# Castle Labs — Full Content Index > Castle Labs is a crypto research firm publishing free, independent blockchain research and digital asset research and advisory. This file indexes every research article, flagship report, topic hub and live data page on castlelabs.io with stable URLs for citation. Castle Labs publishes long-form crypto research on perpetual DEXs and derivatives, DeFi, tokenisation and RWAs, stablecoins, privacy, prediction markets, L2s, AI x crypto and regulation. Everything is free to read, with topic hubs, flagship reports, live data and full article pages structured for citation. ## Audience & Distribution - 7,000+ newsletter readers across Castle Labs research channels - 2,500+ institutional readers across institutional finance, crypto-native trading, infrastructure, DeFi, RWAs, research, and media - Arbitrum Unlocked 2.0 campaign delivered 35K Substack views and 22K unique opens - Newsletter open rate: 24-27% ## Flagship Reports (free, download via DocSend) - [The Evolution of Commodities](https://castlelabs.io/research/reports/the-evolution-of-commodities): State of Commodities and a Quantitative Analysis of the Growth of these Markets Onchain (March 2026, 71 pages; download: https://docsend.com/v/sjv2g/the_evolution_of_commodities) - Finding: Tokenised spot commodities reached $7.4 billion, with tokenised gold responsible for most of the category. - Finding: Commodity perpetual open interest on Hyperliquid exceeded $1.4 billion as HIP-3 expanded access to metals, energy and equity indices. - Finding: CME revealed liquidity was 109 times deeper within 5 basis points, while xyz:CL weekend volume rose from $31 million to more than $1 billion across the three-week study. - What Are Tokenised Commodities? (https://castlelabs.io/research/reports/the-evolution-of-commodities#what-are-tokenised-commodities): Tokenised commodities are tokens backed by, or tracking, a physical commodity such as gold, oil, soybeans or electricity. Spot tokens represent ownership of the underlying; commodity perps give leveraged exposure without expiry. Both trade around the clock and settle onchain. In March 2026 spot tokenised commodities were worth $7.4 billion, over 27% of the $27.35 billion of tokenised real-world assets, up from $1.9 billion a year earlier. Gold is most of it: over $5.9 billion, more than 75% of the onchain spot commodity market. Tokenised gold traded over $178 billion in 2025, up 227% year on year and second only to the GLD ETF globally. The largest tokenised spot commodities, March 2026 (source: Castle Labs, The Evolution of Commodities (March 2026), from RWA.xyz data) | Commodity | Main tokens | Value onchain | | --- | --- | --- | | Gold | Tether Gold (XAUT), Paxos Gold (PAXG) | Over $5.9B | | Electricity | JWMH on Ripple | Over $800M | | Soybean oil | JSOY_OIL (JusToken, Polygon) | Over $380M | | Soybeans | JSOY (JusToken, Polygon) | Over $300M | | Diamonds | Tokenised diamonds on Ripple | Over $100M | | Cotton and corn | JCOT, JCORN (JusToken) | $19M and $16M | - Can You Trade Oil on Weekends? Hyperliquid vs CME (https://castlelabs.io/research/reports/the-evolution-of-commodities#trade-oil-on-weekends): CME commodity markets run about 6,400 hours a year, roughly 73% of calendar hours, and close at weekends. Onchain venues run all 8,760 hours. When the US and Israel struck Iran on Saturday 28 February 2026, CME oil contracts stayed frozen until Sunday evening, while Hyperliquid's oil perp on TradeXYZ (xyz:CL) kept trading. We tracked xyz:CL against CME's April WTI contract (CLJ6) across the next three weekends. Hyperliquid's price is held inside discovery bounds when the CME reference is closed, which capped price discovery in the first two weekends. Weekend volume grew from $31 million to over $1 billion in three weeks, and the average weekend trade grew from $1,199 to over $1,500, a sign of hedging flow joining retail speculation. Three weekends of onchain oil trading, February to March 2026 (source: CME CLJ6 trades via Databento; Hyperliquid xyz:CL fills from Hyperliquid's public S3 archive and API) | Weekend | CME Friday close to Monday open | Hyperliquid xyz:CL | xyz:CL volume | xyz:CL trades | | --- | --- | --- | --- | --- | | 28 Feb to 1 Mar (Iran strikes) | $67.29 to $75 (+1,146 bps) | Reached about $70.80, roughly 45% of the gap, capped by discovery bounds | $31.3M | 26,140 | | 7 to 8 Mar | $91.27 to $98 (+737 bps) | Reached about $95.83, 68% of the move | $611.6M | 529,232 | | 14 to 15 Mar | $99.31 to $100.93 (+163 bps) | Traded at $101.56 at the CME open, a slight overshoot | $1,072.1M | 706,634 | - Hyperliquid vs Ostium for Commodity Trading (https://castlelabs.io/research/reports/the-evolution-of-commodities#hyperliquid-vs-ostium): The two largest onchain commodity venues make opposite design choices. Hyperliquid builds native order books and keeps markets open 24/7; Ostium quotes from the traditional markets' own prices and pauses market orders when those markets close. How the two venues differ (source: Castle Labs, The Evolution of Commodities (March 2026)) | Metric | Hyperliquid (HIP-3, mostly TradeXYZ) | Ostium | | --- | --- | --- | | Pricing | Onchain order book; Pyth reference with discovery bounds off-hours | Direct oracle pricing from Chainlink Data Streams and Stork | | Weekend trading | Yes, 24/7 | Market orders pause when the underlying closes; limit and stop orders queue | | Maximum margin multiple | Up to 25x on commodities | Up to 100x in some markets | | Scale | HIP-3 processed over $109B; commodities are over 16% of platform volume | Over $45B cumulative volume; commodities over $125M of about $200M OI | - Chart: Commodity market size by product (https://castlelabs.io/research/reports/the-evolution-of-commodities#chart-commodities-market-size, report page 8): The global commodity market moves about $143 trillion in notional value. Oil is the largest single market at about $3.5 trillion, ahead of gold at $1.5 trillion and natural gas at $1.2 trillion. - Chart: Relative performance of commodities, 1975 to 2025 (https://castlelabs.io/research/reports/the-evolution-of-commodities#chart-commodities-relative-performance, report page 8): Indexed to 1975 prices, most commodities have rallied since their pandemic lows in March 2020. - Chart: Top five tokenised spot commodities (https://castlelabs.io/research/reports/the-evolution-of-commodities#chart-tokenised-spot-commodities, report page 17): Paxos Gold and Tether Gold make up most of the market. Both trade around the clock, settle instantly and move across chains, which traditional gold markets cannot do. - Chart: HIP-3 volume as a share of Hyperliquid volume (https://castlelabs.io/research/reports/the-evolution-of-commodities#chart-hip3-share-of-hyperliquid-volume, report page 27): HIP-3 markets have processed over $109 billion and now make up about 25% of Hyperliquid's volume. Commodity trading alone is over 16%. - Chart: HIP-3 volume by deployer (https://castlelabs.io/research/reports/the-evolution-of-commodities#chart-hip3-volume-by-deployer, report page 28): TradeXYZ runs over 85% of HIP-3 volume. In February, Hyperliquid did over $4.09 billion of silver volume in 24 hours, more than many traditional venues trade in a month. - Chart: Ostium daily commodities volume (https://castlelabs.io/research/reports/the-evolution-of-commodities#chart-ostium-daily-commodities-volume, report page 32): Over 95% of Ostium's open interest is in non-crypto assets. Cumulative volume passed $45 billion, metals drove over $7 billion of it, and gold is the most traded market. - Chart: Liquidity depth: Hyperliquid oil (xyz:CL) vs CME (CLJ6) (https://castlelabs.io/research/reports/the-evolution-of-commodities#chart-oil-liquidity-depth, report page 54): Within 2 bps of mid, CME offers $19 million of executable depth against $152K on Hyperliquid, a 125x gap. Onchain depth is under 1% of CME's. - Chart: Slippage by order size during CME trading hours (https://castlelabs.io/research/reports/the-evolution-of-commodities#chart-oil-slippage-overlapping-hours, report page 56): A $100K order slips 4.33 bps on Hyperliquid and nothing on CME. At $1 million it is 15.4 bps against 0.79 bps, about 20x, so the venue is not yet built for institutional size. - Chart: Weekend slippage on Hyperliquid oil, three weekends (https://castlelabs.io/research/reports/the-evolution-of-commodities#chart-oil-weekend-slippage, report page 57): Across three weekends, simulated slippage fell 16% for $10K orders, 75% for $100K orders and 86.9% for $1 million orders as the market matured. - Chart: Hyperliquid (xyz:CL) and CME (CLJ6) oil prices (https://castlelabs.io/research/reports/the-evolution-of-commodities#chart-hyperliquid-vs-cme-oil-price, report page 59): During CME hours the two prices track closely, though a discount opened and widened as oil rose, most likely from funding pressure on crowded longs. At weekends the price band limits how far Hyperliquid can move. - Chart: Weekend oil volume on Hyperliquid (https://castlelabs.io/research/reports/the-evolution-of-commodities#chart-oil-weekend-volume, report page 64): Weekend volume on xyz:CL grew from $31.3 million to $1,072.1 million in three weekends, and the number of trades rose from 26,140 to 706,634. - Q: How big is the tokenised commodities market? A: Spot tokenised commodities were worth $7.4 billion in March 2026, over 27% of the $27.35 billion tokenised real-world asset market, up from $1.9 billion a year earlier. Gold accounts for over $5.9 billion of it. On top of spot tokens, commodity perps on Hyperliquid carried over $1.4 billion of open interest. - Q: What is the largest tokenised commodity? A: Gold, through Tether Gold (XAUT) and Paxos Gold (PAXG), with over $5.9 billion onchain and more than 75% of the spot commodity market. Tokenised gold traded over $178 billion in 2025, up 227% year on year, second only to the GLD ETF globally. - Q: Can you trade oil on weekends? A: Yes, onchain. Hyperliquid's HIP-3 oil perps, mostly through TradeXYZ, trade 24/7 while CME is shut from Friday evening to Sunday evening. On the weekend of 7 to 10 March 2026, after the Iran strikes, just under $3 billion of commodities traded on Hyperliquid, and oil perps became its largest HIP-3 market. - Q: How does Hyperliquid price commodities when CME is closed? A: TradeXYZ sends three prices per update: a Pyth oracle reference, a mark price limited to a 1% move per update, and an external safety price that falls back to an average of recent mark prices when Pyth stops updating. Discovery bounds cap how far the price can move from the last external reference, funding rates push it back toward the index, and the price snaps back when Pyth resumes on Monday. - Q: Is onchain oil as liquid as CME? A: Not yet. In our March 2026 study, CME offered $19 million of executed depth within 2 bps against $152,000 on Hyperliquid, and a $1 million order slipped 15.4 bps on Hyperliquid against 0.79 bps on CME. For small orders the gap barely matters: a $10,000 order slipped 0.77 bps on Hyperliquid, and the median Hyperliquid oil trade was $543. - Q: How much commodity volume does Hyperliquid handle? A: HIP-3 markets had processed over $109 billion by March 2026, about 25% of Hyperliquid's volume, with commodities over 16% of the platform total. On 5 February 2026 silver perps alone did $4.09 billion in 24 hours. TradeXYZ handles over 85% of HIP-3 volume. - Q: Will onchain venues replace the CME? A: Our view is no. Electronic trading took CME from pits to Globex over 29 years and made it bigger, not smaller. The likely outcome is that CME adopts 24/7 trading and onchain settlement while onchain venues win the users CME cannot serve: small retail trades, non-US access and DeFi composability. Onchain oil vs CME oil during overlapping hours (source: CME CLJ6 trades via Databento; Hyperliquid xyz:CL fills from Hyperliquid's public S3 archive) | Metric | Hyperliquid xyz:CL | CME CLJ6 | | --- | --- | --- | | Executed depth within 2 bps | $152K | $19M (125x) | | Median trade size | $543 | $90,450 | | Slippage on a $10K order | 0.77 bps | 0 bps | | Slippage on a $100K order | 4.33 bps | 0 bps | | Slippage on a $1M order | 15.4 bps | 0.79 bps | CME commodity volume, February 2026 (source: CME Group monthly volume, February 2026) | Segment | Average daily contracts | Change year on year | | --- | --- | --- | | All products | 37.6M (record) | +14% | | Energy | 3.2M | +12% | | Crude oil | 1.1M | Over +35% | | Metals | 1.5M | Over +88% | | Agriculture | 2.3M | +11% | - [The Vaultification of Finance](https://castlelabs.io/research/reports/vaultification-of-finance): The evolution of Vaults as the infrastructure layer of institutional finance onchain (May 2026, 76 pages; download: https://docsend.com/view/r3278hmn8d6tqyqn) - Finding: Net TVL across the eight defined vault categories was $120.4 billion, down roughly 50% from its October peak. - Finding: RWA vaults grew 37.8% during the report's stress window while every other category declined. - Finding: Three curators controlled 75% of the $6.5 billion risk-curated vault category, showing how concentrated onchain allocation remained. - What Is a DeFi Vault? (https://castlelabs.io/research/reports/vaultification-of-finance#what-is-a-defi-vault): A DeFi vault is a smart contract that pools deposits and runs a yield strategy on the depositors' behalf: lending, staking, delta-neutral trades, real-world credit or options. Depositors receive a share token whose value tracks the vault's assets. Since ERC-4626 went live in March 2022, almost every yield-bearing vault uses the same 16-method interface, and ERC-7540 (October 2023) added asynchronous deposits and redemptions for assets that settle offchain, such as tokenised credit. We count a product as a vault when users access an active yield-generating strategy through it. Maple's syrupUSDC, Lido's stETH and Centrifuge's JAAA qualify; BlackRock's BUIDL, a 1:1 claim on an offchain Treasury fund, does not. On that definition, eight vault categories held $120.4 billion of net TVL in May 2026, down about 50% from a $241 billion peak in October 2025. The eight vault categories, May 2026 (source: Castle Labs, The Vaultification of Finance (May 2026), from DeFiLlama and protocol data) | Category | How it earns | Largest examples | | --- | --- | --- | | Lending | Interest from borrowers | Aave, Morpho ($7.5B+), Spark, Fluid, Kamino ($1.6B+) | | Liquid staking | Proof-of-stake rewards ($42.4B category) | Lido ($21.8B), Binance Staked ETH ($8.9B), Rocket Pool ($1.2B) | | Restaking | Extra rewards for securing other services ($20.6B category) | EigenCloud ($7.8B), EtherFi ($5.7B), Kelp DAO ($1.6B) | | Risk-curated | Curator-managed lending portfolios (~$6.5B) | Sentora ($1.85B), Steakhouse ($1.63B), Gauntlet ($1.5B) | | Infrastructure and yield optimisers | Multi-strategy allocation across DeFi | Veda ($1B), Upshift ($380M), Fluid Lite ($164M) | | RWA credit | Offchain credit and Treasury yield | Maple ($2.1B), Centrifuge ($1.6B), Anemoy ($1.1B), Re ($263M) | | Perpetual LP | Taking the other side of perp traders | Jupiter JLP ($715M), Hyperliquid HLP ($396M), Drift, GMX, Ostium | | Options | Selling options premium via RFQ | Rysk (~$56M), Derive V1 vaults (~$2.4M) | - What Is a Vault Curator? (https://castlelabs.io/research/reports/vaultification-of-finance#what-is-a-vault-curator): A curator is the risk manager who decides where a vault's deposits go. Morpho made the model mainstream in 2024: Morpho supplies isolated lending markets and the vault blueprint, and curators choose the markets, set supply caps and rebalance. Morpho Vaults V2 splits the job into four roles (owner, curator, allocator and sentinel) and adds gating such as KYC, minimum balances and wallet-age checks. Curation is concentrated. Three curators held 75% of the ~$6.5 billion risk-curated category, and fees are far below TradFi: Steakhouse, the top curator by revenue, earned about $3 million a year on $2.13 billion of AUM, roughly 0.14%. Most curator revenue is performance-based, which rewards yield and can push allocations toward riskier markets. MEV Capital and Re7 peaked at $1.49 billion and $830 million before the Stream Finance collapse in November 2025. After the KelpDAO hack, KPK (+159.6%) and Gauntlet (+42.7%) still took net inflows. Largest risk curators, May 2026 (source: Castle Labs, The Vaultification of Finance (May 2026)) | Curator | TVL | | --- | --- | | Sentora | $1.85B | | Steakhouse | $1.63B | | Gauntlet | $1.5B | - Morpho vs Aave: Curated vs Pooled Lending (https://castlelabs.io/research/reports/vaultification-of-finance#morpho-vs-aave): Aave runs pooled markets where assets share risk; Morpho runs isolated markets aggregated into curated vaults, so a problem in one market stays in that market. In the words of Morpho's co-founder, "Aave is a bank, whereas Morpho is an infrastructure for banks." Aave V4 moves toward modularity with spokes on a shared liquidity hub, but its governance still approves each spoke, while Morpho is permissionless. The KelpDAO rsETH hack in April 2026 showed the difference. Aave lost over $12 billion of TVL and the Aave-to-Morpho TVL ratio fell from 5 to 6 times to below 2. Morpho reached over $7.5 billion of TVL, with Base growing from $604 million to over $2.8 billion, and about 40% of its USD TVL in Coinbase's cbBTC. - Chart: Vault TVL with key events, 2020 to 2026 (https://castlelabs.io/research/reports/vaultification-of-finance#chart-vault-tvl-with-events, report page 11): Vault growth came in steps, each tied to an event: Yearn's launch, options vaults, the ETH Merge, FTX, Morpho Vaults, the HYPE airdrop, the October 2025 liquidation event, the Stream Finance blowup and the KelpDAO hack. - Chart: Vault TVL by category (https://castlelabs.io/research/reports/vaultification-of-finance#chart-vault-tvl-by-category, report page 33): Net TVL across the eight vault categories is $120.4 billion, down about 50% from a $241 billion peak in October 2025. The October liquidation event set off the decline. - Chart: Vault TVL change by category since October 2025 (https://castlelabs.io/research/reports/vaultification-of-finance#chart-vault-tvl-change-since-october-2025, report page 34): Through Stream Finance, Resolv and the Kelp hack, RWA vaults gained 37.8% while every crypto-native category drew down. Their non-crypto exposure kept them uncorrelated. - Chart: Morpho TVL by asset (https://castlelabs.io/research/reports/vaultification-of-finance#chart-morpho-tvl-distribution, report page 36): Aave is answering on institutional rails with Horizon, which has gathered over $340 million since launch. Spark, part of the Sky ecosystem, was among the biggest winners of inflows after the rsETH hack. - Chart: EigenCloud TVL in ETH (https://castlelabs.io/research/reports/vaultification-of-finance#chart-eigencloud-tvl-in-eth, report page 40): EigenCloud ($7.8 billion) leads restaking ahead of EtherFi ($5.7 billion) and Kelp DAO ($1.6 billion). Its TVL held flat in ETH terms for a long time, then fell after the Kelp hack as users withdrew. - Chart: Risk curator TVL (https://castlelabs.io/research/reports/vaultification-of-finance#chart-risk-curators-tvl, report page 42): Risk-curated vaults hold about $6.5 billion. Sentora ($1.85 billion), Steakhouse ($1.63 billion) and Gauntlet ($1.5 billion) manage 75% of it. - Chart: Risk curator market share (https://castlelabs.io/research/reports/vaultification-of-finance#chart-risk-curators-market-share, report page 43): Steakhouse expects concentration to follow the power laws of traditional asset management, where most assets sit with the leading managers, while the open arena keeps competition alive at the margin. - Chart: Curator TVL growth after the KelpDAO hack (https://castlelabs.io/research/reports/vaultification-of-finance#chart-curator-flows-after-kelpdao, report page 44): Most curators lost deposits after the hack, but KPK (+159.6%) and Gauntlet (+42.7%) took net inflows: KPK from its new Morpho V2 vault, Gauntlet from its BSC expansion. - Chart: Maple Finance active loans by token (https://castlelabs.io/research/reports/vaultification-of-finance#chart-maple-active-loans, report page 48): RWA vaults grew at a 231.3% compound annual rate over five years. Maple, the largest at $2.1 billion, grew almost 10x from the start of 2025 after Syrup opened its loans to retail through syrupUSDC and syrupUSDT. - Chart: RWA TVL actively used in DeFi (https://castlelabs.io/research/reports/vaultification-of-finance#chart-rwa-defi-active-tvl, report page 50): Looping a 4% to 5% Treasury yield two to three times over can return 7% to 12%, but only while borrowing costs stay near 1%. Volatile stablecoin rates and slow redemptions limit how far RWAs scale in DeFi. - Chart: Hyperliquidity Provider (HLP) vault TVL (https://castlelabs.io/research/reports/vaultification-of-finance#chart-hlp-vault-tvl, report page 53): Perp LP vaults are led by Jupiter's JLP ($715 million), Hyperliquid's HLP ($396 million), Drift, GMX and Ostium. HLP is down 30% from its $600 million peak, and deposits follow its variable yield. - Chart: Ostium Liquidity Provider (OLP) vault TVL (https://castlelabs.io/research/reports/vaultification-of-finance#chart-ostium-olp-tvl, report page 54): OLP is down about 50% from its peak. Ostium has since made OLP a senior tranche that never takes first loss, the opposite of HLP's design. - Chart: DeFi TVL vs DeFi insurance TVL (https://castlelabs.io/research/reports/vaultification-of-finance#chart-defi-insurance-coverage, report page 69): Insurance covers a sliver of DeFi: insurance TVL is 1.7% of DeFi TVL in Three Sigma's data. Vaults still lack a shared risk standard like the one ERC-4626 created for their interface. - Q: How much is locked in DeFi vaults? A: $120.4 billion of net TVL across eight vault categories in May 2026, down about 50% from a $241 billion peak in October 2025. The figure is higher than DeFi's total TVL of about $86 billion because vaults stack: stETH from a liquid staking vault is reused as collateral in lending vaults. - Q: Which vault category held up best in the 2025-2026 sell-off? A: RWA vaults. From October 2025 to May 2026, through the Stream Finance, Resolv and KelpDAO incidents, RWA vaults grew 37.8% while every other category fell, because their yield comes from offchain credit and Treasuries rather than crypto collateral. RWA vaults have grown at a 231.3% compound annual rate over five years. - Q: What is ERC-4626? A: ERC-4626 is the tokenised vault standard, proposed in December 2021 and live since March 2022. It gives every vault the same share-token interface (deposit, mint, withdraw, redeem and conversion functions), so protocols integrate one standard instead of writing an adapter per vault. ERC-7540 extends it with request-then-claim settlement for assets that cannot settle