In this week’s newsletter, Alex Thorn explains the significance of the SEC’s classification system for tokenized securities; Lucas Tcheyan breaks down Ethereum’s new ERC-8004 standard for AI agents; and Thad Pinakiewicz examines the Pendle protocol’s tokenomics overhaul.

📘SEC Defines Tokenization Terms

The Securities and Exchange Commission staff released guidance on tokenized securities Wednesday. Issued jointly by the divisions of Corporation Finance, Trading and Markets, and Investment Management, the document is largely definitional, providing a taxonomy for different types of tokenized securities and explaining their designs.

The SEC divides tokenized securities into two broad categories:

• Issuer-sponsored tokenized securities, in which a company makes its securities available on a blockchain, either by allowing the transformation from traditional to tokenized format, or by issuing securities natively onchain.

• Third party-sponsored tokenized securities, in which a third party (a tokenization company, crypto exchange, or asset manager) offers exposure to a security by some intermediary method. The market has experimented with two forms of third party-sponsored securities:

  • Custodial tokenized securities in which a company unaffiliated with the issuer creates a security entitlement that may or may not grant the holder economic exposure, governance, or legal rights to an underlying security. Market participants typically do this by placing shares of a security into a fund structure (like a special-purpose vehicle, or SPV) and then tokenizing the fund’s shares.

  • Synthetic tokenized securities may offer economic exposure to a security through a “linked security,” in which the third party issues its own security and designs it such that its return is linked to that of the underlying asset. Alternatively, the third party may issue a “security-based swap” (SBS) in tokenized form, another method of providing synthetic exposure with different regulatory obligations.

The statement doesn’t offer a qualitative view on the different models but instead creates a framework to inform investors about differences between instrument types and help guide discussions between market participants and the Commission. “We stand ready to engage regarding any questions,” the divisions write.

OUR TAKE:

Clarifying taxonomies is crucial for ensuring that a diverse group of stakeholders can engage constructively, particularly on an issue as complex and technical as the lifecycle, format, trading, and settlement of securities. These types of definitional advancements are also key parts of market structure and stablecoin legislation, and indeed most if not all comprehensive legal and regulatory frameworks. If the various stakeholders can’t agree on language, they won’t be able to proceed to designing new systems.

Galaxy has very publicly engaged in the first category in its tokenization of GLXY Class A Common Stock, which is an issuer-sponsored tokenized security. We criticized the third party-sponsored model (which has also been referred to as “wrappers”) in July and then followed up a week later when SEC Commissioner Hester Peirce wrote that such structures might be security-based swaps. We read this document to be consistent with Commissioner Peirce’s view regarding SBS, but it also properly describes the differences between the two high-level designs. When you own tokenized $GLXY, you own the same security as if you owned the traditionally issued version, with all associated legal, economic, and governance rights. When you own a tokenized security issued by a third-party wrapper, you enter into a relationship with the third party that may or may not convey various rights to the underlying security. As we wrote on July 11, “the wrapped structure obviates the relationship between issuer and shareholder to the detriment of both.”

But the wrapper model has benefits, particularly if the third-party sponsor doesn’t restrict transfers to allow-listed addresses, as is true for xStocks. The absence of such allowlists lets holders easily deploy the tokens in DeFi applications today. This is also a much more scalable way to get a bunch of stocks onto blockchains because the model does not require the acquiescence (or even knowledge) of the issuer itself. Essentially, each xStock is like a tokenized fund or even a modern stablecoin: of course, the collateral is one underlying asset rather than a basket of cash equivalents, but it also has freeze and seize capability without an allowlist (similar to USDC and USDT), allowing the token to flow to any unknown party but retaining the ability for xStocks to comply with sanctions. By volume, xStocks are the most successful such tokenized securities to date with $2 million to $4 million of DeFi trading volume per day. U.S.-based issuer-sponsored securities, like tokenized GLXY, in contrast, have almost no DeFi volume because they are true, native securities and the issuers are seeking further clarity from the SEC. The presence of an allowlist means that the issuer can also prevent most forms of DeFi trading, by requiring that an autonomous pool’s address be added to the allowlist.

