6 min read · 12 sources

Coinbase put four of the most traded equities on the planet on a public L2 yesterday. NVDAc, AAPLc, METAc and GOOGLc are live on Base under the new B20 standard, with each token backed 1:1 by shares held at Alpaca under Abu Dhabi Global Market supervision. The wrappers are real ERC-20s, which means they plug straight into Aave, Morpho, Aerodrome, 0x and 1inch without any new router code. On day one the four tokens did $4.55M in combined onchain value, $3.06M in DEX liquidity and $10.8M in 24-hour volume across roughly 50 integrated protocols.

Dividends and splits settle through an onchain multiplier that keeps your balance static, so a stock split can’t break your collateral position on a lending market. That detail matters more than it sounds: it is the difference between a token that lives in DeFi and a token that has to be unwound every corporate action. The launch is currently restricted to non-US users in eligible jurisdictions, but the technical surface is the point. Whatever you build on Base is now composable with equities.

Five new L2s raised $1.06B for roughly 99,000 DAU and generated $2.5M in combined chain fees, a 0.08% capture rate.

AI agents are crypto's next billion users, and they pay in stablecoins

Source: coindesk.com ↗

Coinbase’s head of AI product put the current moment as the “Napster/LimeWire era” of agentic payments. The framing matters: machine-initiated transactions happen without a human in the approval loop, which means onboarding funnels, KYC and the entire retail UX of crypto don’t apply. The agents already spend stablecoins for API calls, compute and services, and the rails that capture that wallet share now will be the settlement layer for machine commerce at scale.

For protocol teams, the implication is that stablecoin-denominated micropayments are a structural demand driver, not a cyclical one. A billion agents paying ten cents at a time is a different volume curve than retail traders flipping memecoins, and the throughput and fee design that works for one won’t work for the other.

Elysium is what HyperEVM should have been

Source: x.com ↗

Kinetiq, which already holds the majority of HyperEVM TVL, is launching Elysium as an L2 on top of Hyperliquid. The pitch is straightforward: HyperEVM is dual-block, throughput-constrained, and charges up to $20 per swap. Elysium swaps USDC gas for HYPE, ships an L1Read precompile that gives PropAMMs top-of-block oracle freshness, and targets block-time parity with HyperCore.

The interesting part for builders is the path it lays out. A native token generation platform feeds long-tail AMMs, then Elysium PropAMMs, then HyperCore spot, then HIP-3 perpetuals, all inside one ecosystem. Sequencer fees split 50/50: half goes to programmatic open-market KNTQ purchases burned into the Hyperliquid Assistance Fund, half split between builders and the Kinetiq treasury. Kinetiq claims no preferential treatment from Hyperliquid Labs or the Foundation, which is the part to watch.

EntropyIO puts a private company on Hyperliquid

Source: threadreaderapp.com ↗

EntropyIO launched liquid perpetual markets for private company exposure on Hyperliquid, with Anthropic as the first underlying. Most of the value in the AI stack is locked in private companies accessible only through VC rounds or secondaries with high minimums. Routing that demand through Hyperliquid’s permissionless order book gives you continuous price discovery on a name of Anthropic’s scale, and positions EntropyIO against trade[xyz] in the Hyperliquid-based equity perps niche. Whether the oracle problem is actually solved for private marks is the open question.

Kamino, Jupiter Lend, Loopscale: pick your risk model

Source: x.com ↗

DefiLlama-tracked stablecoin lending on Solana is roughly 92% concentrated across three protocols, and each one makes a different bet. Kamino runs a shared pool for main assets with isolated markets for riskier collateral, and has the deepest published security record in the segment: 15 audits, 4 formal verifications, dedicated fuzzing by OtterSec, Sec3, Offside Labs, Certora and Ackee, and no public exploits since its 2023 launch.

Jupiter Lend routes both depositor capital and leveraged borrowing through a single Liquidity Layer, with dynamic withdrawal ceilings that throttle liquidity-shock risk. Loopscale uses a Credit Order Book that pairs lenders and borrowers at fixed rates and durations, prioritising rate certainty over pool depth. Loopscale is also the only one of the three with a confirmed exploit: a $5.8M price feed attack in April 2025 where all funds were recovered in 48 hours via a negotiated 10% bounty. If you’re parking treasury stablecoins, the choice is variable-rate depth, shock-controlled earn-and-borrow, or fixed-rate predictability with thinner books.

Aria, Gauntlet's vault curator, is gated by humans

Source: gauntlet.xyz ↗

Gauntlet’s writeup on its Aria agent is the most useful operational post in the newsletter. Aria runs through staged autonomy: tool, scoped delegate, orchestrator, and every production vault change still requires one or two human approvals before going onchain. In production it has triaged 450+ risk alerts and reviewed 200+ pull requests, with 61 merged post-approval.

The architecture is the part to study. Offchain reasoning has unconstrained access to live data and compute; onchain authority is enforced through Aera’s merkle-root-committed guardian permissions, so delegation is cryptographically precise and auditable. Gauntlet built its own harness instead of using a commercial platform, citing ownership of controls and action verifiability as non-outsourceable in adversarial DeFi. The hardest requirement, per the post, is data integrity: any single price feed can be wrong, thin or stale, and disagreement between independent sources is treated as its own signal. General-purpose models with stale DeFi views do not meet the bar.

How every major exchange structures tokenized stocks

Source: threadreaderapp.com ↗

A side-by-side of Coinbase, Binance, Kraken, Robinhood and Backpack on redemption control, trading venue and dividend handling. Ondo still controls roughly 70% of the tokenized-stock market, but every major exchange is now shipping its own wrapper, and the structural differences (who can redeem, where the underlying sits, how dividends flow) are diverging fast. If you’re integrating any of these, the wrapper architecture is the integration surface, not the ticker.

Five new L2s raised $1.06B and made $2.5M in fees

Source: threadreaderapp.com ↗

The L2 post-mortem worth quoting directly: Monad, MegaETH, Plasma, Katana and Ink raised roughly $1.06 billion and pulled in nearly $3 billion in combined TVL, but generated only $2.5 million in chain fees to date. That is a 0.08% fee capture rate, and it works out to about $10,700 of capital raised per DAU across roughly 99,000 daily users. Plasma peaked at $5.5B post-launch then dropped 93% as early investors exited; Katana peaked at $677M pre-TGE; Monad is the only one with a consistent TVL trajectory. Monad leads in DEX activity, Ink leads in perps. TVL is not value capture, and the post is the clearest evidence of that gap the newsletter has run in months.

Crypto cards that aren't really non-custodial

Source: threadreaderapp.com ↗

A short thread on neobanks that market themselves as non-custodial but force users to delete their account to export private keys. The technical claim is that the double-spend risk (card swipe plus simultaneous onchain drain) is solvable with smart contract wallets and account abstraction, so the forced-deletion pattern is a product decision dressed up as a security one. Worth reading if you are evaluating wallet infrastructure for a card product.

Quick links

Source: threadreaderapp.com ↗

A security incident in the Cosmos EVM module prompted Cosmos Labs to advise validators across affected chains to halt operations while engineering investigates. On Plasma One, a €10 Wise deposit converted to $11.64 versus Google-quoted $11.66, a roughly 0.2-0.3% spread. Base tokenized stock quotes through Aerodrome came in about 32 cents under IBKR’s tiered plan and 97 cents under IBKR’s fixed plan at close. Fasset, an Islamic crypto neobank that just raised at a $1B valuation, is shipping a gold-backed card with Tether paying up to 6% cashback in XAUt.