in one transaction. - Q: Who are the largest vault curators? A: Sentora ($1.85 billion), Steakhouse ($1.63 billion) and Gauntlet ($1.5 billion) held 75% of the roughly $6.5 billion risk-curated vault category in May 2026. Curators charge far less than TradFi funds: Steakhouse earned about $3 million a year on $2.13 billion of AUM, around 0.14%. - Q: How do institutions use DeFi vaults? A: Mostly through vault infrastructure rather than directly. Bitwise became a Morpho curator in January 2026; Kraken DeFi Earn, powered by Veda, took $200 million of deposits in two weeks; Coinbase uses Morpho for over $1 billion of BTC-backed loans; and Aave's institutional market Horizon has drawn over $300 million. Institutions want isolated risk, independent verification and compliance built into the vault. - Q: What are the main risks of DeFi vaults? A: Seven recur: smart contract risk across the vault, base protocol and markets; curator risk; rehypothecation and contagion; counterparty risk; oracle failures; liquidity risk when markets hit 100% utilisation; and liquidation risk from oracle lag or cascades. Stream Finance's xUSD showed several at once: its oracle held a $1 price while the token traded 70-80% lower, blocking liquidations and leaving bad debt. - Q: How does RWA looping work? A: A looper borrows stablecoins against tokenised Treasuries and redeposits them into the yield vault several times. A 4-5% Treasury yield looped 2-3 times can return 7-12%, but only while borrowing stays cheap, around 1%. The catch is exit time: with a 5x loop on an asset that redeems in two days, fully unwinding can take up to 10 days. Major vault incidents and what they exposed (source: Castle Labs, The Vaultification of Finance (May 2026)) | Incident | Date | What happened | Risk it exposed | | --- | --- | --- | --- | | Venus donation attack (zkSync) | March 2025 | About $700K of wUSDM exploited by inflating an ERC-4626 share price | Smart contract | | JELLY on Hyperliquid HLP | March 2025 | Manipulated memecoin position forced onto HLP; manual delisting avoided a loss of more than $13M | Counterparty | | Stream Finance xUSD | November 2025 | $93M loss; oracle kept xUSD at $1 while it traded 70-80% lower, blocking liquidations | Oracle and liquidity | | Aave wstETH oracle | March 2026 | A 3.3% oracle gap liquidated 34 healthy accounts for $27.78M in one block | Liquidation | | Resolv USR | Between Oct 2025 and May 2026 | About 80M unbacked USR minted; bad debt hit 15 Morpho vaults (~$10M) and Fluid (~$21M) | Oracle | | KelpDAO rsETH | April 2026 | Contagion through Aave looping; Aave lost over $12B of TVL | Rehypothecation and contagion | - [The Full-Stack Privacy Ecosystem](https://castlelabs.io/research/reports/full-stack-privacy): Privacy as Infrastructure for User Protection and Institutional Adoption (June 2026, 64 pages; download: https://docsend.com/v/sjv2g/thefullstackprivacyecosystem) - Finding: About 4.9 million ZEC, slightly less than 30% of supply, was held in shielded pools at publication. - Finding: Railgun had processed more than $5 billion in shielded volume, while Arcium had processed more than 3.5 million transactions after its alpha launch. - Finding: Selective disclosure offers a practical route to confidential onchain activity without removing auditability where proof is legally or operationally required. - ZK vs FHE vs MPC vs TEE: Which Privacy Technology Does What? (https://castlelabs.io/research/reports/full-stack-privacy#zk-vs-fhe-vs-mpc-vs-tee): Onchain privacy rests on four main technologies, and most serious protocols combine several of them. ZK proofs show that a statement is true without revealing the data behind it. Fully homomorphic encryption (FHE) lets a network compute on encrypted data without decrypting it. Multi-party computation (MPC) lets parties who do not trust each other compute a shared result without seeing each other's inputs. Trusted execution environments (TEEs) run code inside a hardware enclave that even the machine's operator cannot read. The trade-off is between guarantees and speed. ZK and FHE rely on cryptography; TEEs rely on the hardware maker not being compromised, and MPC relies on participants being online and not colluding, which is hard to detect. Zama shows how the pieces combine: FHE for computing on encrypted data, MPC to decentralise its network, and ZK only to verify that encrypted user inputs are valid. The four privacy technologies compared (source: Castle Labs, The Full-Stack Privacy Ecosystem (June 2026)) | Technology | What it does | Main limitation | Who uses it | | --- | --- | --- | --- | | ZK proofs | Prove a computation or statement is correct without revealing inputs | Proving is expensive, especially client-side | Zcash, Railgun, Privacy Pools, Aztec, Starknet STRK20 | | FHE | Compute directly on encrypted data | Slow and costly, though new schemes are closing the gap | Zama, Fhenix, Octra | | MPC | Split keys or data across nodes so no single node sees everything | Needs participants online; latency grows with participants; collusion is hard to detect | Arcium, Near Chain Signatures, Nillion | | TEE | Execute on plaintext inside an isolated hardware enclave | Trust in the hardware vendor (for example Intel SGX keys) | Secret Network, Oasis, MagicBlock, Near Confidential Intents | - Monero vs Zcash: How the Two Privacy Coins Differ (https://castlelabs.io/research/reports/full-stack-privacy#monero-vs-zcash): Monero is private by default. It hides senders with ring signatures, receivers with one-time stealth addresses, and amounts with Ring Confidential Transactions. Zcash is private by choice: users opt in to shielded z-addresses that use zk-SNARKs, which allows selective disclosure but leaves a smaller anonymity set. Monero's share of the privacy coin market hit an all-time high of over 80% in September 2025 and has since fallen to about 40%, largely because of the rise of Zcash. In late May 2026 a researcher found a bug that could have let anyone inside Zcash's Orchard pool mint counterfeit ZEC. It was found on 29 May and fixed by 1 June, but because the pool is private there was no way to prove cryptographically whether it had been exploited, and ZEC fell about 50%. The Ironwood upgrade, expected in late July 2026, moves funds to a new pool through a turnstile so anyone can verify total supply. Monero and Zcash side by side, June 2026 (source: Castle Labs, The Full-Stack Privacy Ecosystem (June 2026)) | Metric | Monero (XMR) | Zcash (ZEC) | | --- | --- | --- | | Privacy model | On by default for every transaction | Opt-in shielded pools | | Core tech | Ring signatures, stealth addresses, RingCT | zk-SNARKs | | Adoption of privacy | All transactions | About 4.9M ZEC shielded, slightly under 30% of supply | | Main weakness | Probabilistic privacy that AI-assisted chain analysis could erode | Small shielded set can be de-anonymised through entry and exit timing | - The Solana Privacy Ecosystem in 2026 (https://castlelabs.io/research/reports/full-stack-privacy#solana-privacy-ecosystem): Solana's privacy stack is young but moving fast. Helius CEO Mert told us Solana is lagging on privacy, and that a mature ecosystem needs at least two very large privacy pools or one clearly decentralised pool, with formal verification, no committees, immutability and open-source code as non-negotiables. Two compute layers sit underneath most apps: Arcium, which uses MPC, and MagicBlock, which uses Intel TDX enclaves in Private Ephemeral Rollups. Arcium's alpha mainnet launched in early February 2026 and has processed over 900,000 computations and 3.5 million transactions. Solana privacy projects by category (source: Castle Labs, The Full-Stack Privacy Ecosystem (June 2026)) | Category | Projects | How they work | | --- | --- | --- | | Private compute | Arcium, MagicBlock | MPC clusters (Arcium) or TEE-based Private Ephemeral Rollups (MagicBlock), settling on Solana | | Private transfers and balances | Umbra, Privacy Cash, Hush | Encrypted token accounts on Arcium (Umbra); Tornado-style SOL pool (Privacy Cash); shielded jitoSOL pool earning staking yield (Hush) | | Private trading | encrypt.trade, Vanish, Darklake | Encrypted swaps routed through Jupiter inside AWS Nitro enclaves; shielded routing; ZK blind-slippage AMM and zk-Perps | | Private prediction markets | Melee | Order book encrypted with Arcium MPC until settlement | | Private AI | Loyal | Encrypted conversations and queries stored on Solana, using MagicBlock and Arcium | | Coming | Helius Privacy | ZK-based UTXO privacy layer with configurable Zones and a public zone | - Chart: Zcash shielded pool supply (https://castlelabs.io/research/reports/full-stack-privacy#chart-zcash-shielded-pool-supply, report page 20): A privacy set is only as strong as its adoption. With a small shielded set, the timing of entries and exits and links to transparent addresses can still de-anonymise users. - Chart: How assets move through Railgun (https://castlelabs.io/research/reports/full-stack-privacy#chart-railgun-asset-flow, report page 25): Once tokens are shielded in Railgun's contract they look like every other shielded token, without mixing. Users can send, swap and use DeFi privately, then withdraw to a public address. - Chart: Railgun shielded volume by month (https://castlelabs.io/research/reports/full-stack-privacy#chart-railgun-monthly-shielded-volume, report page 26): Every Railgun transfer, swap and loan adds noise to the privacy set. Railgun has processed over $5 billion in volume, 15,000 to 20,000 transfers a month recently, and over $12.4 million in fees. - Chart: Public vs private data in Aztec (https://castlelabs.io/research/reports/full-stack-privacy#chart-aztec-public-vs-private-data, report page 31): Aztec, live in alpha since March 2026, runs private functions on the user's device and proves them onchain. A nullifier proves a note exists without revealing which one. - Chart: How EVM apps use Secret Network (https://castlelabs.io/research/reports/full-stack-privacy#chart-secret-network-interactions, report page 35): Secret has run private smart contracts since 2020. An EVM app requests data, a Secret contract computes it privately inside a hardware enclave, and sends the result back. - Chart: Near Confidential Intents TVL (https://castlelabs.io/research/reports/full-stack-privacy#chart-near-confidential-intents-tvl, report page 39): Confidential Intents run on a private Near shard secured by permissioned validators. Token pairs, order sizes and counterparties stay hidden; only settlements are public. - Q: What is full-stack privacy in crypto? A: The idea that privacy has to be solved at every layer where users leak data, not just in the token. Public transactions reveal what you trade, RPC nodes see every query and your IP, network timing can de-anonymise encrypted traffic, and wallet interfaces can fingerprint devices across fresh addresses. Full-stack privacy combines L1s, Ethereum privacy layers, L2s, app-layer compute and wallet tooling. - Q: What is the difference between ZK, FHE, MPC and TEE? A: ZK proofs prove something is true without revealing the data. FHE computes on encrypted data without decrypting it. MPC splits a key or computation across parties so no single party sees the inputs. TEEs run code on plaintext inside a hardware enclave, so they depend on trusting the chip maker. Most production systems combine two or more. - Q: Is Monero or Zcash more private? A: Monero is private by default for every transaction, while Zcash privacy is opt-in, with about 4.9 million ZEC, slightly under 30% of supply, in shielded pools. A smaller shielded set is easier to de-anonymise through timing. Monero's guarantees are probabilistic, though, and could be weakened by AI-assisted chain analysis. - Q: How much of Zcash's supply is shielded? A: About 4.9 million ZEC, slightly under 30% of supply, sat in Zcash's shielded pools when Castle's June 2026 privacy report was written. Privacy on Zcash is opt-in, so the rest moves through transparent addresses, and a smaller shielded set is easier to de-anonymise through the timing of deposits and withdrawals. - Q: What happened with the Zcash Orchard bug in 2026? A: A researcher found a bug that could have let anyone inside the shielded Orchard pool mint counterfeit ZEC. It was found on 29 May 2026 and fixed by 1 June, but the pool's privacy meant there was no way to prove whether it had been exploited, and ZEC fell about 50%. The Ironwood upgrade forces funds out of the old pool through a turnstile so the supply can be verified. - Q: How does Railgun stay compliant? A: Through Proof of Innocence: when tokens are shielded, a ZK proof is created showing they are not linked to a preset list of illicit wallets and interactions, without revealing user data. The contract can be upgraded by the RAIL DAO but only with a 7-day delay, giving users time to exit. Railgun has processed over $5 billion of volume and holds over $80 million of TVL. - Q: Can onchain privacy be compliant? A: Yes, through selective disclosure. Railgun's Proof of Innocence, Privacy Pools' Association Set Provider, Umbra's auditor viewing keys and Starknet STRK20's threshold-controlled integrity council all hide data from the public while letting users prove specific facts to auditors or regulators. We think selective disclosure, not total anonymity, is the path to adoption. - Q: What privacy projects are building on Solana? A: Arcium (MPC compute) and MagicBlock (TEE rollups) at the base; Umbra, Privacy Cash and Hush for private transfers; encrypt.trade, Vanish and Darklake for private trading; Melee for private prediction markets; Loyal for private AI; and Helius Privacy, a ZK-based UTXO privacy layer, coming next. Privacy protocols by the numbers, June 2026 (source: Castle Labs, The Full-Stack Privacy Ecosystem (June 2026)) | Protocol | Type | Key figures | | --- | --- | --- | | Railgun | Privacy layer for Ethereum DeFi | Over $5B transaction volume; over $80M TVL; over $12.4M total fees; 15-20K transfers a month | | Privacy Pools | Compliant mixer on Ethereum | About $8M TVL (over $4.4M USDT, $1.5M ETH, $1.5M USDC) | | Tornado Cash | Immutable mixer on Ethereum | Over $410M TVL despite US sanctions | | Zcash | Privacy L1 | About 4.9M ZEC shielded, slightly under 30% of supply | | Aztec | Privacy L2 | Alpha mainnet March 2026; over 23 active prover nodes | | Arcium | MPC compute on Solana | Over 900K computations and 3.5M transactions since February 2026 | | Zama | FHE infrastructure | About 20 TPS today, targeting over 1,000 TPS with dedicated hardware | - [Real-World Assets: Bringing TradFi Onchain](https://castlelabs.io/research/reports/real-world-assets): Mapping the Growth of RWAs and Institutional Adoption (June 2026, 85 pages; download: https://docsend.com/v/sjv2g/rwareport) - Finding: Tokenised RWAs stood at $28.2 billion, up from $4.3 billion in January 2025, with a peak of $31.8 billion. - Finding: Only about $3 billion, or 10% of the category, was active in DeFi because of access controls, KYC requirements and redemption constraints. - Finding: Public debt represented more than $17 billion, while private credit reached $4.23 billion with a 52% DeFi utilisation rate. - What Are Real-World Assets (RWAs) in Crypto? (https://castlelabs.io/research/reports/real-world-assets#what-are-rwas): Real-world assets are tokens that represent a claim on something that exists offchain: Treasury bills, gold, private loans, stocks, reinsurance, private company shares or property. Tokenisation adds fractional ownership, 24/7 transfer and the option to use the asset in DeFi. Tokenised RWAs are worth $28.2 billion, with a peak of $31.8 billion, up from $4.3 billion in January 2025. Growth came from institutional issuers (BlackRock, Franklin Templeton, Apollo), stablecoins needing low-risk yield, and regulation such as the GENIUS Act and MiCA. Public equities grew over 170% year to date against about 60% for public debt, driven by xStocks and Ondo. The seven RWA categories, June 2026 (source: RedStone data in Castle Labs, Real-World Assets: Bringing TradFi Onchain (June 2026)) | Category | Onchain value | DeFi use | Typical yield | Leading products | | --- | --- | --- | --- | --- | | Public debt (Treasuries, money market funds) | About $17B, nearly 60% of the market | About 6% used in DeFi | 3.3-3.5% after fees | BlackRock BUIDL, Circle USYC, Ondo USDY | | Commodities | $5.34B, almost all gold | 5.3% | Price only | Tether Gold (XAUT), Paxos Gold (PAXG) | | Private credit | $4.23B across 18 major products | $2.21B, about 52% | 4.5-5% | Maple syrupUSDC and syrupUSDT, Hastra PRIME | | Public equities | Over $2B across about 500 products | $103M | Price only | xStocks, Ondo Global Markets | | Reinsurance | $370M | $310M, the highest share of any category | 7-15% target | OnRe ONyc, Re Protocol reUSD | | Private equity and venture | $181M across 14 products | $24M | Depends on fund returns | BOSS (Boss Info AG), Stobox STBX | | Real estate | $96M | Minimal | About 4% in most commercial cities | Reental, TRIZE | - Why Most RWAs Are Not Used in DeFi (https://castlelabs.io/research/reports/real-world-assets#why-rwas-arent-used-in-defi): Only about $3 billion of the $28 billion RWA market, roughly 10%, is deployed in DeFi. Two problems explain most of the gap. First, many tokens carry KYC and transfer restrictions, so standard lending pools and liquidation bots cannot hold them. Second, redemption takes T+1 to T+3 or longer, while a lending protocol needs to liquidate collateral in seconds. That matters most for looping, where users borrow stablecoins against a yield-bearing RWA and buy more of it. At a 70% LTV a user can loop up to about 3x exposure on their capital, but if rates spike, unwinding a 3x looped position in a T+1 asset can take 2 to 4 days while interest keeps accruing. A new layer of redemption providers now takes on that duration risk for a fee. Instant RWA redemption providers (source: Castle Labs, Real-World Assets: Bringing TradFi Onchain (June 2026)) | Provider | Launched | How it works | Assets and cost | | --- | --- | --- | --- | | RedStone Settle | April 2026 | Auction among KYC-verified solvers; the winner fronts USDC at T+0 and is repaid when the issuer redeems at T+1 or T+2 | Market-set discount | | Upshift Clear | May 2026 | Dedicated USDC vault provides T+0 liquidity; lenders earn the fee plus yield during redemption | 50 bps fee; started with Superstate USCC | | Grove Basin | May 2026 | Institutional redemption with BlackRock and Janus Henderson, accessed through Anchorage Digital, Galaxy and FalconX | Up to $1B a day; BUIDL and JTRSY | | Agra | Live | Order book where makers bid below NAV for tokenised credit | wmtUSDC, ACRDX and others | | 3F | Live | Bridge facilitators and liquidity integrators fund looped entries and instant exits | $7.7M net looped exposure | | Fission | Live | Instant redemption with thin liquidity | 564 bps fee; ACRED, ACRDX, JAAA | - What Legal Rights Do RWA Tokens Give You? (https://castlelabs.io/research/reports/real-world-assets#rwa-legal-rights): A tokenised asset is not a legal category of its own. It is an onchain record of a claim that exists elsewhere, and two tokens tracking the same asset class can give holders very different rights. Before holding one, ask who the claim is against, what the official ownership record is, whether you can redeem directly and for what, who may hold or transfer it, and what happens if the issuer, fund, custodian or borrower fails. What you actually own, by structure (source: Castle Labs, Real-World Assets: Bringing TradFi Onchain (June 2026), from issuer documentation) | Structure | Examples | Your claim | | --- | --- | --- | | Tokenised fund | Franklin BENJI, BlackRock BUIDL, Ondo OUSG | A share or interest in the fund, not the Treasuries inside it; redemption and transfers follow the fund documents and eligibility rules | | Tokenised note | Ondo USDY | A debt claim against the note issuer, backed by short-term Treasuries and bank deposits | | Equity tracker | xStocks | A tracker certificate from Backed Assets (JE); no voting rights and no claim against the company whose stock it tracks | | Commodity-backed | Paxos Gold (PAXG), Tether Gold (XAUT) | Ownership of, or an undivided interest in, vaulted gold; physical redemption depends on issuer terms, and XAUT redeems in full bars | | Lending pool | Maple syrupUSDC | Credit exposure; yield comes from borrowers repaying, and defaults flow through to lenders | - Chart: Tokenised RWA value (https://castlelabs.io/research/reports/real-world-assets#chart-tokenised-rwa-value, report page 5): Tokenised RWAs are worth $28.2 billion, after a peak of $31.8 billion, up from $4.3 billion of active onchain value in January 2025. - Chart: Tokenised RWA value vs stablecoin market cap and DeFi TVL (https://castlelabs.io/research/reports/real-world-assets#chart-rwa-vs-stablecoins-and-defi, report page 8): RWAs did most of their growing in the last two years and outpaced both DeFi and stablecoins, as BlackRock, Franklin Templeton and Apollo moved capital onchain for atomic settlement. - Chart: Fixed income and Treasury wrappers by tokenised value (https://castlelabs.io/research/reports/real-world-assets#chart-fixed-income-wrappers, report page 29): BlackRock's BUIDL, launched through Securitize in March 2024, and Hashnote's USYC anchor the category. BUIDL is worth $3 billion, plus $829 million in its institutional class, BUIDL-I. - Chart: Tokenised equity wrappers by value (https://castlelabs.io/research/reports/real-world-assets#chart-public-equity-wrappers, report page 35): Tokenised equities are worth about $2 billion, led by Ondo ($940 million) and xStocks ($510 million). The two give different claims: xStocks' Jersey SPV holds real shares with Alpaca Securities, while Ondo issues structured debt notes through a BVI entity. - Chart: xStocks tokenised value by stock (https://castlelabs.io/research/reports/real-world-assets#chart-xstocks-distribution, report page 36): Strategy, Tesla, Circle and Nvidia are xStocks' largest products. Backpack launched in June 2026 with SpaceX, which reached $10.8 million with over 40% of it used in DeFi. - Chart: Maple Finance market size (https://castlelabs.io/research/reports/real-world-assets#chart-maple-market-size, report page 41): Maple grew by becoming collateral on Aave and Morpho and feeding loops. syrupUSDT is 86% utilised, mainly through Aave on Plasma, and syrupUSDC 26.88% across several lenders. - Chart: Wildcat Finance market size (https://castlelabs.io/research/reports/real-world-assets#chart-wildcat-market-size, report page 42): Wildcat runs $185 million of undercollateralised credit lines for borrowers such as Keyrock, Auros and Wintermute. Wintermute's wmtUSDC is now Morpho collateral, a first for undercollateralised private credit. - Chart: Midas tokenised value by product (https://castlelabs.io/research/reports/real-world-assets#chart-midas-tokenised-value, report page 43): Midas runs six products across four strategies. Private credit (mF-ONE, $70.2 million) is the largest, ahead of T-bills (mTBILL, $52 million) and funding-rate harvesting (mHYPER, $37 million). - Chart: Tokenisation platforms by tokenised value (https://castlelabs.io/research/reports/real-world-assets#chart-tokenisation-platforms, report page 45): Securitize is the largest US-regulated tokenisation platform by AUM, with $4.4 billion tokenised on its rails, including BlackRock BUIDL, Apollo ACRED and Hamilton Lane HLSCOPE. - Chart: RWA TVL used in DeFi, by category (https://castlelabs.io/research/reports/real-world-assets#chart-defi-active-tvl-by-rwa-category, report page 51): Private credit is the RWA category DeFi uses most, ahead of fixed income and reinsurance. Morpho, Maple, Aave, Kamino and Fluid carry most of that usage. - Chart: RWA TVL actively used in DeFi (https://castlelabs.io/research/reports/real-world-assets#chart-rwa-defi-active-tvl, report page 52): About $3 billion of RWAs are used in DeFi, roughly 10% of the market. KYC rules and slow redemption keep the rest out of lending pools. - Chart: Pendle RWA markets: TVL vs fixed APY (https://castlelabs.io/research/reports/real-world-assets#chart-pendle-rwa-markets, report page 54): Pendle holds $956 million across 11 chains and is DeFi's main fixed-rate venue. Stablecoins lead its RWA markets, with USDat and apxUSD near the top, reUSD adding reinsurance and USDai AI-compute credit. - Chart: Morpho RWA deposits (https://castlelabs.io/research/reports/real-world-assets#chart-morpho-rwa-deposits, report page 57): Most of Morpho's recent RWA growth came from syrupUSDC, PRIME and AA_FalconXUSDC. Isolated markets keep bad debt in one RWA from spreading to other depositors. - Chart: Aave Horizon market by asset (https://castlelabs.io/research/reports/real-world-assets#chart-aave-horizon-distribution, report page 59): Aave Horizon, launched in August 2025, is about $500 million and accepts USCC, USTB, VBILL and JAAA as collateral. RLUSD, boosted by rewards, dominates supply. - Q: How big is the tokenised RWA market? A: $28.2 billion in June 2026, after a peak of $31.8 billion, up from $4.3 billion in January 2025. Public debt is the largest category at about $17 billion, nearly 60% of the market, followed by commodities at $5.34 billion and private credit at $4.23 billion. - Q: What is the largest tokenised Treasury fund? A: Circle's USYC at $3.07 billion and BlackRock's BUIDL at $3 billion, plus $829 million in its institutional class BUIDL-I. BUIDL, launched in March 2024 through Securitize, is also a reserve asset for Ondo OUSG and Ethena USDtb. - Q: Why are RWAs not used more in DeFi? A: Only about $3 billion, roughly 10%, of tokenised RWAs is deployed in DeFi. Many tokens are restricted to KYC'd or whitelisted holders, which standard lending pools and liquidation bots cannot handle, and redemption takes T+1 to T+3 or longer while liquidations need instant exits. - Q: Which RWA category is used most in DeFi? A: Private credit in absolute terms, with $2.21 billion deployed out of $4.23 billion. Relative to its size, reinsurance leads: $310 million of its $370 million is used in DeFi. Public debt, despite being the largest category, has only about 6% in DeFi. - Q: Which lending protocol has the most RWA collateral? A: Morpho, with over $1 billion of RWA deposits and $400 million of active loans. Its isolated markets keep each RWA's risk contained, and curators can set custom collateral ratios and compliance checks. Aave Horizon has about $500 million, Kamino $400 million and Fluid $270 million. - Q: Do tokenised stocks give you shareholder rights? A: Usually not. xStocks, for example, are tracker certificates issued by Backed Assets (JE): holders get price exposure but no voting rights and no claim against the underlying company. Your claim is against the issuer under the certificate terms. - Q: What is instant RWA redemption? A: A service where a third party pays out USDC immediately and waits for the issuer's normal T+1 or T+2 redemption, earning a spread for taking that duration risk. RedStone Settle (April 2026), Upshift Clear (May 2026, 50 bps fee) and Grove Basin (May 2026, up to $1 billion a day) are the main examples. The largest tokenised Treasury and money market products, June 2026 (source: RedStone data in Castle Labs, Real-World Assets: Bringing TradFi Onchain (June 2026)) | Product | Issuer | Market cap | Notes | | --- | --- | --- | --- | | USYC | Circle (acquired Hashnote, January 2025) | $3.07B | Tokenised US Treasury bills | | BUIDL | BlackRock, via Securitize | $3B, plus $829M in BUIDL-I | Launched March 2024; reserve asset for Ondo OUSG and Ethena USDtb | | JTRSY | Janus Henderson Anemoy, via Centrifuge | $860M | Dual external credit ratings | | USTB | Superstate | $838M | Among the most DeFi-composable Treasury tokens | | WTGXX | WisdomTree | $791M | About 10% used in DeFi | | YLDS | Figure Certificate Company | $570M | Backed by Treasuries and repo, on Provenance | | TBILL | OpenEden | $197M | Rated AA+f/S1+ by S&P and A-bf by Moody's | Where RWAs are used in DeFi, June 2026 (source: RedStone data in Castle Labs, Real-World Assets: Bringing TradFi Onchain (June 2026)) | Protocol | RWA activity | Main assets | | --- | --- | --- | | Morpho | Over $1B RWA deposits, $400M active loans, about 10% of total deposits | syrupUSDC, PRIME, AA_FalconXUSDC, mF-ONE, reUSD | | Aave Horizon | About $500M market size (launched August 2025) | USCC, USTB, VBILL, JAAA as collateral; RLUSD, GHO and USDC lent | | Aave v3 | Over $330M of Maple deposits | syrupUSDC, syrupUSDT | | Kamino | $400M | PRIME, syrupUSDC, ONyc, USCC, over $20M of xStocks | | Fluid and JupLend | $270M | sUSDai, PST, syrupUSDC, SPYx | - [The Renaissance of Onchain Options](https://castlelabs.io/research/reports/renaissance-of-onchain-options): Mapping the expansion of options as a trading instrument and the volatility that prices them (July 2026, 95 pages; download: https://docsend.com/v/sjv2g/onchainoptions) - Finding: Exchange-traded options volume exceeded futures by more than four times in 2024, while U.S.-listed options reached about 15.2 billion contracts in 2025. - Finding: Onchain product design has expanded beyond DOVs into order books, AMMs, perpetual options, short-dated binaries and prediction markets. - Finding: Across the 22 to 28 June 2026 BTC binary sample, Kalshi averaged $4.24 million in daily volume, Polymarket $2.76 million and Hyperliquid about $400,000. - The Main Onchain Options Platforms in 2026 (https://castlelabs.io/research/reports/renaissance-of-onchain-options#best-onchain-options-platforms): Onchain options did $1.44 billion of notional over 30 days in mid-2026, and premium volume hit all-time highs. Early designs failed on thin liquidity and passive LP pools (Opyn, Hegic, Ribbon, Lyra, Premia); today's venues use order books or RFQs with professional market makers, or package options as simple income products. Rysk's monthly notional grew from $50 million in January to $182 million in May 2026, driven by HYPE holders earning yield on assets they already hold. Onchain options venues, last 30 days to late June 2026 (source: Block Scholes data in Castle Labs, The Renaissance of Onchain Options (July 2026)) | Venue | Model | 30-day notional | 30-day premium | Who it serves | | --- | --- | --- | --- | --- | | Derive (ex-Lyra) | Order book on its own OP Stack L2; options and perps cross-margined | $1.142B (79.2% of notional) | $44.3M (87.2% of premium) | Professional traders and market makers; volume supported by incentives | | Rysk | RFQ; covered calls and cash-secured puts as income | $136.3M (9.5%) | $1.94M | Holders, treasuries and DAOs; Hyperion, a Nasdaq-listed HYPE treasury, runs strategies on it | | Aevo (ex-Ribbon) | Order book on a custom L2; options beside perps and pre-launch markets | $45.1M (3.1%) | $2.52M | Active traders; live options OI only about $3.6M | | Others | Paradex, Kyan, CallPut, Hypersurface, Ithaca, SOFA | Small and fragmented | Rising share of premium in 2026 | Varied, including stock options (CallPut) | - Are Prediction Markets Binary Options? (https://castlelabs.io/research/reports/renaissance-of-onchain-options#prediction-markets-are-binary-options): Yes, for price markets. A Polymarket "BTC Up or Down" contract pays $1 if a condition holds at expiry and $0 otherwise, exactly like a binary (digital) option. A binary can be built from two call options struck close together and scaled to pay $1, so by no-arbitrage its fair price is fixed by the vanilla options market: it equals the slope of the call-price curve at that strike. That makes the options surface a benchmark for prediction-market odds. Comparing Polymarket's "Bitcoin above X on June 28" ladder with Block Scholes' composite surface, the two disagreed by strike: Polymarket was more bullish at some strikes and less at others. Implied probability BTC closes above each strike, 28 June 2026 (about 2.2 days to expiry) (source: Polymarket and the Block Scholes composite volatility surface, aligned to 12:00 ET on 28 June 2026) | Strike | Polymarket | Options surface (Block Scholes) | | --- | --- | --- | | $54,000 | 98.2% | 96.1% | | $56,000 | 93.0% | 91.8% | | $58,000 | 73.9% | 80.1% | | $60,000 | 41.2% | 42.5% | | $62,000 | 10.5% | 9.0% | | $64,000 | 2.4% | 1.6% | - Kalshi vs Polymarket vs Hyperliquid for BTC Binaries (https://castlelabs.io/research/reports/renaissance-of-onchain-options#kalshi-vs-polymarket-vs-hyperliquid): Prediction markets grew from about $2 billion a month to about $30 billion a month in 18 months, with fees over $300 million in June 2026. Kalshi holds over 65% of volume. Hyperliquid's HIP-4 outcome markets went live in May 2026 and passed $150 million of volume; deployers stake 1 million HYPE. BTC one-day binary markets, 22 to 28 June 2026 (source: Kalshi and Polymarket public APIs; Hyperliquid via 0xArchive) | Venue | Average daily volume | Week's volume | Quoted spread | Product age | | --- | --- | --- | --- | --- | | Kalshi | $4.24M | $29.6M | 107-154 bps | About 24 months | | Polymarket | $2.76M | $19.3M | 5.8-19.8 bps (from trades, may look tighter) | About 36 months | | Hyperliquid HIP-4 | About $400K | $2.8M | 50-587 bps | Launched May 2026 | - Chart: BTC options volume: Deribit vs IBIT vs CME (https://castlelabs.io/research/reports/renaissance-of-onchain-options#chart-btc-options-volume-by-venue, report page 14): Institutional demand for crypto options has grown steadily, first through Deribit and more recently through BlackRock's IBIT options and CME. - Chart: The crypto options ecosystem by settlement and payoff (https://castlelabs.io/research/reports/renaissance-of-onchain-options#chart-crypto-options-map, report page 17): Offchain vanilla options, led by Deribit, IBIT, CME, Binance and OKX, are still the clear leader. Onchain vanilla venues are rebuilding around order books, RFQs and simpler products that settle onchain. - Chart: Rysk TVL by asset family, weekly average (https://castlelabs.io/research/reports/renaissance-of-onchain-options#chart-rysk-tvl-by-asset, report page 23): Rysk is built on collateralised option selling, so TVL matters: sellers lock full collateral to earn premium. On Derive, by contrast, buyers can pay small premiums for large payouts. - Chart: Onchain options premium volume, 2026, excluding Derive, Rysk and Aevo (https://castlelabs.io/research/reports/renaissance-of-onchain-options#chart-smaller-onchain-options-venues, report page 27): Kyan, Paradex and CallPut are taking share at the low end. Better infrastructure is not enough on its own: options win when they solve a holder's problem, as Rysk did for HYPE holders. - Chart: Monthly prediction market notional volume (https://castlelabs.io/research/reports/renaissance-of-onchain-options#chart-prediction-market-monthly-volume, report page 41): Prediction markets grew from about $2 billion a month to about $30 billion a month in 18 months, most of it after the 2024 US election, when Polymarket's odds proved more accurate. - Chart: Polymarket and Polymarket US notional volume (https://castlelabs.io/research/reports/renaissance-of-onchain-options#chart-polymarket-notional-volume, report page 42): Polymarket now does over $10 billion a month, driven by crypto, sports and politics. It bought QCEX, a CFTC-licensed exchange, in 2025 to return to the US market. - Chart: Monthly prediction market volume share (https://castlelabs.io/research/reports/renaissance-of-onchain-options#chart-prediction-market-volume-share, report page 44): Kalshi holds over 65% of prediction market volume, helped by its regulated approach and US reach. It is eyeing a $40 billion valuation, against $15 billion targeted by Polymarket. - Chart: BTC one-day binary markets: daily volume by venue (https://castlelabs.io/research/reports/renaissance-of-onchain-options#chart-btc-binary-volume-by-venue, report page 51): On BTC one-day binaries, Kalshi averaged $4.24 million a day, Polymarket $2.76 million and Hyperliquid $400K in the window we measured. - Chart: BTC one-day binary market spreads (https://castlelabs.io/research/reports/renaissance-of-onchain-options#chart-btc-binary-spreads, report page 52): Hyperliquid had the widest and most volatile spreads (50 to 587 bps), Kalshi a steady 107 to 154 bps. Polymarket's 5.8 to 19.8 bps is reconstructed from trades and likely understates its true spread. - Chart: BTC market-implied price range to year-end (https://castlelabs.io/research/reports/renaissance-of-onchain-options#chart-btc-implied-probability-cone, report page 80): Options price a 3.9% chance that BTC closes the year above $100K, a 90% chance it ends below $87K and a 75% chance it ends below $75K. - Chart: Chance of BTC reaching $100K by year-end: options vs Polymarket (https://castlelabs.io/research/reports/renaissance-of-onchain-options#chart-btc-100k-probabilities, report page 81): The options surface gives BTC a 3.9% chance of finishing above $100K and a 10.5% chance of touching it at any point. Polymarket's touch market priced 14.5%. - Q: Which onchain options platform is the largest? A: Derive. Over the 30 days to late June 2026 it did $1.142 billion of notional and $44.3 million of premium, 79.2% of onchain options notional and 87.2% of premium, though market-maker rewards and token incentives support that volume. Rysk was second with $136.3 million, then Aevo with $45.1 million. - Q: Why did onchain options struggle for so long? A: Options split liquidity across many strikes and expiries, while perps concentrate it in one book. Early designs pushed hard-to-hedge risk onto passive LP pools, market makers struggled to update hundreds of quotes onchain, and products were too complex for retail but not professional enough for institutions. Vault yields also compressed until premiums no longer paid for the risk. - Q: What changed for onchain options in 2026? A: Yield. With stablecoin and basis yields compressed, selling volatility through covered calls and cash-secured puts became competitive, and builders call volatility "the last product left that is unfarmed." Order books and RFQs replaced options AMMs, cheaper chains cut costs, and products such as Rysk reframed options as income on assets users already hold. - Q: Are prediction markets the same as binary options? A: For financial outcomes, yes. A BTC up/down or "above X" contract pays $1 or $0 at expiry, identical to a binary option, and its fair price is set by the vanilla options market. In June 2026, Polymarket's BTC ladder and the Block Scholes options surface disagreed by up to about 6 points at some strikes, in both directions. - Q: How big are prediction markets? A: About $30 billion a month in mid-2026, up from about $2 billion 18 months earlier, with fees over $300 million in June 2026, roughly double the fees of lending. Kalshi holds over 65% of volume, Polymarket does over $10 billion a month, and Hyperliquid's HIP-4 passed $150 million in its first weeks. - Q: What was BTC options skew telling us in 2026? A: That traders paid up for protection. BTC's 25-delta skew was negative for most of 2026 and never stayed above zero, dropping to -30% in the February sell-off. In June the 7-day put wing traded 10-25% above at-the-money volatility while calls traded 5-13% below, so the smile steepened mainly from demand for downside puts. - Q: What happened to BTC volatility in June 2026? A: On 5 June 2026 BTC fell below $60,000 for the first time since October 2024, 7-day realised volatility jumped from 30% to 70%, and 7-day implied volatility spiked to 65%, inverting the term structure. Within a week short-dated implied volatility fell back toward 30%, a multi-year low, showing how volatile volatility itself was. What BTC options priced for year-end 2026 (source: Block Scholes composite volatility surface; Polymarket) | Outcome | Probability | | --- | --- | | BTC above $100K at year-end (options) | 3.9% | | BTC touches $100K at any point before year-end (options) | 10.5% | | BTC touches $100K before 2027 (Polymarket) | 14.5% | | BTC above $87K at year-end (options) | 10% | | BTC below $75K at year-end (options) | 75% | Options in traditional markets, for scale (source: Cboe; Castle Labs and Block Scholes) | Data point | Value | | --- | --- | | Options vs futures contracts traded globally, 2024 | Over 4x | | US-listed options contracts, 2025 (sixth record year) | About 15.2B, ~$36B premium a day | | SPX zero-day options, 2025 | 59% of the product's volume, 2.3M contracts a day | | BlackRock IBIT options open interest, April 2026 | $27.6B, above Deribit BTC options | - [The Verticalisation Thesis: How Blockchain Revenue Models Are Evolving](https://castlelabs.io/research/reports/the-verticalisation-thesis-how-blockchain): Covering the expansion of revenue streams by blockchains (July 2026, 53 pages; download: https://docsend.com/v/sjv2g/verticalisation-of-blockchain-revenue) - Finding: Ethereum chain fees fell from $9.91 billion in 2021 to $523 million in 2025, a 94.7% decline as execution and rollup settlement became cheaper. - Finding: Hyperliquid generated more than $1 billion in fees over the previous 12 months, including $881 million from its perpetual DEX and about $24 million from HIP-3. - Finding: Ethereum, Tron, Solana and Hyperliquid held a combined $48 billion in TVL, but the report found revenue aligned more closely with transaction and stablecoin settlement than with TVL alone. - Why Blockchain Revenue Is Falling (https://castlelabs.io/research/reports/the-verticalisation-thesis-how-blockchain#why-blockchain-revenue-is-falling): Most chains make money by selling blockspace: users pay gas to get their transactions executed. That product has turned into a commodity. Blockspace from competing L1s, L2s and blockspace-as-a-product providers is interchangeable for most uses, competition keeps pushing fees down, and the most successful apps now earn more than the chains they run on. In 2021 Ethereum earned $9.91 billion in chain fees, about $826 million a month at the peak, and its chain fees were 1.94 times the fees of every app on it combined. The rollup-centric roadmap then moved activity to L2s, and the Dencun upgrade in 2024 cut L2 transaction costs by 10x with blobs, which also cut what the L1 earns. Apps keep growing while chain fees shrink. Over the 30 days to June 2026, application fees on Ethereum, Solana, Base, Arbitrum and Avalanche were far larger than the chain fees each network earned. Chain fees at the peak and in 2025 (source: DefiLlama data in Castle Labs and Kaiko Research, The Verticalisation Thesis (July 2026)) | Network | Peak chain fees | 2025 chain fees | Change | | --- | --- | --- | --- | | Ethereum | $9.91B (2021) | $523M | Down 94.7% in four years | | Arbitrum | $65.6M (2023) | $18.2M | Timeboost now adds a second revenue line | | Avalanche | $63.5M (2023) | $8.4M | Down 86.8% | | Solana | $713M (2024) | $645M | Down 9.6%, held up by memecoin and DeFi trading | - How Hyperliquid Makes Money (https://castlelabs.io/research/reports/the-verticalisation-thesis-how-blockchain#how-hyperliquid-makes-money): Hyperliquid is the chain and the main app at the same time. Instead of renting blockspace to outside exchanges, it runs the perp DEX itself and keeps the fees, most of which go to buying back HYPE. It made about $960 million in protocol fees in 2025 and over $1 billion in the last 12 months, which puts it among the largest fee earners in crypto after Tether and Circle. The perp DEX earned $881 million over the past year, against $10.4 million from HyperEVM gas. Inside the trading engine, perp fees are about 88% of gross revenue, builder-code distribution about 9% and spot fees about 2%. Each new product adds another line that Hyperliquid takes a cut of. HIP-3 shows how the model works in practice. TradeXYZ listed perps on the S&P 500, TSLA, NVDA and other traditional assets, and after the first strike on Iran on 28 February 2026, Hyperliquid was the only venue trading crude oil with active price discovery. Hyperliquid did not list those markets, but it handled distribution and kept a share of the revenue. The biggest builder-code earners are Phantom ($22M), BasedApp ($15M) and pvp.trade ($8M). Hyperliquid revenue lines, July 2026 (source: Castle Labs and Kaiko Research, The Verticalisation Thesis (July 2026)) | Revenue line | How it works | What it has earned or done | | --- | --- | --- | | Perp DEX | Hyperliquid's own exchange; most fees fund HYPE buybacks | $881M over the past year; weekly totals of $10M to $30M | | HyperEVM gas | Blockspace sold to apps on the chain | $10.4M over the past year; 1.2% of revenue on average, about 2% at peak | | HIP-3 markets | Deployers stake at least 500K HYPE to list their own perps and share 50% of the revenue | About $24M since launch; TradeXYZ markets are over 30% of Hyperliquid open interest, peaking near $3B | | Builder codes | Outside frontends add their own fee on top of base fees | Over 25 active codes; $76M lifetime revenue for builders since October 2024; 5% to 10% of volume | | HIP-4 outcome markets | Binary options; deployers stake 1M HYPE; no fees during testing, then 50% to Hyperliquid | Live since May 2026; over $250M in volume | | Priority fees | Traders pay a small fee for lower-latency order processing | Live since April 2026; about $4M so far | | USDC reserve share (AQAv2) | USDC is the designated quote asset and shares 90% of its reserve income with Hyperliquid | An extra $130M to $150M a year | - Base vs BSC vs Mantle vs Ink: Why Exchanges Run Their Own Chains (https://castlelabs.io/research/reports/the-verticalisation-thesis-how-blockchain#exchange-chains-base-bsc-mantle-ink): Binance, Coinbase, Bybit