For the issuer-sponsored model to prevail, which we believe is essential for the long-term health and durability of any emergent tokenized securities market, the SEC must work to level the playing field between the two models. Through wrappers, third-party sponsors can grow and indirectly give Americans access through DeFi while avoiding onshore rules. If the SEC can’t figure out how to let issuer-sponsored securities flow into DeFi and compete with the offshore wrappers, the market will gravitate toward structures that provide more utility regardless of what the rules say. – Alex Thorn

🤖Ethereum’s ERC-8004: When AI Gets an Identity

On Thursday, the Ethereum Foundation released ERC-8004, an open protocol for AI agent discovery and trust. At a high level, ERC-8004 defines a framework for representing agents as first-class onchain entities, with standardized identity, reputation, and validation mechanisms other applications and agents can rely on. The standard introduces three core components.

An Identity Registry, implemented as an ERC-721 non-fungible token (NFT), allows agents to publish canonical information about what they do, how they can be engaged, and which wallet they use to receive payments. A Reputation Registry enables users and counterparties to rate agents based on their interactions, with feedback cryptographically signed to ensure it comes from verifiable participants. And a forthcoming Validation Registry is designed to allow independent validators to re-execute tasks (using approaches such as zero-knowledge machine learning or zkML, trusted execution environments, or staked proofs) to verify outcomes and strengthen agent reputations over time.

Proposed in August, ERC-8004 builds on ideas pioneered in offchain agent frameworks, including Google’s Agent-to-Agent protocol, but extends them with blockchain-native trust and settlement primitives. The standard was developed with input from the foundation’s decentralized AI team, alongside contributors from MetaMask, Google, and Coinbase.

OUR TAKE:

A primary driver of crypto adoption in recent years has been its strength as a settlement layer for financial activity. Regulatory clarity has accelerated the adoption of stablecoins and tokenization, extending crypto’s relevance far beyond purely crypto-native use cases. But a parallel thesis is beginning to form (one we’ve been writing about since 2024), oriented not around humans, but around the role blockchains may play as critical infrastructure for AI and autonomous agents.

The internet is shifting from being dominated by human-initiated interactions to mediated by autonomous agents. Agents are beginning to search, transact, negotiate, and execute tasks continuously, often at a speed and scale that traditional systems were never designed to support. As AI systems scale, the limiting factor shifts from model capability to coordination; how autonomous systems interact, transact, and compose with one another without relying on a single intermediary.

Centralized platforms can and will address many of these challenges within their own ecosystems. At the same time, AI is dramatically lowering the barrier to creating software (i.e. vibe coding), unbundling application development from traditional engineering roles and enabling a much larger and more dynamic long tail of agents, tools, and services. As the supply of software expands, friction around onboarding, billing, and permissioning becomes increasingly costly, particularly for applications that fall outside the core focus of any single platform.

As autonomous systems come to represent a growing share of online activity, demand will rise for programmable layers capable of supporting identity, permissions, and payments for interactions with real economic value. Standards such as ERC-8004 and Coinbase’s x402 represent early building blocks of this emerging stack. ERC-8004 provides agents with persistent identity, while x402 enables native, programmatic payments for tools, data, and services. Together, they suggest how crypto’s role as a financial settlement layer could evolve into a broader coordination layer for AI — one designed not just to move value, but to make autonomous systems interoperable at scale. – Lucas Tcheyan

🗳️ Vote Escrow? Nope, Let’s Go

Pendle pulled the plug on its tokenomics this week, swapping out its entire governance incentive structure for a liquid staked version of its token. The vePENDLE era is officially over. In its place: sPENDLE, a liquid staking token with a 14-day withdrawal window (or instant exit ... for a 5% fee), and a fully algorithmic incentive system called AIM. Pendle is simplifying its governance, cutting emissions by ~30%, and admitting something many DeFi protocols avoid confronting: the system wasn’t working as intended, incentives were misaligned and third parties were stepping in to fill in client demand for a product Pendle didn’t offer: a transferable, yield-bearing, governance token.

