Sam Green, founder of Cambrian Network, and Austin Griffith, Head of Builder Growth at the Ethereum Foundation, appeared on episode 73 of The Block's The Crypto Beat podcast on April 10, hosted by Tim Copeland and Kelvin Sparks, to discuss how AI agents are reshaping crypto development, trading, and risk management. This article draws on reporting from that conversation alongside secondary sources; direct claims from the episode have not been independently transcription-verified, and figures sourced from industry publications are attributed throughout.

Their conversation arrives as data from blockchain protocols and industry analyses begins to reflect what has largely been theoretical: software agents are moving real money, executing trades, and interacting with decentralized protocols with minimal human oversight.

---

The Numbers Behind the Hype

Cambrian's Q1 2026 Agentic Finance Landscape Report, published February 23, 2026, counted 51 live projects across four categories: trading and portfolio optimization (23 projects, including AskJimmy, HeyAnon, Bankr, Glider, and Velvet Capital), yield agents (21 projects, including AFI, Almanak, Kamino, Lulo, and ARMA by Arrakis Finance), prediction and betting agents (3), and analysis agents (4).

The report flags an important caveat: the systems managing the most assets are still rule-based rather than fully autonomous. Cambrian maps projects along two axes, Intelligence and Autonomy, and the leading high-asset-under-management platforms cluster toward fixed, rule-based logic rather than adaptive decision-making via large language models. Full AI autonomy over significant capital remains the frontier, not the norm.

Still, the infrastructure supporting those agents is scaling fast. The x402 protocol, an open payments standard built on the HTTP 402 status code that allows AI agents to pay for services in stablecoins without human involvement, is currently live on Base and Solana. It logged more than 15 million transactions in a single 30-day window as of Q1 2026. Cambrian puts cumulative x402 volume at $50 million across that same window; x402.org's own site claims $600 million total since launch, a discrepancy that reflects different measurement windows rather than conflicting data. On the Solana deployment specifically, post-launch data shows more than 35 million transactions and over $10 million in volume, context that is particularly relevant for regions where Solana's low-fee infrastructure supports financial accessibility arguments.

Roughly 500,000 active AI wallets are now interacting with the protocol. Early adopters include Coinbase, Google, AWS, Visa, Mastercard, Stripe, Shopify, and Microsoft.

Separately, ERC-8004, an Ethereum standard that functions as an identity and trust registry for AI agents, launched on mainnet on January 29, 2026, and had more than 24,000 agents registered shortly thereafter.

A second standard, ERC-8211 (informally called "Smart Batching"), co-developed with Biconomy, allows agents to execute multi-step transactions without locking in every parameter at the time of signing, which is a meaningful operational improvement for systems that need to adapt mid-process.

---

Ethereum's Explicit Play for the Machine Economy

The Ethereum Foundation is not treating AI agent activity as incidental. It created a dedicated team called the dAI team, led by Davide Crapis, with a stated goal of ensuring AI commerce does not default to closed, centralized platforms. "If AI doesn't have the properties we care about (self-sovereignty, censorship resistance, privacy) and then we use AI for everything, basically no one has those properties anymore," Crapis told CoinDesk in March.

Crapis has also described the challenge in more concrete terms. He envisions the trust infrastructure for AI agents as "a decentralized version of Google Reviews combined with payment rails," and has framed Ethereum's strategic ambition plainly: "In a world where AI is in the wild, we want Ethereum to be the place with the big lock."

Griffith's contribution to this effort includes ETHSkills, a project designed to close the knowledge gap between AI agents and what production Ethereum development actually requires.

He presented on agentic workflows and DAO governance participation at ETHDenver 2026, emphasizing that trustlessness is not a native property of current AI models and that security must be the starting point for any agentic system built on public blockchains.

---

Security Risks Are Already Materializing

The episode also covered DeFi risk layering and the specific vulnerabilities that AI-driven systems introduce. Those risks are not hypothetical. According to KuCoin's security analysis, a January 2026 breach at Step Finance drained roughly $40 million, involving more than 261,000 SOL transfers. This figure comes from a single commercial source and has not been independently corroborated at time of publication. Separately, AI-assisted social engineering and deepfake attacks caused approximately $45 million in losses, according to the same analysis.