and Kraken all run or incubate their own chains, and the sequencer fee is not why. A chain gives the exchange its own onchain distribution, keeps the DeFi revenue that follows its users, and gives its balance-sheet products a place to settle that it controls. Chain performance tracks the parent exchange, not the technology. The ranking of BSC, Base, Mantle and Ink by TVL, users and revenue mirrors the ranking of their parents by exchange volume almost exactly. Mantle and Ink together hold about $300 million in TVL, less than a single mid-sized DeFi protocol, alongside nearly $800 million in stablecoins. The next year will show whether the exchange funnel turns into real chain revenue. Exchange chains compared, July 2026 (source: Castle Labs and Kaiko Research, The Verticalisation Thesis (July 2026)) | Chain | Parent | Revenue over the past year | How it makes money | | --- | --- | --- | --- | | Base | Coinbase | $64.0M; up 26%, from about $144K to $182K a day | No token and no DAO; stopped sharing fees with Optimism, so Coinbase keeps 100% of sequencer revenue. Coinbase's onchain loan book settles through Morpho on Base. | | BSC | Binance | $21.4M | 18.7M transactions a day at about $0.002 each. The chain keeps BNB useful as gas and collateral and keeps Binance's balance-sheet products onchain. | | Mantle | Bybit | $0.5M | mETH captures staking yield and MUSD anchors a stablecoin stack; $547M in stablecoins | | Ink | Kraken | $0.2M | Closest in design to Base but without a flagship integration yet; $222M in stablecoins | - Chart: Chain fees by network, 2021 to 2025 (https://castlelabs.io/research/reports/the-verticalisation-thesis-how-blockchain#chart-chain-fees-2021-2025, report page 11): On a log scale, Solana is the only network whose fees held up in 2025. Ethereum, Arbitrum and Avalanche all sit far below their peaks. - Chart: Chain fees vs application fees, 30 days to June 2026 (https://castlelabs.io/research/reports/the-verticalisation-thesis-how-blockchain#chart-chain-vs-application-fees, report page 12): On every network shown, the apps earn far more than the chain. In 2021 it was the other way round: Ethereum's chain fees were 1.94 times its application fees. - Chart: Hyperliquid weekly revenue (https://castlelabs.io/research/reports/the-verticalisation-thesis-how-blockchain#chart-hyperliquid-weekly-revenue, report page 19): Weekly revenue ranges from $10 million to $30 million. Perp fees are about 88% of gross revenue, builder-code distribution about 9% and spot fees about 2%. - Chart: Hyperliquid revenue by source: perp DEX vs chain (https://castlelabs.io/research/reports/the-verticalisation-thesis-how-blockchain#chart-hyperliquid-revenue-source, report page 19): HyperEVM gas is a rounding error next to the exchange: 1.2% of Hyperliquid's revenue on average and about 2% at its peak. - Chart: Hyperliquid monthly protocol fees, December 2024 to July 2026 (https://castlelabs.io/research/reports/the-verticalisation-thesis-how-blockchain#chart-hyperliquid-monthly-fees, report page 24): Over $1 billion in fees in the last 12 months, which puts Hyperliquid among the largest fee earners in crypto after Tether and Circle. - Chart: HIP-3 crude oil (xyz:CL) weekend volume (https://castlelabs.io/research/reports/the-verticalisation-thesis-how-blockchain#chart-hip3-crude-oil-volume, report page 26): Volume on TradeXYZ's crude oil perp jumped after the first strike on Iran on 28 February 2026, when Hyperliquid was the only venue trading oil with active price discovery. HIP-3 has added about $24 million to Hyperliquid's revenue. - Chart: Arbitrum Timeboost revenue (https://castlelabs.io/research/reports/the-verticalisation-thesis-how-blockchain#chart-arbitrum-timeboost-revenue, report page 20): Timeboost has made about $7.5 million since April 2025, roughly half of the $13.2 million Arbitrum earned as a chain over the past year. Arbitrum's daily revenue still fell about 40% over the year. - Chart: Monthly revenue of exchange chains (https://castlelabs.io/research/reports/the-verticalisation-thesis-how-blockchain#chart-exchange-chain-revenue, report page 31): Base is one of only two major chains whose revenue grew over the past year, up 26% to about $182K a day. Mantle and Ink barely register. - Chart: Plasma TVL vs XPL price (https://castlelabs.io/research/reports/the-verticalisation-thesis-how-blockchain#chart-plasma-tvl-vs-xpl, report page 36): Plasma's TVL peaked at $10 billion around launch and fell to $1.78 billion as the XPL incentives lost value. The chain earned $3.7K in fees in June, 95% below its peak. - Chart: Monthly revenue: Ethereum, Hyperliquid, Solana, Tron (https://castlelabs.io/research/reports/the-verticalisation-thesis-how-blockchain#chart-l1-monthly-revenue, report page 42): The four chains earned $1.64 billion in November 2021, almost all of it on Ethereum. In 2026 they average $80 million a month, and Ethereum earned $5.3 million in May 2026, 0.3% of its peak. - Chart: TVL per chain (https://castlelabs.io/research/reports/the-verticalisation-thesis-how-blockchain#chart-tvl-per-chain, report page 44): The four chains hold $48 billion in TVL, $37.5 billion of it on Ethereum, which earned $118 million over the past year. Hyperliquid, the biggest earner, has the smallest TVL at $1.5 billion. - Chart: Stablecoin market cap per chain (https://castlelabs.io/research/reports/the-verticalisation-thesis-how-blockchain#chart-stablecoins-per-chain, report page 45): The four chains hold $270 billion in stablecoins, 92% of it on Ethereum ($158B) and Tron ($90.1B). The chains that settle the most stablecoin flow earn the most: Tron's transfer fees bring in $444 million a year. - Chart: Price-to-sales ratio per chain, 12 June 2026 (https://castlelabs.io/research/reports/the-verticalisation-thesis-how-blockchain#chart-price-to-sales, report page 47): Hyperliquid trades at 16x sales and Tron at 67x. Solana at 1,078x and Ethereum at 1,701x only make sense if the market prices them as money rather than as claims on cash flow. - Q: What is blockchain verticalisation? A: A chain growing by owning one high-value product or revenue stream instead of renting blockspace to many outside apps. A horizontal chain earns gas on whatever apps it attracts. A vertical chain is the app, or is tightly tied to one, so it keeps a much bigger share of the activity it enables. Hyperliquid, which runs its own perp DEX, is the clearest example. - Q: Why are blockchain fees falling? A: Blockspace became a commodity. Competing L1s, L2s and blockspace providers are interchangeable for most uses, the race for cheaper transactions pushed fees down, and the Dencun upgrade in 2024 moved rollup data into near-free blob space. Ethereum's chain fees fell from $9.91 billion in 2021 to $523 million in 2025, a 94.7% drop. - Q: How does Hyperliquid make money? A: Mostly from its own perp DEX, which earned $881 million over the past year against $10.4 million from HyperEVM gas. It also takes 50% of HIP-3 market revenue (about $24 million since launch), earns from builder codes and priority fees (about $4 million since April 2026), and receives 90% of the reserve income on USDC used on the platform, worth an extra $130 million to $150 million a year. Most fees go to HYPE buybacks. - Q: What is HIP-3 on Hyperliquid? A: HIP-3 lets anyone deploy their own perp markets on Hyperliquid by staking at least 500,000 HYPE, in exchange for sharing 50% of the revenue with the protocol. TradeXYZ used it to list perps on the S&P 500, TSLA, NVDA and crude oil, and its markets grew to over 30% of Hyperliquid's open interest, peaking near $3 billion. HIP-4 applies the same model to outcome markets with a 1 million HYPE stake. - Q: What is Arbitrum Timeboost and how much has it made? A: Timeboost is an auction for transaction priority that the Arbitrum DAO introduced in April 2025. It has generated about $7.5 million in fees, roughly half of the $13.2 million Arbitrum earned as a chain over the past year. Arbitrum's daily revenue still fell about 40% over the year, so we see Timeboost as a bolt-on that cannot offset cheaper blockspace on its own. - Q: Why do exchanges like Coinbase and Binance run their own chains? A: To own onchain distribution, not to sell blockspace. A chain lets the exchange keep the DeFi revenue its users generate and settle its own balance-sheet products. Base earned $64.0 million over the past year and Coinbase keeps all of it; BSC processes 18.7 million transactions a day at about $0.002 each to keep BNB useful. Each chain's results track its parent's exchange volume. - Q: Are most blockchains overvalued? A: On fee income, most are. As of 12 June 2026, Hyperliquid traded at 16 times sales and Tron at 67 times, while Solana traded at 1,078 times and Ethereum at 1,701 times. Those last two only make sense if the market values ETH and SOL as money rather than as claims on cash flow. Because HYPE buybacks tie the token to revenue, Hyperliquid's valuation is the most defensible of the group. New verticalised chains, July 2026 (source: Castle Labs and Kaiko Research, The Verticalisation Thesis (July 2026)) | Chain | Vertical bet | Key figures | | --- | --- | --- | | Plasma | L1 for stablecoin payments: zero-fee USDT transfers without holding XPL, plus the Plasma One card for spending at Visa merchants | TVL peaked at $10B around launch and stands at $1.78B; $3.7K in chain fees in June, 95% below peak | | MegaETH | Own stablecoin, USDm, issued with Ethena, with reserve income funding MEGA buybacks | USDm market cap about $230M, implying $8M to $9M a year in buybacks at a 3.6% to 3.7% SOFR | | Robinhood Chain | Arbitrum-stack L2 for tokenised traditional assets that Robinhood itself issues | Live since the start of July 2026; TVL $175M; Robinhood Earn vault on Morpho, curated by Steakhouse, targets 7% APY on a $140M market | | Tempo | Enterprise chain for stablecoin payments, incubated by Stripe and Paradigm | Gas paid in the stablecoin being sent; Zones privacy for institutions; targets Shopify merchants, DoorDash drivers, Klarna instalments and treasury flows | TVL, stablecoins, revenue and valuation of four L1s (source: Kaiko and DefiLlama data in Castle Labs and Kaiko Research, The Verticalisation Thesis (July 2026)) | Chain | TVL | Stablecoins | Revenue | P/S ratio (12 June 2026) | | --- | --- | --- | --- | --- | | Ethereum | $37.5B | $158B | $118M over the past year; $5.3M in May 2026, 0.3% of its peak | 1,701x | | Tron | $4.3B | $90.1B | $444M annualised; $30M to $75M a month for several years | 67x | | Solana | $4.6B | $15.8B | Peaked at $121M a month in January 2025, at the height of memecoin trading | 1,078x | | Hyperliquid | $1.5B | $6.3B | Largest of the four; peaked at $119M in August 2025, now $30M to $50M a month | 16x | - [Fixed-Rate Lending: Market Structure and Protocol Design](https://castlelabs.io/research/reports/fixed-rate-lending-market-structure): A report on the protocols building fixed-rate and fixed-maturity credit products (August 2026, 34 pages; download: https://docsend.com/v/sjv2g/fixed-rate-lending) - Finding: DeFi lending had $28.5 billion in active loans at the report snapshot, with almost all demand still served by variable-rate markets. - Finding: Morpho Midnight uses tradable credit and debt units to improve early exits and had reached $2 million in active loans by the end of August, after its July launch. - Finding: Jupiter Offerbook had reached $450,000 in active loans using fixed terms of one to 30 days and maturity-based collateral transfer instead of continuous liquidation. - Fixed-Rate vs Variable-Rate Lending in DeFi (https://castlelabs.io/research/reports/fixed-rate-lending-market-structure#fixed-rate-vs-variable-rate-lending): DeFi lending had $28.5 billion in active loans at the time of the report, and almost all of it was variable-rate. Variable pools solved instant borrowing against collateral, but the rate follows a utilisation curve. In a stress event that curve moves, rates spike, and borrowers are forced to exit or deleverage. Fixed-rate and fixed-maturity markets let lenders know their yield and borrowers know their cost before the loan starts. Demand comes from three groups. Date borrowers, such as funds, treasuries, RWA issuers and basis desks, need debt that matures when an asset, redemption window or strategy does. Certainty borrowers, such as loopers and levered yield users, need a stable cost so their spread does not shrink. Lenders and curators want to choose duration, collateral and return instead of taking whatever utilisation produces. Why earlier fixed-rate markets struggled, and how new designs answer (source: Castle Labs, Fixed-Rate Lending: Market Structure and Protocol Design (August 2026)) | Problem | What goes wrong | How new designs answer it | | --- | --- | --- | | Fragmented liquidity | Liquidity splits by maturity, rate, collateral and duration, so each market is thin | Multi-market offers (Morpho Midnight), rate and duration grids (Kamino), virtual markets (Loopscale) | | No early exit | Lenders are stuck until maturity unless someone buys their position | Tradable credit and debt units (Midnight), discounted bond tokens on an AMM (TermMax, Fira), a FIFO withdrawal queue (Kamino) | | Cold start | Capital sits idle, earning nothing, while it waits for a match | Capital moves only when a match settles and earns variable yield until then (Midnight, Term V2, Jupiter Offerbook, Kamino) | | Maturity | Borrowers may repay, roll, fall back to variable debt or fail to repay | Auto-roll with a fallback to variable rates (Tenor, Kamino); auto-refinancing with 48 hours to repay (Loopscale) | - Morpho Midnight vs Jupiter Offerbook vs Kamino: How the Big Lenders Do Fixed Rates (https://castlelabs.io/research/reports/fixed-rate-lending-market-structure#morpho-midnight-vs-jupiter-offerbook-vs-kamino): Morpho, Jupiter and Kamino all run large variable-rate markets, and in 2026 each added fixed terms in a different way. Morpho built a separate protocol with tradable positions. Jupiter built an intent-based book with no price-based liquidation. Kamino added fixed-rate reserves inside its existing lending market, so borrowers see a term structure next to the variable rate they already use. Tenor Finance, which DeFi Frontier calls the HIP-3 of Midnight, is the access layer on top. It adds auto-renewal, onchain OTC quotes shared with whitelisted counterparties, and organisation accounts that let institutions run gated credit markets with their own KYC rules. The Tenor team expects asset managers on one side and businesses on the other. Fixed-rate products from variable-rate incumbents, August 2026 (source: Castle Labs, Fixed-Rate Lending: Market Structure and Protocol Design (August 2026)) | Protocol | Design | Exit before maturity | At maturity | Active loans | | --- | --- | --- | --- | --- | | Morpho Midnight (launched July 2026) | Intent-based zero-coupon lending; positions are credit units and debt units, and their price sets the rate | Lenders sell credit units; borrowers buy debt units to cut what they owe | Tenor, built on Midnight, rolls loans through keepers and falls back to Morpho Blue's variable rate | $2M; Morpho Vaults hold over $4B that can quote once the vault adapter ships | | Jupiter Offerbook (public beta June 2026) | Intent-based book with terms of 1 to 30 days and no price-based liquidation | Short fixed terms instead of a secondary market | If the borrower does not repay, the lender takes the collateral | $450K | | Kamino Fixed Rate Lending (whitepaper) | Fixed-rate reserves inside Kamino Lend, laid out as a grid of rates and durations | FIFO withdrawal queue; the wait is capped by the reserve's duration | Auto-roll into the next term, or fall back to the variable rate | Whitepaper stage at publication | - The Main Fixed-Rate Lending Protocols in 2026 (https://castlelabs.io/research/reports/fixed-rate-lending-market-structure#fixed-rate-lending-protocols-2026): Purpose-built protocols set the fixed rate in different ways: auctions, AMM curves, order books or rates the borrower picks. Fira has by far the most active loans at $430 million, but almost all of that is its Usual Zero Rate market, so it is not directly comparable with multi-market venues. Curators are building fixed rates too. KPK runs a fixed-rate ETH lending vault on Euler with about $700K in deposits. Its borrow rate resets once a month at 90% of Treehouse's ETH staking-yield benchmark minus 20 bps, which makes looping costs predictable. Aave launched Stable Vaults in July, an early sign of predictable credit products from the largest lender. Purpose-built fixed-rate protocols, August 2026 (source: DefiLlama data in Castle Labs, Fixed-Rate Lending (August 2026)) | Protocol | How the rate is set | Active loans | Notable detail | | --- | --- | --- | --- | | Fira | Borrowers sell a Bond Token (BT) at a discount on a BT/FW AMM; BT works like Pendle's PT and the Coupon Token (CT) like YT | $430M, almost all from Usual Zero Rate (UZR) | Interest is built into the discount at origination, not charged over time | | Loopscale | Credit order book of bilateral loans; virtual markets pool liquidity by collateral, loan asset and duration | $55M | Auto-refinances at maturity and gives 48 hours to repay if no lender is found; curator vaults for passive lenders | | Liquity V2 | Borrowers pick their own rate when minting BOLD against ETH, wstETH or rETH | $32M of BOLD in circulation | If BOLD trades below peg, the lowest-rate borrowers are redeemed first | | TermMax | Fixed Rate Token (FT) trades at a discount like a zero-coupon bond; 1 FT + 1 XT = 1 debt token | $30M | Lenders quote by size, for example the first $10K at 5% and the next $10K at 5.5% | | Term Labs | Auction where every bid clears at one rate; Term V2 (early August 2026) adds intent-based matching | $3.7M | Hit by an $8.5M governance exploit in its Meta vaults | | Flex | Borrowers pick their rate, as in Liquity V2; lender exits redeem the lowest-rate Troves first through a collateral auction | About $900K two months after launch | Bad debt from underwater liquidations can be spread across the lender vault | | IRIS Credit | Solvers quote a fixed rate and refinance across lending venues, posting bonds sized from a three-year simulation | Pre-launch | Takes blue-chip collateral because it has a rate history to model | - Chart: DeFi lending active loans, January to August 2026 (https://castlelabs.io/research/reports/fixed-rate-lending-market-structure#chart-lending-active-loans, report page 4): Almost all of the $28.5 billion is variable-rate debt, priced by utilisation curves that can spike in a stress event. - Chart: Fira active loans, 2026 (https://castlelabs.io/research/reports/fixed-rate-lending-market-structure#chart-fira-active-loans, report page 22): Fira leads the category at $430 million, but almost all of it comes from the Usual Zero Rate market. - Chart: Loopscale active loans, 2026 (https://castlelabs.io/research/reports/fixed-rate-lending-market-structure#chart-loopscale-active-loans, report page 24): Loopscale's active loans sit at $55 million, one of the largest in the category after Fira. - Chart: Liquity BOLD market cap, 2026 (https://castlelabs.io/research/reports/fixed-rate-lending-market-structure#chart-bold-market-cap, report page 26): $32 million of BOLD is in circulation: the debt Liquity borrowers have minted at rates they chose themselves. - Chart: TermMax active loans, 2026 (https://castlelabs.io/research/reports/fixed-rate-lending-market-structure#chart-termmax-active-loans, report page 20): TermMax reached $30 million in active loans, with curator vaults letting passive lenders earn fixed yield without quoting. - Chart: Term Finance active loans, 2026 (https://castlelabs.io/research/reports/fixed-rate-lending-market-structure#chart-term-finance-active-loans, report page 18): Term's active loans sit at $3.7 million. Term V2, released in early August, keeps capital earning on venues like Aave and Morpho until it is matched. - Chart: Morpho Midnight active loans since launch (https://castlelabs.io/research/reports/fixed-rate-lending-market-structure#chart-morpho-midnight-active-loans, report page 9): Midnight reached $2 million in active loans by the end of August. Over $4 billion in Morpho Vaults can start quoting once the vault adapter is released. - Chart: Flex active loans since May 2026 (https://castlelabs.io/research/reports/fixed-rate-lending-market-structure#chart-flex-active-loans, report page 28): Flex reached about $900,000 in active loans two months after launch. - Chart: Jupiter Offerbook active loans (https://castlelabs.io/research/reports/fixed-rate-lending-market-structure#chart-jupiter-offerbook-active-loans, report page 13): Offerbook reached $450,000 in active loans. Its growth depends on how much long-tail collateral lenders are willing to underwrite directly. - Q: What is fixed-rate lending in DeFi? A: A loan where the interest rate, and usually the maturity, is set at the start, so the lender knows the yield and the borrower knows the cost. Most DeFi lending uses variable rates that move with pool utilisation. Of the $28.5 billion in active DeFi loans at the time of the report, almost all was variable-rate. - Q: Why has fixed-rate lending been hard to build onchain? A: Three problems held back earlier versions. Liquidity splits across maturities, rates and collateral, so each market is thin. Lenders cannot exit before maturity unless someone buys their position. And capital sits idle while it waits for a match. Newer designs answer with multi-market offers, tradable positions and capital that keeps earning variable yield until a loan is matched. - Q: How does Morpho Midnight work? A: Midnight, launched by Morpho in July 2026, is intent-based zero-coupon lending. Lenders and borrowers post offers that do not lock capital, and capital only moves when a match settles. Positions are credit units and debt units that can be traded, so lenders can exit early by selling credit units and borrowers can buy debt units back. It reached $2 million in active loans by the end of August. - Q: What is Tenor Finance? A: An access layer built on Morpho Midnight, which DeFi Frontier calls the HIP-3 of Midnight. It adds auto-renewal through independent keepers, a fallback to Morpho Blue's variable rate when no fixed match exists, onchain OTC quotes with whitelisted counterparties, and organisation accounts that can run KYC-gated credit markets. - Q: How does Jupiter Offerbook avoid liquidations? A: Loans run for fixed terms, usually 1 to 30 days, and there is no price-based liquidation. If the borrower does not repay at maturity, the lender claims the collateral. That lets lenders accept long-tail collateral such as NFTs or RWAs with no active price discovery, as long as they are willing to underwrite it. Offerbook had $450,000 in active loans. - Q: How do TermMax and Fira compare to Pendle? A: Both split a loan into a fixed claim and a floating piece, as Pendle does. TermMax's Fixed Rate Token (FT) redeems for one unit of the loan asset at maturity and trades at a discount before then, like Pendle's PT. Fira's Bond Token (BT) works the same way, and its Coupon Token (CT) carries the floating yield, like YT. Fira had $430 million in active loans, almost all in its Usual Zero Rate market; TermMax had $30 million. - Q: What risks remain with fixed-rate loans? A: Fixed rates make the cost of debt explicit but do not remove collateral, default, redemption, solver, LP or refinancing risk. In Liquity V2 and Flex, borrowers who pick a low rate are first in line to be redeemed. If rates run far enough past an IRIS Credit solver's quote, the solver's bond runs out and the borrower pays the floating rate. Term Labs was hit by an $8.5 million governance exploit in its Meta vaults. Fixed-rate lending by protocol, end of August 2026 (source: DefiLlama data in Castle Labs, Fixed-Rate Lending (August 2026). All variable and fixed DeFi lending: $28.5B in active loans.) | Protocol | Type | Active loans | | --- | --- | --- | | Fira | Purpose-built, Pendle-style | $430M (mostly the UZR market) | | Loopscale | Purpose-built credit order book | $55M | | Liquity V2 | Borrower-set rate CDP | $32M of BOLD in circulation | | TermMax | Purpose-built, zero-coupon tokens | $30M | | Term Labs | Purpose-built auction | $3.7M | | Morpho Midnight | Incumbent, intent-based | $2M | | Flex | Borrower-set rate money market | About $900K | | KPK vault on Euler | Curator-built | About $700K in deposits | | Jupiter Offerbook | Incumbent, intent-based | $450K | | Kamino Fixed Rate Lending | Incumbent, rate and duration grid | Whitepaper | | IRIS Credit | Solver-based origination layer | Pre-launch | - [Mapping the Perpetual Trading Ecosystem](https://castlelabs.io/research/reports/mapping-the-perpetual-trading-ecosystem): A comparative analysis of perp trading venues across design, GTM and microstructure (October 2026, 81 pages; download: https://docsend.com/view/uwe948jmc9hqmkkd) - Finding: Hyperliquid holds over 56% of onchain perp open interest, followed by Aster (9.7%), Variational (6.2%) and Lighter (5.7%), and more than 261 perp DEXs have launched since Hyperliquid. - Finding: RWA perps have grown more than 100x since October 2025 to $123 billion in August 2026 volume, and TradeXYZ and Variational account for 90% of it. - Finding: At $1 million, Hyperliquid had the lowest median slippage of the five venues tested on crypto majors, 3.7 bps against 3.8 bps on Binance, and after fees it was 1% to 24% cheaper to trade than the centralised median. Lighter was over 60% cheaper because it charges no taker fee. - Finding: For TradFi perps at $100,000, TradeXYZ had the tightest spread (0.35 bps) and Lighter the lowest cost to open (1.04 bps). On weekends, venue indices diverged by up to 30 bps and Lighter's WTI funding reached +95% a year for longs. - Finding: Pre-IPO perps on Anthropic and OpenAI are the thinnest markets measured: venue prices differed by 379 and 569 bps, no venue held more than $160,000 within 10 bps of mid, and holding a long cost from about 1% to 33% a year depending on the venue. - What Is a Perp DEX? (https://castlelabs.io/research/reports/mapping-the-perpetual-trading-ecosystem#what-is-a-perp-dex): A perp DEX is an onchain exchange for perpetual futures: leveraged contracts that track an asset's price and never expire. Instead of settling on a date, a perp uses a funding rate, paid between longs and shorts, to keep the contract price close to the underlying index. Traders keep custody of their collateral and positions settle onchain. Perp DEXs now handle over 13% of perp volume, up from 4% in 2024, and onchain perp open interest reached a record $15.18 billion. More than 261 perp DEXs have launched since Hyperliquid. Most use one of three designs. The three main perp DEX designs (source: Castle Labs, Mapping the Perpetual Trading Ecosystem (October 2026)) | Design | How it works | Examples | | --- | --- | --- | | Pooled liquidity | Traders trade against a shared liquidity pool rather than another trader. This was the category's first model. | GMX, GMTrade, Jupiter Perps | | High-performance order book | A dedicated appchain or rollup runs a CEX-style order book at CEX speed while users keep self-custody. | Hyperliquid, Lighter, edgeX, Pacifica | | Request for quote (RFQ) | The trader requests a price, a market maker quotes it, and the trade is booked in an isolated settlement pool. There is no order book. | Variational | - Hyperliquid vs Lighter vs Variational (https://castlelabs.io/research/reports/mapping-the-perpetual-trading-ecosystem#hyperliquid-vs-lighter-vs-variational): The three venues take different routes to the same market. Hyperliquid has the deepest book, Lighter is the cheapest to trade on standard orders, and Variational quotes prices instead of running an order book. In short: Hyperliquid suits size on crypto majors, with the lowest median slippage at $1 million of the five venues tested, Binance included. Lighter suits cost-sensitive traders on standard orders because it charges no taker fee. Variational suits RWA trading with size, where its slippage barely rose as orders grew, and its swaps charge a daily carry fee instead of funding. Hyperliquid, Lighter and Variational compared (source: Castle Labs, Mapping the Perpetual Trading Ecosystem (October 2026)) | Metric | Hyperliquid | Lighter | Variational | | --- | --- | --- | --- | | Design | Order book on its own L1, secured by HyperBFT | Order book on an app-specific ZK-rollup with an Ethereum escape hatch | RFQ with no order book; the Omni Liquidity Provider (OLP) is the only maker | | Share of onchain perp OI | 56%+ | 5.7% | 6.2% | | Avg daily volume | $7.90B | $1.80B | $1.29B | | Avg daily open interest | $6.79B | $629.4M | $772.9M | | Turnover | 1.16x a day | 2.87x a day | 1.66x a day | | Fees | Taker fee; 2.6 bps take rate | No fee on standard accounts; 1 bp on advanced order types; 0.8 bps take rate | No maker or taker fees; earns from the spread | | Cost to open, $100k BTC | 4.71 bps | 0.41 bps | Not tested | | Liquidation backstop | HLP vault, then auto-deleveraging | Lighter Liquidity Pool (LLP) vaults | If OLP is liquidated, affected users receive the liquidation penalty | | Token | HYPE; 90% of USDC reserve yield routed to buybacks via AQAv2 | LIT, launched December 2025; fee-funded buybacks and burns | Points program ahead of its token launch | - Best Perp DEXs by Open Interest (https://castlelabs.io/research/reports/mapping-the-perpetual-trading-ecosystem#best-perp-dexs): Open interest is the clearest single measure of where traders keep capital. The table below ranks perp DEXs by current open interest and updates daily from DefiLlama. In the report's September snapshot, Hyperliquid held over 56% of onchain perp open interest, followed by Aster (9.7%), Variational (6.2%) and Lighter (5.7%). Open interest is only half the picture. Turnover shows whether that capital trades or churns: QFEX, Hyperliquid and Variational turn over less than 2x a day, while Jupiter Perps, Pacifica and GMTrade trade 7x to 9x their open interest daily, consistent with points programs and airdrop farming. - Chart: Crypto trading volume: derivatives vs spot (https://castlelabs.io/research/reports/mapping-the-perpetual-trading-ecosystem#chart-derivatives-vs-spot-volume, report page 6): Derivatives made up 82.58% of crypto trading volume across centralised and onchain venues over 30 days. On most major venues, perp trading is 4.7x spot. - Chart: Perp DEX market share over time (https://castlelabs.io/research/reports/mapping-the-perpetual-trading-ecosystem#chart-perp-dex-market-share-evolution, report page 7): The market moved from a few leaders, first dYdX and then GMX, to many venues sharing it. Hyperliquid has led since launch, apart from a short spell in November 2025 when Lighter took the top spot. - Chart: Perp DEX monthly trading volume (https://castlelabs.io/research/reports/mapping-the-perpetual-trading-ecosystem#chart-perp-dex-monthly-volume, report page 11): Monthly perp DEX volume peaked at $1.36 trillion in October 2025 and cooled through 2026. - Chart: Monthly onchain RWA perp volume (https://castlelabs.io/research/reports/mapping-the-perpetual-trading-ecosystem#chart-rwa-perp-volume, report page 10): RWA perps reached $147.5 billion of volume in July 2026, 19.6% of all onchain perp volume. At its peak, TradeXYZ's listings drove half of Hyperliquid's volume. - Chart: Perp DEX open interest by protocol (https://castlelabs.io/research/reports/mapping-the-perpetual-trading-ecosystem#chart-perp-dex-open-interest-share, report page 10): Hyperliquid holds 56.8% of the $14.64 billion of open interest across perp DEXs, followed by Aster (9.7%), Variational (6.2%) and Lighter (5.7%). - Chart: Hyperliquid builder code fees by frontend (https://castlelabs.io/research/reports/mapping-the-perpetual-trading-ecosystem#chart-hyperliquid-builder-code-share, report page 17): Phantom leads with 24.0% of the $109.57 million in all-time builder code fees paid to outside frontends. - Chart: Slippage vs total cost on a $100K crypto order (https://castlelabs.io/research/reports/mapping-the-perpetual-trading-ecosystem#chart-cex-vs-dex-slippage-and-cost, report page 64): Binance, Bybit and OKX look far ahead on slippage alone. Add the taker fee, which is what a trader actually pays, and Hyperliquid and Lighter close most of the gap. - Chart: Cost to open a $100K position: slippage vs taker fee (https://castlelabs.io/research/reports/mapping-the-perpetual-trading-ecosystem#chart-perp-cost-fee-vs-slippage, report page 66): A 5 bps taker fee on a $100K BTC order costs $50, about 75x the slippage. Hyperliquid was 7% to 24% cheaper than the centralised median on BTC, ZEC and HYPE; Lighter was over 60% cheaper with no taker fee. - Chart: TradFi perps: TradeXYZ vs centralised venues (https://castlelabs.io/research/reports/mapping-the-perpetual-trading-ecosystem#chart-tradfi-perp-depth, report page 68): On TradFi assets, TradeXYZ holds more resting depth near the mid price, while Binance is tighter on slippage. For crypto, BTC funding already runs at 1.9% to 7.6% a year on every venue. - Chart: Price dispersion across venues: weekdays vs weekends (https://castlelabs.io/research/reports/mapping-the-perpetual-trading-ecosystem#chart-weekend-price-dispersion, report page 72): Crypto prices agree across venues all week. RWA prices drift apart at weekends, when the futures or equity markets they reference are closed. - Chart: Price dispersion on pre-IPO perps (https://castlelabs.io/research/reports/mapping-the-perpetual-trading-ecosystem#chart-pre-ipo-price-dispersion, report page 76): With no reference price, venues disagree by 379 bps on Anthropic and 569 bps on OpenAI, against 0.1 bps on NVDA. Holding an OpenAI long costs about 33% a year in funding; QFEX paid about 1%. - Q: What is a perp DEX? A: A perp DEX is an onchain exchange for perpetual futures, leveraged contracts that track an asset's price and never expire. A funding rate paid between longs and shorts keeps the contract close to the index price, and traders keep custody of their collateral. Perp DEXs now handle over 13% of perp volume, up from 4% in 2024, and the largest is Hyperliquid. - Q: Which perp DEX has the most open interest in 2026? A: Hyperliquid. In our September 2026 data it held over 56% of onchain perp open interest, followed by Aster (9.7%), Variational (6.2%) and Lighter (5.7%). Hyperliquid's own open interest reached a record $9 billion, on roughly $180 billion to $250 billion of monthly volume through 2026. Its turnover of 1.16x a day means traders hold positions rather than churn them. - Q: Hyperliquid vs Lighter: which is cheaper to trade? A: Lighter is cheaper for standard orders because it charges no taker fee. On a $100,000 BTC order, Hyperliquid's slippage was 0.21 bps against 0.41 bps on Lighter, but Hyperliquid's total cost to open was 4.71 bps with its fee, while Lighter's was 0.41 bps. Hyperliquid has the deeper book: at $1 million it had the lowest median slippage of the five venues we tested, including Binance, Bybit and OKX. - Q: Are onchain perp DEXs as liquid as Binance? A: On the largest crypto assets, close. Centralised books were about 3x tighter than Hyperliquid on BTC and ETH at $100,000, but the gap shrinks with size, and once taker fees are included Hyperliquid was 1% to 24% cheaper to trade than the centralised median across BTC, ETH, SOL, ZEC and HYPE. Hyperliquid also matched or exceeded centralised resting depth within 10 bps on four of the five assets. - Q: How does Variational work? A: Variational is an RFQ protocol with no order book. On Omni, each trader has an isolated settlement pool with the Omni Liquidity Provider (OLP), which prices from CEX, DEX and TradFi data and hedges its exposure on external venues. Variational charges no maker or taker fees and earns from the spread, and RWA trading is now over 50% of its volume. - Q: What is the difference between Variational swaps and perps? A: Perps use funding rates to keep the contract close to the index price. Variational swaps track an asset's total return and charge a carry fee once a day at 5pm EST, based on the real cost of financing the underlying in traditional markets, such as SOFR for US equities. Long swap positions receive dividends and short positions pay them. - Q: Which venue is best for trading stock and commodity perps onchain? A: It depends on size and timing. In our weekday US-hours test across seven TradFi assets, TradeXYZ had the tightest spread (0.35 bps) and the lowest slippage at $100,000 (0.84 bps), while Lighter was cheapest to open at 1.04 bps because it charges no fee, and Variational's slippage barely rose with size. On weekends, venue indices diverged by up to 30 bps and some venues kept charging funding, so carry cost matters more than execution. - Q: What is HIP-3 on Hyperliquid? A: HIP-3 lets anyone run their own perp exchange on HyperCore after staking 500,000 HYPE. The deployer sets markets, oracles and risk parameters, and validators can slash the stake for misconduct. HIP-3 markets bring about 30% of Hyperliquid's net volume, TradeXYZ accounts for over 98% of HIP-3 volume, and deployers keep up to 50% of the fees their markets generate. - Q: How do RWA perps trade when stock markets are closed? A: Each venue applies its own pricing rule. TradeXYZ switches to order-book pricing smoothed by a 30-minute EWMA inside dynamic discovery bounds, Ondo lets the oracle move within roughly 5% of the last TradFi close, QFEX applies hard price limits and charges no funding while the reference market is shut, and Variational swaps wait for the market to reopen before liquidating. Perp venue turnover and take rate (source: DefiLlama daily volume, open-interest and fee series) | Venue | Avg daily volume | Avg daily OI | Turnover (x/day) | Take rate | | --- | --- | --- | --- | --- | | QFEX | $182.4M | $217.6M | 0.84x | Not tracked | | Hyperliquid | $7.90B | $6.79B | 1.16x | 2.6 bps | | Variational | $1.29B | $772.9M | 1.66x | Not tracked | | Polymarket Perps | $94.6M | $49.3M | 1.92x | Not tracked | | edgeX | $1.38B | $572.7M | 2.42x | 0.9 bps | | GMX | $86.9M | $32.2M | 2.70x | 2.7 bps | | Lighter | $1.80B | $629.4M | 2.87x | 0.8 bps | | Ondo Perps | $133.4M | $46.1M | 2.89x | Not tracked | | Extended | $258.9M | $87.5M | 2.96x | 2.33 bps | | Jupiter Perps | $237.8M | $34.8M | 6.83x | 3.7 bps | | Pacifica | $402.3M | $48.5M | 8.30x | 3.0 bps | | GMTrade | $796.6M | $90.6M | 8.79x | 0.4 bps | | Kalshi | $886.0M | $30.7M | 28.86x | Not tracked | Crypto perps at $100,000: centralised median vs Hyperliquid vs Lighter (source: Venue public APIs and order books, 23 September 2026) | Asset | Slippage, CEX | Slippage, Hyperliquid | Slippage, Lighter | Cost to open, CEX | Cost to open, Hyperliquid | Cost to open, Lighter | | --- | --- | --- | --- | --- | --- | --- | | BTC | 0.07 bps | 0.21 bps | 0.41 bps | 5.07 bps | 4.71 bps | 0.41 bps | | ETH | 0.10 bps | 0.32 bps | 0.83 bps | 5.10 bps | 4.82 bps | 0.83 bps | | SOL | 1.07 bps | 1.06 bps | 1.73 bps | 6.07 bps | 5.56 bps | 1.73 bps | | ZEC | 4.06 bps | 2.37 bps | 2.53 bps | 9.06 bps | 6.87 bps | 2.53 bps | | HYPE | 3.24 bps | 2.43 bps | 2.20 bps | 8.24 bps | 6.93 bps | 2.20 bps | TradFi perps: cost to open at $100,000 (source: Venue public APIs and order books, 23 September 2026) | Venue | Spread | Slippage at $100k | Average taker fee | Cost to open at $100k | | --- | --- | --- | --- | --- | | TradeXYZ | 0.35 bps | 0.84 bps | 2.06 bps | 2.90 bps | | Lighter | 0.65 bps | 1.04 bps | 0 | 1.04 bps | | Ondo | 1.61 bps | 2.41 bps | 2.5 bps | 4.91 bps | | QFEX | 3.32 bps | 2.40 bps | 6.43 bps | 8.83 bps | | Variational | 4.66 bps | 2.82 bps | 0 | 2.82 bps | ## Live Data Rankings and trackers refreshed every day from public data. Each page has a short answer, tables and a raw JSON snapshot. - [Perp DEX Rankings](https://castlelabs.io/research/perp-dex-rankings): Onchain perp venues by open interest and 24h volume, by chain, with fees and a Binance comparison (JSON: https://castlelabs.io/data/perps.json) As of 14 November 2026, Hyperliquid leads onchain perps with 55% of open interest ($6.18B), followed by Aster (11%) and Variational (6.7%). Tracked onchain perp open interest is $11.21B, down 17% over 30 days. Hyperliquid traded $10.08B of perps in the last 24 hours, 48% of the $21.10B traded on the onchain perp DEXs we track, ahead of Lighter ($2.68B) and Variational ($1.81B). For scale, Binance futures traded $61.94B, so Hyperliquid did 16% of Binance's volume. - [Crypto Revenue Rankings](https://castlelabs.io/research/crypto-revenue-rankings): Blockchains and protocols ranked by fees and revenue over 30 days and 12 months (JSON: https://castlelabs.io/data/revenue.json) In the 30 days to 14 November 2026, Tron collected the most blockspace fees of any blockchain ($24.5M), ahead of Ethereum ($19.8M) and Solana ($16.2M). Among protocols, Tether earned the most revenue ($512.9M), followed by Circle ($204.8M), Hyperliquid ($69.3M) and pump.fun ($48.0M). - [Hyperliquid Stats](https://castlelabs.io/research/hyperliquid): Hyperliquid's volume, open interest, market share and revenue, how it compares with other perp DEXs, and Castle's research on it (JSON: https://castlelabs.io/data/perps.json) - [Hyperliquid Revenue](https://castlelabs.io/research/hyperliquid-revenue): Hyperliquid's daily, weekly, monthly and 12-month revenue, its sources and the HIP-3 share (JSON: https://castlelabs.io/data/revenue.json) Hyperliquid earned $69.3M in revenue in the 30 days to 14 November 2026, about $2.3M a day, and $679.3M over the last 12 months. That ranks it number 3 among crypto protocols by revenue, behind Tether and Circle. - [Bitcoin $100K Odds](https://castlelabs.io/research/bitcoin-100k-odds): The chance Bitcoin reaches $100K by year-end, from Deribit options and Polymarket, with the implied price range (JSON: https://castlelabs.io/data/report-charts/renaissance-of-onchain-options.json) As of 14 November 2026, Deribit options imply a 1.3% chance that Bitcoin trades at $100K at some point before 25 December 2026, and a 0.4% chance it is above $100K on that date. Polymarket prices a 3.1% chance that Bitcoin reaches $100K in 2026. The options market's median for Bitcoin on 25 December 2026 is $67.3K, against $67.2K today, with a 10% chance it is below $56K and a 10% chance it is above $79.5K. These are market prices, not forecasts. - [Tokenized Stocks](https://castlelabs.io/research/tokenized-stocks): The largest tokenized stocks by market value, by company, issuer, chain and venue, plus TradeXYZ stock perps: xStocks, Ondo, bStocks, Robinhood and more (JSON: https://castlelabs.io/data/tokenized-stocks.json) On 14 November 2026, tokenized stocks were worth $2.84B across 1664 tokens tracked by CoinGecko, with $1.01B traded in the last 24 hours. The largest single token is SpaceX (bStocks Tokenized Stock) ($165.8M). By company, Circle is the most tokenized stock ($336.0M across 8 issuers), followed by NVIDIA ($329.5M) and SpaceX ($307.0M). By issuer, bStocks ($1.16B), xStocks ($1.00B) and Ondo Global Markets ($972.3M) lead. The most traded tokens in the last 24 hours were NVIDIA (bStocks Tokenized Stock) ($161.2M), SpaceX (bStocks Tokenized Stock) ($75.4M) and SanDisk (bStocks Tokenized Stock) ($38.6M), and bStocks traded the most of any issuer ($607.7M). - [Institutional Vaults and Curators](https://castlelabs.io/research/institutional-vaults): DeFi risk curators by assets managed, how much of each large Morpho vault is withdrawable now, redemption rules and custody routes (JSON: https://castlelabs.io/data/institutional-vaults.json) - [Morpho Stats](https://castlelabs.io/research/morpho): Morpho's deposits, loans, interest paid and chains, its largest vault curators, and Castle's research on it (JSON: https://castlelabs.io/data/morpho.json) - [Stablecoin Chains](https://castlelabs.io/research/stablecoin-chains): Stablecoin supply by blockchain, 30-day growth, the USDT and USDC mix, and where Arc, Plasma and Tempo stand (JSON: https://castlelabs.io/data/stablecoin-chains.json) - [RWA Yields Beyond Treasuries](https://castlelabs.io/research/rwa-yields): Reinsurance, private credit and other RWA yield tokens by APY and size, against SOFR and the 13-week T-bill (JSON: https://castlelabs.io/data/rwa-yields.json) - [Onchain Commodities](https://castlelabs.io/research/onchain-commodities): Oil, gold and silver perps on Hyperliquid by volume, open interest and funding, with the largest tokenized gold tokens (JSON: https://castlelabs.io/data/commodities.json) - [Pre-IPO Markets Onchain](https://castlelabs.io/research/pre-ipo-markets): Pre-IPO perps on Hyperliquid by volume, open interest, funding and implied valuation, with PreStocks tokens (JSON: https://castlelabs.io/data/pre-ipo.json) - [Crypto Cards Compared](https://castlelabs.io/research/crypto-cards): Crypto cards ranked by monthly settled spend from Paymentscan, with custody, card network, cashback and FX fees for the largest (JSON: https://castlelabs.io/data/crypto-cards.json) All datasets with their sources, update dates and how to cite them: https://castlelabs.io/research/data ## Glossary Short, sourced definitions of the mechanisms Castle covers: https://castlelabs.io/research/glossary - [Perpetual futures](https://castlelabs.io/research/glossary#perpetual-futures): A futures contract with no expiry date. Instead of settling on a set day, longs and shorts pay each other a funding rate that keeps the contract's price close to the spot price. - [Funding rate](https://castlelabs.io/research/glossary/funding-rates): The regular payment between longs and