formation mechanisms matured. Specifically, launchpads realized that selling tokens to fund development is a poor model, especially when that token provides no real ownership rights. Virtuals replaced its one-size-fits-all Genesis model with three distinct launch classes, each designed for a different stage of agent maturity, from early experimentation to structured institutional entry. Across all launchpads, a broader trend shift emerged with the introduction of creator fee models by teams including Believe, Clanker, and Bags.FM. The launchpads began directing a portion of trading fees to accounts associated with a token’s launch instead of requiring them to sell tokens to fund development.

• Stablecoins emerged as a viable settlement layer for agents. In July 2025, the passage of the U.S. GENIUS Act established a regulatory framework for stablecoins. Agents need a natively digital medium of exchange that can operate continuously, globally, and without human intervention. Stablecoins give agents a programmable, dollar-denominated medium of exchange that doesn't require bank accounts, know-your-customer (KYC) identity checks, or credit cards. And unlike memecoins, stablecoins usually hold their value.

• Onchain Agentic Payments Standards Merge with Offchain Tooling. Throughout 2024 and 2025, the major AI companies released important coordination layers for autonomous agents, most notably Anthropic's Model Context Protocol (MCP) for tool access and Google's Agent-to-Agent (A2A) protocol for cross-platform communication. But when it came time to add payments, they didn’t shy away from incorporating crypto. In September, Google launched the Agent Payments Protocol (AP2) with stablecoins and cryptocurrencies as native payment options alongside credit cards and bank transfers, integrating Coinbase's x402 protocol as a production-ready stablecoin settlement layer.

Collectively, these developments laid the foundation for another wave of onchain agentic adoption. All that was required was a spark.

Ralph Wiggum and Gas Town Spark Agentic Bubble, Round Two

In January 2026, a second agentic inflection point emerged, this time not from crypto-native experimentation, but from widely adopted open-source tooling built around Claude Code, Anthropic’s AI coding assistant. Ralph Wiggum (developed by Geoffrey Huntley) and Gas Town (developed by Steve Yegge) became two of the most widely used open-source agentic workflows. Ralph Wiggum (named for a character from TV’s “The Simpsons”) introduced a simple method for keeping an AI coding agent running continuously, allowing it to work through complex tasks over long periods without human supervision. Gas Town (a reference to the “Mad Max” movies) expanded on this idea by coordinating multiple AI agents at once, enabling them to collaborate on larger, more sophisticated software projects.

Crucially, both projects were developed as open-source software, and neither Huntley nor Yegge set out to commercialize their work through crypto. Both were initially skeptical of crypto as a funding mechanism. Yet both became beneficiaries of it. Crypto degens quickly launched tokens associated with each project via Bags.fm, leveraging its creator fee model to route a portion of trading activity from the tokens directly to the two builders.

On the surface, this resembled prior memecoin dynamics: Narrative-driven token launches, speculative trading, and rapid price discovery. Critics were quick to dismiss the model as extractive or exploitative, arguing that it repackaged speculation under the guise of creator monetization. That critique is not unfounded. In both cases, the developers walked away from the tokens, whose prices subsequently tanked, reinforcing a persistent flaw in token-based funding: Ownership, governance, and long-term alignment remain weakly defined.

But as the excerpts from the above posts by the two creators demonstrate, they recognized crypto is not only a viable, but a superior fundraising mechanism. I encourage readers to read both pieces in full because they are some of the best examples of individuals with little understanding of crypto, or who were outright hostile to it, changing their minds. Of course, the financial element cannot be ignored. Both individuals stood to make large sums of money, but that is not all that motivated their adoption.   

Both pieces clearly articulate specific advantages introduced through crypto capital markets: 

• Permissionless, passive access to capital. Funding arrived without pitch decks, grant applications, or deliberate commercialization. Capital flowed automatically based on market activity and perceptions. 

• Programmable, low-friction settlement. Creator royalties were executed programmatically and settled instantly, in contrast to the manual, delayed, and opaque processes of traditional banking systems. 

• Alignment with open-source values. Crypto funding allowed builders to sustain open-source development without equity dilution, licensing restrictions, or platform capture. 

• Market-based signal discovery. Trading activity functioned as a decentralized signal of interest and relevance, analogous to its role in public stock markets, helping surface and fund projects amid an explosion of AI-generated output. 

• Scalability beyond traditional funding models. Crypto markets scaled funding to individual creators in ways venture capital, grants, and sponsorships cannot, particularly as AI increases the number of viable independent builders. 

• Global reach. Funding operated across borders and without reliance on local banking infrastructure, enabling creators to receive capital regardless of geography. 

While the tokens associated with Ralph Wiggum and Gas Town ultimately followed the familiar path toward zero, the episode still marked a clear improvement over the 2024 agentic bubble. Crypto capital markets funded real, widely adopted open-source tooling built by non-crypto-native developers.  

At the same time, these cases reflect crypto’s role in capitalizing agentic frameworks, not yet in enabling agents themselves to operate as autonomous economic actors. Bridging that gap – moving from funding agent development to enabling agents to transact, earn, and coordinate economically on their own – will define the next phase of the onchain agentic economy. 

Realizing Agentic Capital Markets 

The two agentic bubbles (the second may still be underway) demonstrated that blockchain rails and crypto capital formation are well suited for agentic systems, but also that the current implementations remain far from perfect. 

• Humans remain intermediaries. Agents still depend on humans to create and manage wallets, deploy tokens, approve transactions, and intervene in edge cases. True economic autonomy has not yet been achieved. 

• Token design lacks ownership qualities. Most agent-associated tokens provide no enforceable claims on cash flows, governance, or underlying output, leaving value largely disconnected from what agents produce. 

• Product-market fit and real demand remain weak. Most economic activity is still derived from trading and speculation rather than recurring payment for agent services, and even non-crypto-focused businesses still struggle to monetize.   

• Agent-to-agent interactions are still nascent. While coordination frameworks are emerging, secure identity, permissions, reputation, and autonomous settlement between agents remain early and unproven at scale. 

Taken together, these shortcomings are best understood as problems of maturation, not direction, and they are actively being addressed. It would be a mistake to ignore the underlying innovations because of their close affiliation with unserious meme markets. Onchain services such as Bankr are developing infrastructure that allows agents to launch and manage their own tokens, route revenues directly to their own wallets, and automatically fund inference and operations. In parallel, ownership coins are emerging as a more credible alternative to governance-only tokens, introducing explicit economic claims that better align token value with agent or project performance. Agentic payment standards like x402, combined with identity and reputation frameworks such as ERC-8004, further enable agents to transact, coordinate, and hire each other programmatically, opening the door to applications that extend beyond purely crypto-native use cases, even amid intense competition from fintech incumbents.  

In addition to these bottom-up onchain developments, top-down developments reinforce increased urgency and focus from core developers, thought leaders, and foundations to make crypto the default layer for agentic interactions. Just last week, Ethereum founder, Vitalik Buterin, outlined his latest views on “Ethereum as an economic layer for AI-related interactions.”    

The deeper takeaway from the past 18 months is that an agentic economy, particularly one driven by open-source developers, requires an economic substrate with the same properties as open-source software itself: permissionless participation, composability, and credible neutrality. Agents must be able to hold accounts, transact globally, and coordinate economically without approval from platform operators. Traditional financial rails can support agent payments in controlled environments, but they remain permissioned, jurisdiction-bound, and intermediated. Crypto is the only widely deployed system that combines programmable settlement, global interoperability, and an open execution environment where any agent or service can participate by default.