A survey cited by KuCoin and cross-referenced with OWASP 2026 data found that 88 percent of organizations reported confirmed AI agent security incidents in the prior year.

---

Regional Stakes: Africa and South Asia

The agentic finance story has direct relevance outside the United States. VALR, a Johannesburg-based exchange backed by Pantera Capital, serves 1.8 million users and 2,000 institutional clients and launched a dedicated AI agent service on April 10. Readers should note that the VALR launch date coincides with The Block's podcast publication date; reporting notes that the timing may reflect coordinated announcements within the agentic finance ecosystem rather than independent corroboration, and that context is worth bearing in mind when assessing VALR's data as a separate data point.

The platform supports agents built with tools including OpenClaw, Anthropic's Claude Code, OpenAI Codex, and OpenCode, positioning African users and developers as direct participants rather than observers.

India's exposure is structural. The country has roughly 150 million active crypto users (a 2025 figure; no updated 2026 count was available at publication), a total that surpasses both the United States and China, making India the world's largest crypto user base by this measure. More than 75 percent of those users are located in Tier-2, Tier-3, and Tier-4 cities, a distribution that shapes the infrastructure requirements for agentic finance significantly.

An estimated 20 to 30 percent of global Web3 developers are Indian, and India is home to more than 1,200 active blockchain startups. Fetch.ai and Bittensor rank among the most cited AI agent developer frameworks in the Indian Web3 community. Yield agents and stablecoin payment rails are particularly relevant for India's large population without access to formal banking, though a precise figure for that population was not available at publication. Regulatory uncertainty around crypto classification remains a barrier to deployment at scale.

A practical gap persists across Sub-Saharan Africa: most AI agent tooling assumes fast, cheap internet and formalized identity systems. Mobile-first, intermittent connectivity and informal financial profiles require infrastructure adaptation that has not yet happened at meaningful scale. One concrete example of early progress comes from Rwanda, where startup Kayko had onboarded 8,500 businesses into automated financial ecosystems by early 2026, illustrating what agentic adoption can look like when infrastructure is built for local conditions. A documented pattern across the continent also involves African entities partnering with Asian firms that carry deep experience in large-scale automation, a model that may accelerate regional rollout.

The x402 protocol's deployment on Solana is directly relevant here, given that Solana's low transaction fees align with the cost constraints of mobile-first users across Africa and South Asia.

---

What Comes Next

Green, who previously co-founded Semiotic Labs and helped build Odos (a decentralized exchange aggregator that has processed more than $90 billion in volume across 3 million users), told HackerNoon in an interview that AI agents will account for the majority of on-chain transactions within two to three years. HackerNoon operates an independent contributor and business blogging program, and the claim has not been independently corroborated, so it is best read as a directional view from a founder with a direct stake in the outcome rather than a consensus projection.

Significant structural obstacles remain, including the continued dominance of rule-based systems among the highest-asset-under-management platforms, infrastructure gaps in the Global South, and ongoing regulatory uncertainty in key markets such as India. Whether the two-to-three-year timeline proves accurate, the general direction toward greater agentic participation in on-chain activity has broad support across the industry participants and data sources reviewed here.

According to KuCoin, a commercial exchange with its own interests in this narrative, AI agent token market capitalization has already exceeded $7.7 billion, with daily trading volume near $1.7 billion. Readers should treat these as industry estimates rather than independently audited figures. A Gartner projection cited in a VALR press release (sourced via Benzinga, rather than a direct Gartner publication) puts 40 percent of business applications integrating AI agents by the end of 2026. Separately, the global AI agents market is projected to grow from $8 billion in 2025 to $50 billion by 2030, though that figure also originates from a press release rather than a primary research citation.

The open question is whether public blockchains, Ethereum in particular, can establish the standards and security guarantees fast enough to anchor that activity before closed systems do.