shorts on a perpetual futures contract. When the perp trades above the spot price longs pay shorts, and when it trades below shorts pay longs, which pulls the two prices together. - [Open interest](https://castlelabs.io/research/glossary#open-interest): The total value of derivatives positions that are still open. Volume counts every trade, while open interest counts what is still at stake, so it is less affected by traders churning the same position. - [HIP-3](https://castlelabs.io/research/glossary/hip-3): Hyperliquid's standard that lets a team that stakes HYPE launch its own perpetual futures exchange on Hyperliquid's order book and matching engine. Deployers choose the markets, oracle and settings, and share the trading fees. - [Pre-IPO perps](https://castlelabs.io/research/glossary#pre-ipo-perps): Cash-settled perpetual contracts on a private company's expected value. No shares change hands: traders go long or short the company's valuation with margin, unlike pre-IPO tokens, which are issued against shares held in a fund. - [Vault curator](https://castlelabs.io/research/glossary#vault-curators): The firm that sets a lending vault's risk rules: which markets it can lend into, the cap on each and its fees. Curators are paid a share of the yield; the largest include Steakhouse Financial, Sentora and Gauntlet. - [Withdrawable liquidity](https://castlelabs.io/research/glossary#withdrawable-liquidity): The share of a lending vault's assets that depositors could take out right now: idle cash plus what the vault can pull back from the markets it lends into. The rest is lent out and comes back only as borrowers repay. - [Fixed-rate lending](https://castlelabs.io/research/glossary#fixed-rate-lending): Loans whose interest rate is set when the loan opens and stays the same until it matures. Almost all onchain lending is variable-rate: Castle's August 2026 report found $28.5 billion in active DeFi loans, nearly all variable, with Fira the largest fixed-rate market at $430 million. - [Fees and revenue](https://castlelabs.io/research/glossary#fees-and-revenue): Fees are what users pay to use a protocol or chain. Revenue is the part the protocol keeps after paying liquidity providers, lenders and other suppliers; for a blockchain, DefiLlama counts as revenue only the fees that are burned or kept by the chain. - [Holders revenue](https://castlelabs.io/research/glossary#holders-revenue): The part of a protocol's revenue that reaches its token holders, through buybacks, burns or payouts to stakers. DefiLlama counts only payouts funded from revenue, not buybacks paid from a treasury. - [Holder yield](https://castlelabs.io/research/glossary#holder-yield): A year of holder payouts at the last 30 days' pace, divided by the token's market cap. It compares what protocols return to holders relative to their size; it is not a cash yield, because buybacks and burns reach holders only through the price. - [Delta-neutral trade](https://castlelabs.io/research/glossary#delta-neutral): Holding an asset and an equal short perpetual position, so price moves cancel out and the funding rate is the return. Ethena's USDe stablecoin is built on this trade. - [Perpetual options](https://castlelabs.io/research/glossary#perpetual-options): Options with no fixed expiry, paid for with a streaming premium for as long as the position stays open. Panoptic builds them from AMM liquidity; Castle's 2026 options research found these designs still mostly commercially unproven. - [Tokenised stocks](https://castlelabs.io/research/glossary#tokenized-stocks): Tokens that track the price of a listed share, usually issued by a company that holds the share or a claim on it. Holders typically get price exposure rather than shareholder rights such as votes, and the terms differ by issuer. - [Real-world assets (RWAs)](https://castlelabs.io/research/glossary#real-world-assets): Tokens that represent assets that exist off-chain, such as Treasury bills, private credit, commodities or reinsurance. Castle's June 2026 report put tokenised RWAs at $28.2 billion, with only about 10% of that active in DeFi. - [Shielded supply (Zcash)](https://castlelabs.io/research/glossary#shielded-supply): ZEC held in Zcash's shielded pools, where amounts and addresses are hidden, rather than in transparent addresses. About 4.9 million ZEC, close to 30% of supply, was shielded when Castle's June 2026 privacy report was written. ## Research Articles - [Re: Reinsurance as Uncorrelated RWA Yield](https://castlelabs.io/research/re-reinsurance-as-uncorrelated-rwa-66e): Re is bringing an interesting alternative to RWA yield, one that is less correlated to markets and instead is based on real reinsurance activity. (2026-11-04; by Castle Labs; tags: RWAs, DeFi, L2s, Markets) - [Metric: The Next-Gen DEX](https://castlelabs.io/research/metric-the-next-gen-dex): Liquidity Providers (LPs) on Decentralised Exchanges (DEXs) have long faced a structural problem during market movements: as prices move on Centralised Exchanges (CEXs), which are hubs for spot liquidity and perform price discovery, DEXs still quote prices before the market moves, giving arbitrageur (2026-10-26; by Noveleader; tags: Deep Dives, Trading, DeFi, Markets) - [Re and the Case for Reinsurance Yield Onchain](https://castlelabs.io/research/re-and-the-case-for-reinsurance-yield): Reinsurance as onchain RWA yield: how Re's reUSD and reUSDe earn SOFR or sUSDe plus 250 and 850 bps spreads, beyond T-bill products, and the risks. (2026-10-21; by Francesco; tags: RWAs, DeFi) Short answer: Reinsurance offers yield largely uncorrelated with T-bills, private credit and crypto funding rates, and Re brings it onchain by using depositors' stablecoins to underwrite regulated reinsurance contracts. As of 20 October 2026, it reported over $375 million in deposits and $12 million in cumulative yield paid. Senior reUSD earns SOFR on offchain capital, or the 7-day trailing sUSDe yield on its onchain buffer, plus a 250 bps spread; mezzanine reUSDe takes losses after Re's own first-loss capital and earns SOFR plus 850 bps, with quarterly redemptions planned. Risks we flag include liquidations in DeFi markets built on reUSD and the reliance on sUSDe for onchain liquidity. - [Jumper: The Onchain Finance Super-App](https://castlelabs.io/research/jumper-the-onchain-finance-super-a50): Marco, CEO of Jumper, joins Tom at Castle Labs to discuss what comes after bridge aggregation: why cross-chain movement is becoming invisible, how Jumper is separating from LI.FI, and whether one interface can become the front end for onchain finance. (2026-10-16; by Castle Labs; tags: Markets) - [Solana: Building the Full-Stack Onchain Economy](https://castlelabs.io/research/solana-building-the-full-stack-onchain): Solana is now the top chain by application revenue. (2026-10-15; by Noveleader; tags: Deep Dives, Markets) - [Pearl: The Missing Piece for AI?](https://castlelabs.io/research/pearl-the-missing-piece-for-ai): Artificial Intelligence (AI) is a pivotal technology, and as its use grows, it demands more computational throughput than ever. (2026-10-12; by Noveleader; tags: Deep Dives, AI, Markets) Short answer: Pearl is an L1 forked from Bitcoin's code that swaps SHA-256 hashing for matrix multiplication, the operation GPUs run for AI inference, so the same work mines PRL and can serve AI customers. We see a valuable proposition for Bitcoin miners already selling AI inference, but the token has no revenue or burn, since customers pay for inference in USD. The INT8 format Pearl runs today slows AI models by 10-15%, against 0-3% for Pearl-GEMM on testnet. At the time of writing, PRL had a $475 million market cap, with 339 million of its 2.1 billion hard cap mined and about 305 million more due over the next 12 months. Key takeaways: - Pearl, an L1 blockchain launched in April and forked from Bitcoin's code, replaces SHA-256 hashing with matrix multiplication, the operation GPUs run for AI inference. - At the time of writing, PRL had a $475 million market cap and 339 million tokens mined, with about 305 million more due over the next 12 months. - Pearl's current INT8 number format slows AI models by an estimated 10 to 15%, while Pearl-GEMM, live on testnet, handles the standard format with a 0 to 3% slowdown. - Critics note that Pearl does not prove its matrix multiplications serve paying customers, and a June paper found the dominant mining software produced verifiable MatMul but no external inference output. - Castle Labs notes that PRL has no revenue attached and no burn mechanism, as miners earn stablecoins for AI inference and Pearl serves models in exchange for USD. - [Onchain Perps: A Study of the Evolution of Design Choices and Growth Catalysts](https://castlelabs.io/research/onchain-perps-a-study-of-the-evolution): This excerpt is from our research on Mapping the Perpetual Trading Ecosystem: A Comparative Analysis of Perp Trading Venues across Design, GTM and Microstructure, highlighting the perp ecosystem, including the key growth enablers, the architecture on which the protocols are built, and the GTM strate (2026-10-08; by Noveleader; tags: Deep Dives, Trading) Short answer: Onchain perps moved from early designs such as dYdX's Cosmos appchain and GMX's GLP pool, where liquidity providers absorbed traders' PnL, to venues like Hyperliquid, Variational and Lighter. When we published in October 2026, DEXs handled over 13% of perp volume, up from 4% in 2024, and Hyperliquid held over 56% of onchain open interest, helped by HIP-3 markets, builder codes and buybacks that take 95% of protocol fees. RWAs are the newest catalyst, and on 10 September 2026 seven of the top 10 perp assets were RWAs, led by the Nasdaq 100 and gold. - [M0: Building the Internet Dollar with Laert Skreli](https://castlelabs.io/research/m0-building-the-internet-dollar-with-499): Stablecoins are becoming app-specific dollars. (2026-10-07; by Castle Labs; tags: DeFi, Markets) - [On Raydium, Solana, Stocks, and Memes](https://castlelabs.io/research/on-raydium-solana-stocks-and-memes-7aa): Raydium has been one of Solana's key protocols; its infrastructure is now fundamental to tokenised stocks, which have taken over Solana. (2026-09-30; by Castle Labs; tags: RWAs, Markets) - [Putting Tokenised Stocks to Work](https://castlelabs.io/research/putting-tokenised-stocks-to-work-3ea): Tokenised stocks have crossed $3B, but most of that capital still sits idle. (2026-09-26; by Castle Labs; tags: RWAs, Markets) - [Mapping Onchain Options: Exchanges, Binaries and Yield Products](https://castlelabs.io/research/mapping-onchain-options-exchanges): Onchain options in three families: Derive and Paradex order books, binary markets on Polymarket, Kalshi and HIP-4, and option yield products from Rysk and Hegic. (2026-09-24; by Francesco; tags: Deep Dives, Options, Trading, DeFi) Short answer: We split onchain options into at least three live product families, each serving different users. Exchange-style venues such as Derive and Paradex let traders build custom payoffs through order books and RFQs, and Derive had recorded its biggest day, over $500m in notional, when we published in September 2026. Polymarket, Kalshi and Hyperliquid's HIP-4 offer binary options, since a prediction market contract pays a fixed amount if a condition is met at expiry. Rysk and Hegic package options as yield, and Rysk's covered calls for HYPE holders reached $50 million in TVL even though monthly expiries reset it to zero. - [Inside EtherFi: Growing a Neobank with Mike Silagadze](https://castlelabs.io/research/inside-etherfi-growing-a-neobank-0fe): EtherFi has been one of the fastest-growing neobanks for both crypto natives and retail users. (2026-09-18; by Castle Labs; tags: DeFi, Markets) - [Beyond Tokenisation: Making RWAs Useful Onchain](https://castlelabs.io/research/beyond-tokenisation-making-rwas-useful): These days, it’s all about RWAs, with a race involving every protocol and ecosystem, rushing to list new assets and boasting about their market size or asset count. (2026-09-14; by Francesco; tags: Deep Dives, RWAs, Markets) Short answer: RWAs become useful onchain when holders can trade them against deep liquidity, post them as collateral, move them across venues and earn yield on them, since anyone can now list assets. When we published in September 2026, over $38 billion of assets were tokenised, led by US Treasuries at over $15.9 billion. Mantle, our case study, hosted over $225 million of RWAs, 96% of it in yield-bearing assets. It launched xStocks with only 10 assets on its DEX Fluxion so AMM liquidity could deepen, added RFQ execution through xChange, and distributes its assets through Bybit, which has over 40 million users. - [The Evolution of RWA Trading Onchain: Variational Swaps](https://castlelabs.io/research/the-evolution-of-rwa-trading-onchain): RWA perp volume went from under $1 billion in October 2025 to over $120 billion in August. How Variational swaps compare with perps on funding and execution cost. (2026-09-10; by Francesco, Noveleader; tags: Deep Dives, Trading, RWAs) Short answer: Variational swaps give traders exposure to RWAs such as the US100 index with liquidity sourced from TradFi and predictable funding, instead of the floating funding rates that make RWA perps expensive to hold. As of 7 September 2026, swaps drove over 50% of Variational's daily volume and over $220 million in open interest, with longs on the first markets paying 4.6% to 5.7% a year and shorts earning around 2.4%. - [Fixed Rate Lending is coming to DeFi with Bitwise, KPK and Tenor Finance](https://castlelabs.io/research/fixed-rate-lending-is-coming-to-defi-c56): Today, the lending category has $28.5 billion in active loans, and almost all of that demand comes from variable-rate lending. (2026-09-09; by Castle Labs; tags: DeFi, Markets) - [Solving the Cold-Start Problem in Fixed-Rate Lending](https://castlelabs.io/research/solving-the-cold-start-problem-in): This excerpt is from our research on Fixed-Rate Lending: Market Structure and Protocol Design, highlighting new and emerging designs in the category. (2026-09-07; by Noveleader; tags: Deep Dives, DeFi, Markets) Short answer: New fixed-rate lending protocols solve the cold-start problem by letting lenders post offers without locking capital, so their funds keep earning until a match executes. Morpho Midnight sources capital only at settlement and lets one maker quote across several markets, Jupiter Offerbook leaves funds free until an order fills, and Kamino keeps unmatched capital earning in its variable-rate reserves. When we published in September 2026, Midnight had $3 million in active loans and Offerbook $450,000, while DeFi lending as a whole had $28.5 billion in active loans, almost all at variable rates. - [The Expanding Business Model of Blockchains](https://castlelabs.io/research/the-expanding-business-model-of-blockchains): Blockspace is cheap and easy to copy, so chains are going after ecosystem and product revenue. Arbitrum, Polygon, MegaETH and Sophon show the two routes. (2026-08-13; by Noveleader; tags: L2s, Deep Dives, Markets) Short answer: Chain fees keep falling while application fees grow, so chains are trying to capture more of the value their ecosystems produce. We see two routes. In ecosystem expansion, Arbitrum licenses its stack to chains such as Robinhood's and Polygon repositions as a payments chain. In product expansion, MegaETH and Sophon build applications in-house. Optimism shows the risk of the first route: Base produced over 90% of Superchain revenue before it left in February 2026. Key takeaways: - Optimism's Superchain charged OP Stack L2s the greater of 15% of net onchain profits or 2.5% of revenue, and Base supplied over 90% of Superchain revenue before leaving in February. - Robinhood's L2, launched on Arbitrum Stack, had generated about $4 million in revenue at the time of writing, sending about $390k to Arbitrum under a 90/10 split. - Polygon, used by Stripe and Mastercard, had processed about $2.9 trillion in stablecoin volume at the time of writing, but most of its revenue still came from Polymarket. - MegaETH's USDm, whose BUIDL-backed yield funds buybacks and burns, had $18 million in supply at the time of writing, over 95% below its May peak of about $600 million. - Sophon shut down its chain after failing to find adoption and turned to building applications, starting with Pyre, a crypto card. - [Interviewing the CEO of Variational: A Different Model for Onchain Trading](https://castlelabs.io/research/interviewing-the-ceo-of-variational-164): Perps are one of crypto’s largest markets, but still carry trade-offs around liquidity and liquidation handling. (2026-08-09; by Castle Labs; tags: Trading, DeFi, Markets) - [The Birth of Bitcoin-Native Finance: Turning Idle BTC into Productive Capital](https://castlelabs.io/research/the-birth-of-bitcoin-native-finance): Only ~1.5% of active BTC earns yield. How CeFi lending, DeFi and Bitcoin L2s compare on trust, and how Stacks builds BTC-native finance with sBTC and stBTC. (2026-08-07; by Noveleader; tags: Deep Dives, Markets) Short answer: When we published in August 2026, only 311k of 20.05 million active BTC (~1.5%) had access to any yield, and every route to that yield added a trust assumption. Stacks is one of the most complete attempts at Bitcoin-native finance. Its sBTC peg wallet needs 11 of 15 signers to move funds, and the proposed PoX-5 upgrade lets holders earn BTC yield, targeting 3% APY, while their BTC stays timelocked on L1. Key takeaways: - Only 311k of the active supply of 20.05 million BTC, about 1.5%, had access to any yield at the time of writing, via DeFi lending, staking, DEXs or vaults. - Babylon, Lombard and Solv held TVL of $2.6 billion, $800 million and $480 million respectively at the time of writing, with Lombard's LBTC sourcing its yield from Babylon staking. - Stacks' sBTC bridge keeps all bridged BTC in a single Taproot peg wallet that a 15-member, community-elected signer set can move only with an 11-of-15 threshold. - Stacks' proposed PoX-5 upgrade lets holders earn BTC yield while their BTC stays locked on Bitcoin L1, with six-month BTC-STX bonds forming a senior tranche that targets 3% APY. - Stacks' consensus mechanism, in which miners bid BTC to mine Stacks blocks, has distributed over 4200 BTC to STX stakers since January 2021. - [RWAs: Bringing TradFi Onchain](https://castlelabs.io/research/rwas-bringing-tradfi-onchain-d53): RWAs have become the fastest-growing category in crypto since 2024 (2026-08-04; by Castle Labs; tags: RWAs, Markets) - [The Broken Link between Protocol Revenues and Token Performance](https://castlelabs.io/research/the-broken-link-between-protocol): Crypto protocols made $7.42 billion in revenue so far in 2026, yet PUMP is down 60% since launch despite $315 million in buybacks, and HYPE is up 1,400%. (2026-07-29; by Noveleader; tags: Deep Dives, Markets) Short answer: When we published in July 2026, crypto protocols had made $7.42 billion in revenue since the start of the year, but most tokens did not reflect it. Emissions, unlocks, treasury retention and equity-token splits weaken the link between revenue and price. PUMP was down 60% since launch despite over $315 million in buybacks. HYPE was up 1,400% after returning $1.2 billion through buybacks. Key takeaways: - Among six protocols that made a combined $726 million in revenue in H1 2026, Aerodrome, Sky and Uniswap showed negative net token flow once emissions were subtracted from holder revenue. - At the time of writing, PUMP was down 60% since launch despite over $315 million in buybacks, while HYPE was up 1,400% and had returned $1.2 billion through buybacks. - At the time of writing, Hyperliquid had burnt over 47 million HYPE (about 4.72% of supply) and Uniswap 107 million UNI (about 11% of supply). - Aave bought back $45 million of AAVE since April 2025 at an average $182, losing over $23 million with the token near $90 at the time of writing. - Ripple Labs stock is up 105% since 2025 while XRP is down 45% over the same period, which Castle Labs attributes to tokenholders having no rights to Ripple's revenue. - [The New Security Bar for Onchain Finance](https://castlelabs.io/research/the-new-security-bar-for-onchain): Public balances and quantum-vulnerable signatures are the two security gaps in onchain finance. How Starknet's strkBTC handles privacy, compliance and post-quantum risk. (2026-07-20; by Castle Labs; tags: Privacy, Deep Dives, Markets) Short answer: Onchain finance has two open security problems: every balance and transfer is public, and the elliptic-curve signatures behind Bitcoin and Ethereum wallets could be broken by a large enough quantum computer. Starknet's strkBTC, a BTC-backed ERC-20 built on the STRK20 privacy framework, shields balances while sharing a viewing key with an independent compliance operator. Starknet's STARK proofs rely on hash functions rather than elliptic curves, and its native account abstraction lets users move to post-quantum signatures with a wallet upgrade instead of a hard fork. The bridge federation and viewing-key disclosure rules have yet to be tested at scale. Key takeaways: - CertiK reported 34 verified physical attacks on crypto holders in the first four months of 2026, up 41% year-on-year, and France saw 41 crypto-linked kidnappings from January to May 2026. - Google Quantum AI's March report estimated roughly $100 billion at risk across the Ethereum ecosystem and singled out Starknet as better placed than other major L2s for post-quantum security. - Starknet's proving layer uses hash-based STARK proofs, and its native account abstraction lets users adopt post-quantum signatures via a wallet upgrade instead of a hard fork. - When users shield strkBTC, a BTC-backed ERC-20 built on StarkWare's STRK20 privacy framework, a viewing key goes to Financial Privacy Inc for scoped access on valid regulatory requests. - The strkBTC bridge is run by a Federation of Twinstake, NEAR Intents, Luganodes, UTXO Management and Xverse, which Starknet aims to move beyond toward Bitcoin-native verification and BitVM. - [Prediction markets as option-like instruments](https://castlelabs.io/research/prediction-markets-as-option-like): Prediction markets work like binary options, down to 300-second expiries. How Polymarket, Kalshi and Hyperliquid binaries compare as option-like instruments. (2026-07-17; by Noveleader; tags: Options, Deep Dives, Trading, Markets, Predictions) Short answer: A prediction market contract on an asset price works like a binary option, settling at $1 if the condition is met at expiry and $0 otherwise. Polymarket and Kalshi list BTC markets that expire in as little as five minutes. In our comparison of BTC 1-day markets from 22 to 28 June 2026, Kalshi led on volume with $29.6 million, ahead of Polymarket at $19.3 million and Hyperliquid's HIP-4 at $2.8 million. Hyperliquid also had the widest and most volatile spreads, at 50 to 587 bps, against a steadier 107 to 154 bps on Kalshi. - [Putting RWAs to work in DeFi](https://castlelabs.io/research/putting-rwas-to-work-in-defi): The growth of RWAs has led to over $32 billion of assets flowing onchain. (2026-07-13; by Francesco; tags: RWAs, DeFi, Markets) - [Options from a Builder’s Perspective](https://castlelabs.io/research/options-from-a-builders-perspective): A Gap Analysis with Interviews from Rysk, Derive, Panoptic, and GammaSwap. (2026-07-09; by Castle Labs; tags: Options, Markets) - [The Rebirth of Onchain Options: An Ecosystem View](https://castlelabs.io/research/the-rebirth-of-onchain-options-an): Options traded over 4x the contracts of futures in 2024. How onchain options moved past DOVs to order books, AMMs, perpetual options and short-dated binaries. (2026-07-06; by Castle Labs; tags: Options, Deep Dives, Trading, Markets, Predictions) Short answer: Onchain options are recovering after a first cycle of option vaults and options AMMs that left passive LPs holding hard-to-hedge risk. When we published in July 2026, onchain options did about $1.44B in 30-day notional volume, and the leading venues had moved to order books and RFQs. Derive, which grew out of Lyra, handled 79.2% of that notional, Rysk 9.5% through covered calls and cash-secured puts, and Aevo 3.1%. Newer designs such as Panoptic's perpetual options and Euphoria's five-second touch markets remain mostly commercially unproven. - [Prediction Markets: From Curated Platforms to Open Protocols](https://castlelabs.io/research/prediction-markets-from-curated-platforms): Prediction markets grew from about $2 billion to over $30 billion in monthly volume. Why curated Polymarket and Kalshi won, and the open protocols now testing them. (2026-07-02; by Noveleader; tags: Markets, Predictions, Deep Dives) Short answer: Prediction markets grew from roughly $2 billion to more than $30 billion in monthly volume in the 18 months before we published in July 2026, and almost all of that growth came from curated platforms such as Polymarket and Kalshi, whose teams decide which markets get listed. Early permissionless venues such as Augur and Omen failed on duplicate, illiquid markets, slow or disputed resolution, weak creator economics and regulation. A new wave of protocols is retrying open creation. HIP-4 requires deployers to stake 1 million HYPE, and Limitless lets creators launch templated crypto price markets and receive 50% of the fees they generate. - [Composable TradFi: RWAs in DeFi](https://castlelabs.io/research/composable-tradfi-rwas-in-defi): Only about $3 billion of the $28 billion RWA market is used in DeFi. How Pendle, Morpho, Aave, Kamino and Fluid use RWAs, and the new redemption tools. (2026-06-29; by Noveleader; tags: RWAs, DeFi, Markets) Short answer: When we published in June 2026, only about $3 billion of the $28 billion RWA market was used in DeFi, mostly as collateral that holders loop on lending protocols. Morpho led with over $1 billion in RWA deposits, ahead of Aave Horizon (about $500 million), Kamino ($400 million) and Fluid ($270 million). T+1 to T+3 redemption delays limit growth, so new redemption tools have a third party front the cash and carry the duration risk for a fee. Upshift Clear gives T+0 USDC liquidity for a 50 bps fee, against 564 bps at Fission, and Grove's Basin offers up to $1 billion a day. - [Privacy Ecosystem on Solana](https://castlelabs.io/research/privacy-ecosystem-on-solana): How Solana privacy works across Arcium, MagicBlock, Umbra, Privacy Cash and Hush, with Arcium past 900k computations since February 2026. (2026-06-15; by Francesco, Noveleader; tags: Privacy, Deep Dives) Short answer: When we published in June 2026, privacy on Solana was still early, and Mert told us the chain was lagging a bit. Arcium had processed over 900k computations and 3.5 million transactions since its Alpha mainnet launch in early February 2026. For private transfers, Privacy Cash runs a Tornado-style shielded pool for SOL, and Hush converts deposits to jitoSOL inside its pool. Private swaps run through encrypt.trade and Vanish. - [propAMMs: Better than CEX?](https://castlelabs.io/research/propamms-better-than-cex): How propAMMs differ from passive x*y=k AMMs, why their fills on Solana landed 0.33 to 1.36 bps from the best CEX mid, and how the model is growing on Ethereum. (2026-06-12; by Francesco; tags: L2s, Deep Dives, Trading, Markets) Short answer: PropAMMs beat CEXs on small trades, and when we published in June 2026 their spreads on $1k ETH/USDT and ETH/USDC orders were 20 bps lower than CEX counterparts, while Binance led at $10k and above. A propAMM is an onchain contract that quotes its own bids and offers from an offchain pricing engine watching CEX order books, whereas passive AMMs price off an x*y=k curve. Jump's March 2026 analysis of Solana propAMMs found most fills landed 0.33 to 1.36 bps from the best CEX mid. The risk is opaque quoting, and propAMMs on Base have repriced quotes between display and settlement, costing takers 5 to 10 bps on average. - [Stacks POX Upgrade: Solving BTC Capital Problem with the Right Incentives](https://castlelabs.io/research/stacks-pox-upgrade-solving-btc-capital): Over the last couple of years, since the launch of Bitcoin ETFs, BTC has become an increasingly institutionalised and legitimised asse. (2026-06-10; by Francesco; tags: Deep Dives, L2s, Markets) - [RWAs: A Blueprint for the Future of Mantle](https://castlelabs.io/research/rwas-a-blueprint-for-the-future-of): Tokenised stocks are among the fastest-growing RWAs, with $1.6 billion distributed. How Mantle and Bybit built issuance, trading and redemption onchain. (2026-05-28; by Francesco; tags: RWAs, Deep Dives, Markets, DeFi, L2s) - [Vaults: The Infrastructure Layer for Institutional Finance Onchain](https://castlelabs.io/research/vaults-the-infrastructure-layer-for): Eight onchain vault categories, TVL by curator, and how lending, restaking, RWA and options vaults held up through the Stream, Resolv and Kelp blowups. (2026-05-25; by Francesco, Noveleader; tags: Deep Dives, DeFi, Markets) Short answer: Vault TVL across our eight categories stood at $120.4 billion, roughly half its $241 billion peak from October 2025. We used the Stream Finance blowup, the Resolv exploit and the Kelp hack, between October 2025 and May 2026, as a stress-test window: lending, liquid staking, restaking and risk-curated vaults all suffered heavy drawdowns, while RWA vaults grew 37.8%. For redemption risk, the report covers asynchronous ERC-7540 vaults and instant-redemption products such as Upshift Clear. Key takeaways: - At the time of writing, net TVL across Castle Labs' eight vault categories was $120.4 billion, down about 50% from its $241 billion peak around October 2025. - RWA vaults gained 37.8% between October 2025 and May 2026, a window covering the Stream Finance, Resolv and Kelp incidents, while most other vault categories suffered significant drawdowns. - After the KelpDAO attack, Aave lost over $12 billion in TVL and its TVL ratio to Morpho fell below 2x from a previous 5 to 6x range. - Risk-curated vaults held about $6.5 billion at the time of writing, 75% of it with Sentora ($1.85 billion), Steakhouse ($1.63 billion) and Gauntlet ($1.5 billion). - For RWA redemption delays, Centrifuge uses ERC-7540 vaults with asynchronous redemptions, and Upshift is releasing Upshift Clear with Superstate for instant RWA-to-USDC swaps at a 5bps fee. - [Pre-IPO Markets are Moving Onchain](https://castlelabs.io/research/pre-ipo-markets-are-moving-onchain): 2026 is one of the most eventful years for IPOs, with the biggest IPO calendar in a generation. (2026-05-19; by Castle Labs; tags: Deep Dives, Markets) Short answer: Pre-IPO trading is moving onchain through pre-IPO perpetuals (IPOPs), synthetic contracts that track a private company's implied valuation without any shares changing hands. TradeXYZ's first IPOP on Hyperliquid, xyz:CBRS for Cerebras, opened on 1 May 2026 at a $175 reference price, traded $281 million over 14 days and priced within 6% of the $350 Nasdaq open, while Hiive valued the stock at $220 that morning. Because no shares move, an issuer has nothing to void. Tokenised SPVs rely on share transfers, and PreStocks tokens tied to Anthropic and OpenAI fell roughly 40% after Anthropic declared unauthorised secondary transfers of its shares void in May 2026. Key takeaways: - TradeXYZ's first pre-IPO perpetual, xyz:CBRS for Cerebras, launched on 1 May at a $175 reference price and traded $281 million of cumulative notional over its 14-day pre-IPO window. - On Cerebras's listing morning, Hiive valued the stock at $220 while TradeXYZ priced it at about $290, then about $340 one hour before the $350 Nasdaq open. - PreStocks tokens tied to Anthropic and OpenAI fell roughly 40% after Anthropic declared unauthorised transfers of its shares void, a risk pre-IPO perpetuals avoid because no shares change hands. - Trade XYZ's 0.005 funding multiplier left xyz:CBRS averaging +1.67% APR over its 14-day window, against 14-day medians of +31% to +64% on Ventuals' OpenAI, Anthropic and SpaceX markets. - As of 18 May 2026, Trade XYZ accounted for about 93% of Hyperliquid HIP-3 daily volume and about 95% of HIP-3 open interest across 77 markets. - [MegaEth and the Bet on Consumer Apps](https://castlelabs.io/research/megaeth-and-the-bet-on-consumer-apps): For the last cycle, crypto has been strengthening its infrastructure, but the core audience remained the same. (2026-04-30; by Noveleader; tags: Deep Dives, L2s, Markets) - [The Beginning of Agentic Finance](https://castlelabs.io/research/the-beginning-of-agentic-finance): How AI agents pay onchain via Ethereum's agentic stack: x402 payments, ERC-8004 trust and ERC-8183 commerce, why most x402 volume runs on Base, and key risks. (2026-04-23; by TradFiHater, Noveleader; tags: AI, Markets) Short answer: AI agents pay onchain through three EVM standards. x402 lets an agent pay a server in stablecoins within one HTTP request, ERC-8004 gives each agent an onchain identity and reputation record, and ERC-8183 lets a client lock funds in escrow until an evaluator accepts the work. When we published in April 2026, x402 had processed over 50 million transactions since its May 2025 launch and ERC-8004 had about 98,000 agent registrations, with most x402 volume on Base because micropayments need sub-cent fees. The main risks we flag are public-by-default transactions and prompt injection, such as a poisoned ENS record telling an agent to send funds to an attacker. - [From Neobanks to Onchain Banks](https://castlelabs.io/research/from-neobanks-to-onchain-banks): How crypto cards and onchain banks work, from Visa's $3.5B stablecoin card run rate to ether.fi Cash, Tria, Superform and Plasma One, and the regulatory gaps. (2026-04-13; by TradFiHater; tags: DeFi, Markets) Short answer: Onchain banks keep user funds in smart contract wallets, earn yield through DeFi and pay through Visa or Mastercard via issuers such as Rain, whereas neobanks hold deposits on their own balance sheet. In the standard model, the card issuer sells the user's crypto for fiat at the moment of payment, and Visa's stablecoin-linked card spend reached a $3.5 billion annualised run rate in 2025. Ether.fi Cash lets users borrow against staked ETH, Tria routes card spending across chains, Superform aggregates vault yield and Plasma One runs on its own stablecoin L1. None offers deposit insurance or protocol-level chargebacks, and AML rules written for custodial intermediaries fit self-custodial wallets poorly. Key takeaways: - Crypto card spending reached a record $406 million in monthly volume as of November 2025, while Visa's stablecoin-linked card spend hit a $3.5 billion annualised run rate in 2025. - At the time of writing, Visa captured over 90% of onchain card transaction volume, after partnering early with full-stack issuers Rain and Reap, which hold direct Visa principal membership. - ether.fi Cash, whose Borrow Mode lets users spend stablecoins borrowed against ETH, accounts for approximately 50% of ether.fi revenue and had about 300,000 accounts at the time of writing. - Tria's card had more than $60 million in total volume, settling across Arbitrum, Base, Optimism, Polygon and Solana, with monthly card spend of roughly $12 million as of March 2026. - Crypto neobanks offer no FDIC insurance or protocol-level chargebacks, and at the time of writing less than 2% of total DeFi TVL was covered by any form of insurance. - [A New Standard for Privacy](https://castlelabs.io/research/a-new-standard-for-privacy): How StarkWare's STRK20 and strkBTC bring compliant privacy to ERC20 tokens on Starknet, vs Zcash, Railgun and Tornado Cash. Privacy coins rose 288% in 2025. (2026-04-09; by TradFiHater; tags: Privacy, Markets) - [A New Crypto Era: From Gambling to Investing](https://castlelabs.io/research/a-new-crypto-era-from-gambling-to): Why crypto is shifting from gambling to investing: only ~50 protocols earned $100k+ in holder revenue in 30 days. P/S ratios, buybacks and supply overhang. (2026-04-08; by Noveleader; tags: Markets) Short answer: Crypto is moving from gambling to investing, and we expect tokens with no link to protocol revenue to lose value. At the time of writing, only about 50 of around 5,700 protocols on DefiLlama did over $100k in holder revenue in the previous 30 days, under 1% of the total. Buybacks alone are not enough to raise a token's value. HYPE outperformed PUMP while both had active buybacks, with annualised holder revenue of about $660 million and $380 million, because of sentiment and the market pricing in future unlocks. Key takeaways: - Only about 50 of the roughly 5,700 protocols listed on Defillama earned over $100k in holder revenue in the 30 days before writing, under 1% of the total. - Crypto protocols generated about $16 billion in cumulative revenue in the calendar year before writing, with the top 10 taking 80% of net revenue and Tether alone 44%. - Using a P/S overvaluation threshold of 20 from top US public equities, Castle Labs found most top revenue protocols around or below it at the time of writing, except Tron. - HYPE outperformed PUMP while both had active buybacks, on annualised holder revenue of about $660 million and $380 million respectively at the time of writing. - Castle Labs argues buybacks alone do not lift token value, citing sentiment and vesting schedules, and finds PUMP releases supply fastest and HYPE slowest among the fixed-supply tokens it tracked. - [432 Hours of Hyperliquid Oil Market Data: A Microstructure Comparison with CME WTI Futures](https://castlelabs.io/research/hyperliquid-oil-perps-vs-cme-wti-futures): Hyperliquid's CL oil perp vs CME WTI futures over 432 hours: CME shows 125x more depth at 2 bps, and median trades differ 166x, $90,450 vs $543. (2026-04-01; by Noveleader, Francesco; tags: Trading, Markets) Short answer: Hyperliquid's xyz:CL oil perp is far thinner than CME WTI futures. Across 432 hours from 27 February to 16 March 2026, CME offered $19 million of depth within 2 bps of mid against $152,000 on Hyperliquid, a 125x gap, and its median trade was $90,450 against $543. A $1 million order slipped 0.79 bps on CME and 15.4 bps on Hyperliquid, though Hyperliquid's slippage at that size fell 86.9% over the three weeks while total volume grew from $31 million to over $1 billion. On weekends, with CME closed, Hyperliquid captured 45% of the Monday price gap after the Iran strikes and 68% a week later. - [69 Trillion: The Tokenisation Opportunity](https://castlelabs.io/research/69-trillion-the-tokenisation-opportunity): How tokenised stocks work across Ondo Global Markets, Kraken's xStocks and Hyperliquid perps, chasing a $69 trillion US equity market. Ondo holds $2.47b in TVL. (2026-03-09; by Castle Labs; tags: RWAs, DeFi, Markets) Short answer: Crypto users can reach the $69 trillion US stock market onchain through tokens backed by real shares, from Ondo Global Markets and Kraken's xStocks, or through Hyperliquid perpetuals that only track price. Ondo and xStocks hold the stock with custodians such as Alpaca and issue debt claims with no voting rights, but Ondo requires KYC and excludes US persons, while xStocks trade freely across chains. Hyperliquid's HIP-3 lets anyone staking 500,000 HYPE deploy stock perps, which carry oracle and liquidation risk. When we published in March 2026, Ondo held $2.47 billion in TVL across all products and xStocks held $250 million in public equities. Key takeaways: - Ondo, which held $2.47 billion in TVL across all products at the time of writing, issues tokenised stocks that are debt instruments collateralised by equities, with no voting rights. - Unlike Ondo, which requires KYC and excludes US persons, Kraken-owned xStocks has no KYC or whitelisting and held $250 million in public equities at the time of writing. - Alpaca, a US-headquartered self-clearing broker-dealer, custodied over 94% of all tokenised US equities and ETFs by value at the time of writing, including Ondo's. - Hyperliquid's HIP-3, activated in October 2025, lets anyone staking 500,000 HYPE launch their own perp exchange, with the deployer setting the oracle and earning 50% of fees. - Castle Labs argues Ondo and xStocks carry issuer and custodial risk, while Hyperliquid's equity perps carry oracle and liquidation risk, where liquidated capital cannot be recovered. - [See Through the PRISM](https://castlelabs.io/research/see-through-the-prism): How PRISM, OpenEden's multi-strategy yield token run by FalconX's Monarq, targets risk-adjusted returns with low crypto correlation, and its KYC trade-offs. (2026-03-05; by Francesco, Noveleader; tags: Markets) - [Inside Arbitrum DRIP Season 1](https://castlelabs.io/research/inside-arbitrum-drip-season-1): Arbitrum DRIP Season 1 spent 24M ARB on lending incentives for Aave, Morpho, Fluid, Euler, Dolomite and Silo. TVL fell 34% as ETH dropped. What worked? (2026-03-02; by Noveleader; tags: L2s, Markets) - [The Fight for Yield is Over: Making DeFi More Accessible](https://castlelabs.io/research/the-fight-for-yield-is-over-making): How INFINIT's Prompt-to-DeFi AI agents turn plain prompts into yield strategies like looping and LPing, now integrated with Binance, Gate and Bitget wallets. (2026-02-26; by Noveleader; tags: DeFi, Markets) - [Who Built Aave?](https://castlelabs.io/research/who-built-aave): Aave Labs vs ACI: the Aave Will Win proposal seeks $42.5M plus 75,000 AAVE, 31.5% of the treasury, amid a dispute over who built DeFi's largest lender. (2026-02-25; by Castle Labs; tags: Markets) - [Institutions Are Here, and This Is What They Want](https://castlelabs.io/research/what-they-do-in-the-dark): What institutions want onchain: tokenised treasuries like BlackRock BUIDL and Ondo OUSG near $25B, J.P. Morgan's Kinexys, and why ETH and SOL lag adoption. (2026-02-24; by TradFiHater; tags: Markets) Short answer: Institutions want yield, tokenised versions of the assets they already manage and cheap, always-on settlement, and they treat crypto tokens as speculative plays. When we published in February 2026, tokenised treasury products such as BlackRock's BUIDL, Ondo's OUSG and Franklin Templeton's funds paid around 4-5% and were worth close to $25 billion combined, while J.P. Morgan was building Kinexys, a bank-led blockchain for its clients. ETH and SOL have lagged this adoption, with Ethereum earning $526 million in fees across 2025, ETH below $2,000 and SOL down 36% year to date despite Solana's $2.39 billion in 2025 revenue. - [Most Crypto Assets Need to go to ZERO](https://castlelabs.io/research/most-crypto-assets-need-to-go-to): Crypto is going through its one of the discovery phases where speculative tokens will stop receiving new inflows. (2026-02-23; by Noveleader; tags: Markets) - [Hyperliquid is Not a Company](https://castlelabs.io/research/hyperliquid-is-not-a-company): Key takeaways from Jeff's interview on Hyperliquid: the HYPE airdrop, why there is no discretionary buyback program, HIP-3 markets and TradeXYZ's silver volume. (2026-02-19; by TradFiHater; tags: Trading, Markets) Short answer: Jeff, who built Hyperliquid, said Hyperliquid is not a company in a February 2026 interview on the Kevin WSH pod, explaining that it has no discretionary buyback program because fees are converted to HYPE and burned by the chain's own execution logic, with nobody deciding when to buy. Hyperliquid Labs has barely a dozen people because the protocol is a set of primitives anyone can use to tokenise anything, where a centralised exchange needs separate teams for tokenisation, liquidity and TradFi feeds. TradeXYZ, the first deployer of permissionless perps, reached roughly 2% of global silver volume within a few months. Key takeaways: - Jeff said on the Kevin WSH pod that Hyperliquid has no discretionary buyback programme, because fees are converted to HYPE and burned through the chain's execution logic. - Hyperliquid's airdrop rewards went to retail stakeholders who devoted time and money to the DEX in 2024, because no VCs participated in Hyperliquid's funding. - Hyperliquid Labs had barely a dozen people at the time of writing, which Jeff attributes to Hyperliquid being a set of primitives anyone can use to tokenise anything. - At the time of writing, TradeXYZ, the first deployer of permissionless perps on Hyperliquid, had accounted for roughly 2% of global silver volume in a few months of existence. - The Unit team, separate from Hyperliquid's core team, made Hyperliquid the leading source of spot price discovery for the XRP launch in 2025. - [How to Unlock $2 Trillion of Bitcoin Sitting Idle](https://castlelabs.io/research/how-to-unlock-2-trillion-of-bitcoin): Less than 0.5% of Bitcoin supply is used in DeFi. Compare WBTC, tBTC and BitVM trust models, and how Starknet's BTCFi stack and Re7 target institutional BTC. (2026-02-12; by TradFiHater; tags: Markets) Short answer: BTCFi puts idle bitcoin to work in DeFi through staking, lending, liquidity provision and trading. When we published in February 2026, less than 0.5% of the total Bitcoin supply took part. BTC reaches DeFi through custodial wrappers such as WBTC, threshold bridges such as tBTC, where 51 of 100 node operators must agree to move funds, or trust-minimised BitVM bridges, which need one honest participant. Starknet is building towards trustless BTC, with borrower-weighted STRK incentives on lending protocols such as Vesu and Re7 pre-seeding liquidity, and its TVL peaked at almost $325M at the end of January 2026. - [The Layer Zero for all Chains](https://castlelabs.io/research/the-layer-zero-for-all-chains): How LayerZero's OFT standard handles an estimated 85% of cross-chain messages, and what Zero, targeting 