Under vePENDLE (the prefix stands for “vote escrowed”), emissions were manually directed, often steered toward marginal or outright unprofitable pools, propped up by third-party and bootstrap incentives. Pendle’s own data shows over 60% of pools were unprofitable (>$1 of incentives per $1 of protocol revenue) with overall fee performance driven by a small minority.

AIM (which went live Thursday) replaces manual, vote-based emissions with an algorithm that allocates PENDLE according to quantifiable contributions to the protocol: total value locked (TVL) and fees generated. Emissions now flow to where Pendle is genuinely strong: large, liquid markets where it dominates price discovery (often silently; many of Pendle’s deepest markets aren’t even indexed by CoinGecko or CoinMarketCap). New pools still get a launch boost, but incentives taper as markets mature, encouraging early deposits without permanently subsidizing an asset.

At the same time, sPENDLE replaces vePENDLE’s multi-year, non-transferable locks with a 14-day un-staking period. Protocol revenue is used for PENDLE buybacks and distributed to sPENDLE holders who actively participate in governance. Such participation is simpler too; holders are deemed “inactive” only if they skip a proposal vote.

vePENDLE holders aren’t adversely impacted by the transition. They are allocated a one-off, non-transferable boosted sPENDLE balance (up to 4x, depending on remaining lock time), phasing out long-duration locks while explicitly accounting for Pendle’s earliest and long-term participants.

OUR TAKE:

Only ~20% of PENDLE supply was actively engaged under vePENDLE. Complexity, illiquidity, and multi-year locks created a system optimized for power users, supported by a cottage industry of Convex-style wrappers (Equilibria, StakeDAO) that frequently traded at a discount due to one-way vault dynamics.

This change collapses those ancillary markets, but that’s kind of the point. Those markets exist because there is persistent demand for composable (tokenized) yield-generating governance tokens, and the non-transferable, multi-year timescale of vePENDLE locks didn't match the fast money reality of DeFi. Pendle is moving from governance complexity to capital efficiency. Holders benefit from reduced governance complexity, with less need for frequent, manual vote management. And critically, sPENDLE can now be used across DeFi including, eventually, Pendle opening principal token and yield token (PT/YT) markets on its own LST and eating its own cooking.

This is Pendle maturing. It’s less friendly to frontier launches and mercenary liquidity, more biased toward scale, profitability, and investor demand. In a cycle where many protocols (including Curve and its new offshoot Yield Basis) are still clinging to ve- models out of habit, Pendle is one of the first to seriously unwind it.

This is just a preview of what is to come. Crypto is moving away from arm’s length agreements between foundations and DAOs, and unsustainable economics, to true ownership coins and direct, sustainable incentives. But as a case study in how to survive a major economics change without blowing up your community, this is about as clean and as fair as it gets. – Thad Pinakiewicz

Charts of the Week

The market cap of tokenized stocks on the Ethereum and Solana blockchains continues to climb to new highs, driven by demand for and experimentation with onchain stocks and related assets. As of January 28, the market cap of stocks on the two chains was $618 million across 277 tokens. Ondo’s SPYon is the largest market cap stock token on Ethereum with a $26.7 million market cap; xStocks’ TSLAx is the largest market cap stock token on Solana with a $50.8 million market cap.

Other News

🇺🇸White House to convene crypto, bank exec confab on market structure

🔮CFTC writing new rules for prediction markets, chairman says

🔐Private version of Circle’s USDC goes live on Aleo blockchain

🔷Ethereum seeds $220m security fund with unclaimed funds from The Dao

🪙Fidelity to launch its own stablecoin; Tether debuts U.S.-regulated USAT

🔏Bitwise and Kraken introduce DeFi vault products

🦄Crypto payments network Mesh raises $75m at $1b valuation

🌲WisdomTree expands tokenized funds to Solana

👁️OpenAI considers World Orb eyeball scanner for social network

💰Fairshake PAC amasses $193m war chest ahead of November midterms

🍊Citrea activates mainnet for ZK rollup on top of Bitcoin network