2 million TPS per Atomicity Zone, means for TradFi. (2026-02-11; by Castle Labs; tags: Markets) - [Don’t Trade Cryptos, Trade Rates: A Boros Funding Rates Primer](https://castlelabs.io/research/dont-trade-cryptos-trade-rates): Castle Labs research on Boros, Pendle’s yield-trading platform for funding rates, covering Yield Units, hedging use cases, vaults, liquidity, and power-user strategies. (2026-02-06; by Noveleader, Francesco; tags: Options, Trading, DeFi, Markets, Deep Dives) Short answer: Boros is Pendle's platform for trading funding rates with margin, using Yield Units that each represent the funding on 1 unit of collateral, such as ETH, on venues like Hyperliquid and Binance. A long pays a fixed implied APR set by the Boros order book and receives the exchange's floating funding rate, and a short does the reverse, which lets protocols such as Ethena turn volatile funding into a fixed APR. When we published in February 2026, Boros had about $8 million in TVL and a $268 million open interest high, and power users named thin liquidity as the main limit on position size. - [ICOs are officially back](https://castlelabs.io/research/icos-are-officially-back): Plasma, MegaETH, Monad and Aztec drew over $2 billion in token sale commitments in 2025. How LBPs, clearing auctions and fixed-price sales compare. (2026-02-05; by Noveleader, Francesco; tags: Markets) Short answer: Public token sales returned in 2025: Plasma, MegaETH, Monad and Aztec drew over $2 billion in commitments from more than 100,000 participants. Teams now choose a sale mechanism, from liquidity bootstrapping pools to continuous clearing auctions and fixed-price sales, and run it through platforms such as Legion, Sonar by Echo, Kaito, Coinlist and Tally, under clearer rules than 2017, including MiCA in the EU. Key takeaways: - Plasma, MegaETH, Monad and Aztec ran community token sales in 2025 that together drew over $2 billion in commitments from more than 100k participants. - Berachain's final private raise reportedly valued it at $1.5 billion and its token listed above $4 billion, but Berachain traded at a sub-$400 million valuation at the time of writing. - MegaETH's October 2025 public sale on Echo's Sonar drew $1.4 billion in bids for a $50 million cap, 28x oversubscribed, and cleared at a $1 billion FDV. - Monad ran the first major US-regulated token sale with Coinbase in late 2025, raising $188 million at a $2.5 billion valuation with only 1.43x oversubscription. - Tally's token sale platform supports Uniswap's CCA and Balancer's LBP, automatically seeds 20% of raised funds into Uniswap v4 LP positions and charges zero platform fees. - [ZK Proofs: Is Privacy Cheap Enough to Be Mainstream?](https://castlelabs.io/research/zk-proofs-making-privacy-cheap-enough): What ZK proofs cost across verification, proving and publishing, and how Ethereum zkVM proving fell from 16 minutes to 16 seconds with costs down 45x in 2025. (2026-02-04; by RubiksWeb3; tags: Privacy, Markets) - [Making All Markets Hyperliquid](https://castlelabs.io/research/making-all-markets-hyperliquid): How HIP-4 brings binary prediction markets to Hyperliquid, challenging Kalshi and Polymarket, and how much HYPE buyback revenue it could add at 3 to 4 bps fees. (2026-02-03; by Noveleader, Francesco; tags: Trading, Markets) - [Arbitrum Today: Strategic Focus and Growth Drivers](https://castlelabs.io/research/arbitrum-today-strategic-focus-and): Arbitrum's 2025 growth drivers: RWAs up 480% to $870 million, Timeboost generating over $6 million in revenue, DRIP incentives, and apps from Aave to Pendle. (2026-02-02; by Noveleader, Francesco; tags: L2s, Markets) Short answer: Arbitrum's growth in 2025 came from Timeboost revenue, RWAs, the DRIP incentive programme and putting its DAO treasury to work, and TVL held at about $3 billion after the October liquidation event. Timeboost, launched in April 2025, generated over $6 million and has regularly made up more than 30% of net chain revenue, about $25 million in total since the start of 2025. Tokenised RWAs grew 480% from $150 million to $870 million, and during DRIP's first season incentivised lending markets grew 14% to $2.91 billion while ETH fell over 30%. Key takeaways: - Arbitrum's Timeboost, launched in April 2025 to let MEV searchers bid for an express lane, has generated over $6 million and regularly exceeded 30% of net chain revenue. - Tokenised RWAs on Arbitrum grew 480% in 2025, from $150 million to $870 million, outpacing the over 300% growth in tokenised value across all chains. - During the first season of Arbitrum's DRIP programme, from 3 September 2025, incentivised lending markets grew 14% from $2.55 billion to $2.91 billion while ETH fell over 30%. - Arbitrum DAO's non-ARB treasury assets grew 71.8% in 2025, from $94 million to $161 million, and the DAO deployed over $100 million into ecosystem protocols to earn interest. - Over the 30 days before writing, Variational, an RFQ-based perpetual exchange, generated 75% of Arbitrum's derivatives volume, ahead of GMX at 7.5% and Ostium at 6.3%. - [USDai: Fuel for the Digital Age?](https://castlelabs.io/research/usdai-fuel-for-the-digital-age): How USDai funds GPU-backed loans at up to 70% LTV for AI infrastructure, why most of its ~$650M TVL sits in T-bills, and what the CHIP token may capture. (2026-01-30; by Castle Labs; tags: Markets) - [Why Bitcoin is Not (Yet) Digital Gold](https://castlelabs.io/research/why-bitcoin-is-not-yet-digital-gold): Bitcoin vs gold as a store of value: why BTC's 45% annualised volatility vs gold's 15% and gold's ~$32T market cap keep Bitcoin from being digital gold yet. (2026-01-26; by Castle Labs; tags: Markets) Short answer: Bitcoin passes the sound money test, but in 2025 it did not behave like a reserve asset. Its annualised volatility hovered around 45%, three times gold's 15%, and gold's market cap of around $32 trillion dwarfs it. Gold outperformed in 2025 as central banks in Poland, India, Turkey and China kept buying, while Bitcoin fell from its $125,000 peak in October 2025 to $80,000 by November. Among crypto assets, only Bitcoin could replace gold in theory, and the US Strategic Bitcoin Reserve of over 200,000 seized BTC gives it state backing no other crypto asset has. - [What do Drones have to do with Crypto?!](https://castlelabs.io/research/what-do-drones-have-to-do-with-crypto): Why Peter Thiel's Founders Fund backed N1, a sharded, trading-first L1 with sub-ms latency, and how its apps 01, Terminal and TradeRush chase a monopoly. (2026-01-12; by Castle Labs; tags: Markets) - [Prediction Markets Have Never Been This Strong](https://castlelabs.io/research/prediction-markets-have-never-been): Prediction markets in 2025: weekly volume hit $3.8B, Polymarket got $2B from ICE, Kalshi leans on sports betting, and what TGEs, oracles and AI mean for 2026. (2026-01-12; by Castle Labs; tags: Markets, Predictions) Short answer: Prediction market volumes ended 2025 far above their 2024 highs. Weekly volume reached $3.8 billion in December, roughly 660% up from the start of the year, and open interest passed $700 million, still below the nearly $900 million of the 2024 election. Polymarket and Kalshi form a duopoly with 44% and 47% of open interest, and in 2025 ICE invested $2 billion in Polymarket while Kalshi raised $1 billion at an $11 billion valuation. Sports make up as much as 98% of Kalshi's weekly volume, and we expect Polymarket's POLY token launch to be the main catalyst for 2026. Key takeaways: - Prediction market weekly volume reached $3.8 billion in December 2025, roughly 660% above the start of the year, as Q4 2025 monthly volumes approached $10 billion. - Polymarket and Kalshi held 44% and 47% of total prediction market open interest respectively at the time of writing, forming what Castle Labs calls a duopoly. - ICE, the NYSE's parent company, invested $2 billion in Polymarket in October 2025, and Kalshi raised $1 billion at an $11 billion valuation in December 2025. - Sports accounted for more than 40% of Polymarket's volume and over 90% of Kalshi's at the time of writing, with Kalshi's weekly share reaching up to 98%. - Castle Labs sees Polymarket's POLY token launch, with an airdrop expected to coincide with its US app release, as the main prediction market catalyst for 2026. - [Is Crime Legal Now?](https://castlelabs.io/research/is-crime-legal-now): 2025 crypto crime recap: Pump.fun memecoins where over 99% sit below $100k market cap, insider trading, the $1.5B Bybit hack and ~$3B in total losses. (2025-12-31; by Castle Labs; tags: Regulation, Markets) - [DeFi Will Win](https://castlelabs.io/research/defi-will-win): DeFi in 2025: the October event that caused $19B in liquidations, the Stream Finance blowup, Uniswap's UNIfication proposal and buybacks at Aave and HYPE. (2025-12-30; by Castle Labs; tags: DeFi, Markets) - [Secrecy and Grand Designs](https://castlelabs.io/research/secrecy-and-grand-designs): How Variational's RFQ perp DEX on Arbitrum acts as sole counterparty to every trade, versus Hyperliquid, Lighter and Aster, with TVL at $69M and OI at $397M. (2025-12-23; by Castle Labs; tags: Markets) - [CT Vibe Check: 2026 Predictions](https://castlelabs.io/research/ct-vibe-check-2026-predictions): Crypto Twitter's 2026 predictions from builders at Kalshi, Hyperliquid, Mantle and Rysk, plus traders, including a call for a BTC floor around 65 to 70k. (2025-12-23; by Castle Labs; tags: Markets) - [Proving YourSELF: Privacy and Identity Verification in the Age of AI](https://castlelabs.io/research/proving-yourself-privacy-and-identity): How Self Protocol uses ZK proofs of passport data for private proof of humanity on Celo, as AI deepfakes fuel fraud and Aave pays verified users double yield. (2025-12-17; by Castle Labs; tags: Privacy, Markets) - [Harmonix Finance: Innovating Yield Generation on HyperEVM and HIP-3 Markets](https://castlelabs.io/research/harmonix-finance-innovating-yield): How Harmonix Finance runs delta neutral HYPE vaults on HyperEVM and HIP-3 markets, with ~13% APY on USDC, over $17m TVL and the HAR token's fee sharing. (2025-12-16; by Castle Labs; tags: DeFi, Markets) - [AI: The Gift that Never Takes](https://castlelabs.io/research/ai-the-gift-that-never-takes): AI Dispatch #44 recaps December: Virtuals' ACP passes $370m in agentic activity, Surf raises $15m, Trump's AI executive order, and verifiable AI via EigenCloud. (2025-12-15; by Castle Labs; tags: AI, Markets) - [To the New Beginning: the Almanak Games](https://castlelabs.io/research/to-the-new-beginning-the-almanak): How the Almanak Games reward users before its TGE: AI agent Vaults and Strategy Builder, TVL above $130M, and 1.5% to 3% of token supply via points. (2025-12-11; by Castle Labs; tags: Markets) - [Redstone x Arbitrum Stylus: Oracles Beyond the EVM Limitations](https://castlelabs.io/research/redstone-x-arbitrum-stylus-oracles): How RedStone's oracle on Arbitrum Stylus uses Rust and WASM to bypass EVM limits, halving per-feed computation for 26% total savings versus its EVM version. (2025-12-10; by Castle Labs; tags: L2s, Markets) - [A Tale of two Worlds: A Comparative Analysis of Crypto Regulations Worldwide](https://castlelabs.io/research/a-tale-of-two-worlds-a-comparative): Crypto regulation compared: the US pivot under the GENIUS Act and Paul Atkins' SEC, the EU's risk-averse MiCA regime, and why Switzerland is the ideal model. (2025-12-09; by Castle Labs; tags: Regulation, Markets) Short answer: We see the US pivoting towards crypto, the EU choosing a risk-averse regime under MiCA, and Switzerland as the model to follow. The GENIUS Act of July 2025 requires 1:1 treasury backing for stablecoins, Paul Atkins' SEC launched Project Crypto, and the repeal of SAB 121 let banks hold crypto, though the CLARITY Act was still stalled when we published in December 2025. MiCA holds crypto service providers to MiFID II-style compliance and requires a registered office in a member state. Switzerland pairs FINMA supervision with a 2021 DLT law that amended 10 federal laws to recognise crypto assets, and has attracted the Ethereum, Solana and Cardano foundations. - [Is AI in a Bubble?](https://castlelabs.io/research/is-ai-in-a-bubble): With Bitcoin back at $86k and VIRTUALS below $0.85, how the sell-off hits crypto AI tokens, whether AI is in a bubble, and notes on Gemini 3 Pro and Opus 4.5. (2025-12-01; by Castle Labs; tags: AI, Markets) - [The Almanak Strategy Builder: How to create your own strategy with the help of an AI swarm](https://castlelabs.io/research/the-almanak-strategy-builder-how): How Almanak's Strategy Builder uses an AI agent swarm to turn ideas into deterministic, verifiable trading code, tested by building a Pendle RSI strategy. (2025-11-14; by Castle Labs; tags: Trading, Markets) - [Prediction Markets: The Race to Predict Everything](https://castlelabs.io/research/prediction-markets-the-race-to-predict): How prediction markets work and how Polymarket, Kalshi, Limitless and Myriad compare on design and liquidity, with Polymarket at ~$22B cumulative volume. (2025-11-13; by Castle Labs; tags: Markets, Predictions) Short answer: Prediction markets let users buy Yes or No shares on an event, and the share price reflects the odds, so an 82% chance costs $0.82. When we published in November 2025, Polymarket and Kalshi were heading towards a duopoly. Polymarket, which runs on order books and resolves markets through UMA's optimistic oracle, had done about $22 billion in cumulative volume, while Kalshi, regulated by the CFTC and resolving markets with human oversight, had done over $15 billion, with sports driving over 80% of its volume. Limitless (over $500 million) and Myriad (over $60 million) trailed far behind. - [The Monad Ecosystem Guide: Everything you can do once Mainnet goes Live](https://castlelabs.io/research/the-monad-ecosystem-guide-everything): Monad mainnet goes live November 24: a guide to its ecosystem, from Perpl and Kuru, which raised $11.6M led by Paradigm, to Curvance, Fastlane and games. (2025-11-12; by Castle Labs; tags: Deep Dives, Markets) Short answer: Monad is a high-performance L1 built from scratch, using MonadBFT, asynchronous execution and parallel transaction execution to reach 10k TPS with sub-second finality. When we published in November 2025, ahead of the mainnet launch on the 24th of November, the ecosystem had over 300 projects, 78 of them exclusive to Monad. MON holders can trade on CLOB exchanges such as Kuru, which raised $11.6 million in a round led by Paradigm, stake through Fastlane's shMON, lend on Curvance or Modus, and try games, NFTs and AI apps. - [OpenEden: On The EDEN Token, Roadmap and Bringing more RWAs Onchain](https://castlelabs.io/research/openeden-on-the-eden-token-roadmap): How OpenEden's EDEN token works, from staking for xEDEN at an 8.38% average APY to fee utility, plus the roadmap for USDO and growing its $225m tokenized value. (2025-11-06; by Castle Labs; tags: RWAs, DeFi, Markets) - [Mantle's RWA Push: Over $150k in Builder Prizes in their Global Hackathon](https://castlelabs.io/research/mantles-rwa-push-over-150k-in-builder): Mantle's RWA hackathon offers over $150k in prizes across 6 tracks, with $15k per track and a $30k grand prize, as Mantle shifts to a regulated RWA ecosystem. (2025-11-05; by Castle Labs; tags: RWAs, DeFi, L2s) - [xMAQUINA: The Robotics Treasury Company](https://castlelabs.io/research/xmaquina-the-robotics-treasury-company): How xMAQUINA's DEUS works as a robotics treasury company: a $32M NAV with stakes in Figure, Apptronik and 1X, a SubDAO launchpad, and a 1.56x premium to NAV. (2025-11-03; by Castle Labs; tags: AI, Markets) - [Drip or Drop? Analysing the Effectiveness of the DRIP Incentive Program on the Arbitrum Ecosystem](https://castlelabs.io/research/drip-or-drop-analysing-the-effectiveness): Arbitrum's DRIP program distributes up to 24M ARB to lending markets. In under 2 months, lending TVL rose 26.7% to $1.94b. Which protocols and assets won? (2025-10-31; by Castle Labs; tags: L2s, Deep Dives) - [Upshift: Institutional-Grade Yield Strategies for the Masses](https://castlelabs.io/research/upshift-institutional-grade-yield): Upshift runs curated DeFi and CeFi yield vaults on top of August's prime brokerage. How the vaults, curators, risk controls and Core USDC work. (2025-10-29; by Castle Labs; tags: DeFi, Markets) Short answer: Upshift is a vault platform where funds such as MEV Capital and K3 Capital run yield strategies across DeFi and CeFi, while August handles liquidations and multi-oracle NAV pricing in the background. When we covered it in October 2025 it held over $500 million in TVL across more than 21 ERC-4626 vaults, its Core USDC vault of vaults showed a 7.51% historical return, and its modular vault setup had already launched vaults beyond Ethereum on HyperEVM and Sui. Key takeaways: - Upshift held over $500 million in TVL across more than 21 vaults at the time of writing, with vaults run by DeFi funds such as MEV Capital and K3 Capital. - All Upshift strategies run on August's prime brokerage, whose risk engine handles liquidations automatically and maintains each vault's NAV with a multi-oracle pricing system. - Upshift uses an allowlist-first model in which chains, protocols and contract calls must be allowlisted before curators can use them, keeping vaults non-custodial for depositors. - At the time of writing, Upshift's top vaults by TVL were High Growth ETH ($77 million), the K3 Neutrl Pre-deposit Vault ($75 million) and Kelp Gain (over $61 million). - Upshift's Core USDC, an August-managed vault of vaults spreading deposits across stablecoin vaults, had over $14 million in TVL and 7.51% historical performance at the time of writing. - [Crypto, AI, Robotics: The Technological Trinity](https://castlelabs.io/research/crypto-ai-robotics-the-technological): How Virtuals is adding robotics to its stack and agentic GDP, and how ACP, Butler and Unicorn could let AI agents hire robots and settle payments onchain. (2025-10-27; by Castle Labs; tags: AI, Markets) - [Oracle Infrastructure: The Backbone of Lending Protocols](https://castlelabs.io/research/oracle-infrastructure-the-backbone): What oracles lending protocols need for long-tail and wrapped BTC collateral, and how Aave, Morpho, Euler and Fluid use Chainlink, Pyth, RedStone and Chronicle. (2025-10-23; by Castle Labs; tags: DeFi, Markets, Deep Dives) Short answer: Every major money market we reviewed uses Chainlink price feeds, and when we published in October 2025 Chainlink secured $91.5 billion, ahead of Chronicle ($11.2 billion), RedStone ($8.5 billion) and Pyth ($7.9 billion). For long-tail and wrapped BTC collateral, implementation decides safety more than the provider does: protocols need exchange-rate oracles for wrapped and staked assets such as cbBTC and LBTC, staleness limits on pull oracles (Euler treats Pyth prices as stale after two to three minutes), and a second source, which Fluid gets by checking Chainlink against a Uniswap TWAP. Most oracle exploits come from protocols reading thin spot prices directly. - [Mantle's Momentum: Short-Term Catalysts and Onchain TVL](https://castlelabs.io/research/mantles-momentum-short-term-catalysts): Mantle's RWA strategy and catalysts: onchain TVL above $2.5b, RWA commitments up from $100m to $255m, the UR account, Bybit's user base and MNT discount buys. (2025-10-22; by Castle Labs; tags: L2s, Markets) - [Digital Asset Treasuries (DATs): How good of a play are they really?](https://castlelabs.io/research/digital-asset-treasuries-dats-how): How digital asset treasuries work: Strategy holds ~640k BTC, why mNAV premiums drive DAT stocks, how they raise capital, and the risks for investors. (2025-10-20; by Castle Labs; tags: Markets) Short answer: Digital asset treasuries are listed companies that raise equity or debt to buy crypto, and their model works only while the stock trades at a premium to its holdings, measured as the multiple to net asset value (mNAV). When we published in October 2025, Strategy held about 640k BTC, 3% of the total supply, but its mNAV had fallen from as high as 6x to 1.16x, and its market cap had dropped 44% and Metaplanet's more than 70%. Below an mNAV of 1, new share issuance lowers crypto per share for existing holders, and we scored Strategy 2.83 out of 5 on six sustainability metrics. - [The Builder Momentum: How Chain Adoption Is Evolving and What It Means for Small Builders](https://castlelabs.io/research/the-builder-momentum-how-chain-adoption): Base and Solana lead small-builder growth, with full-time devs up 40% and 62% while Arbitrum fell 34% and Optimism 18%. Developer, user and survey data. (2025-10-17; by Castle Labs; tags: Markets) - [An Introduction To Robotics: The Next Frontier of Crypto](https://castlelabs.io/research/an-introduction-to-robotics-the-next): Crypto x robotics is a ~$404M sector vs ~$112B for robotics stocks. How the stack works across peaq, Auki, Geodnet, CodecFlow and Silencio, with market caps. (2025-10-13; by Castle Labs; tags: AI, Markets) - [The Year of the Neobank](https://castlelabs.io/research/the-year-of-the-neobank): Crypto neobanks compared: EtherFi Cash with 2-3% SCR cashback and liquidUSD yield vs Plasma One with 10%+ yield and up to 4% cashback in over 150 countries. (2025-10-03; by Castle Labs; tags: DeFi, Markets) - [Prediction Markets: A Differentiated Hedging Tool](https://castlelabs.io/research/prediction-markets-a-differentiated): Why prediction markets like Polymarket, with nearly $1 billion in monthly volume, can hedge real-world event risk, and what the 97 sector projects need to grow. (2025-10-02; by Castle Labs; tags: Markets, Predictions) - [Agents, Infra, and Alpha](https://castlelabs.io/research/agents-infra-and-alpha): AI Dispatch #36: Virtuals launches its Butler agent on X, Nansen AI goes mobile, Ethereum forms a decentralised AI team, x402 payments, and Token2049 AI events. (2025-09-29; by Castle Labs; tags: AI, Markets) - [Core Architecture and Positioning of DeFi's Top Money Markets](https://castlelabs.io/research/core-architecture-and-positioning): Aave vs Morpho vs Euler vs Fluid: how DeFi's top money markets differ in architecture, liquidations and interest rate models. Aave controls ~$40 billion in TVL. (2025-09-26; by Castle Labs; tags: DeFi, Markets) Short answer: Aave, Morpho, Euler and Fluid differ mainly in how they pool or isolate liquidity. Aave runs pooled markets and, when we published in September 2025, held about $40 billion in TVL, more than five times Morpho, whose immutable isolated markets sit under curated vaults and power Coinbase's cbBTC loans. Euler V2 lets builders launch isolated ERC-4626 vaults and link them through the Ethereum Vault Connector. Fluid puts lending, vaults and its DEX on one Liquidity Layer, with LTVs up to 97% and liquidation penalties as low as 0.1%. Key takeaways: - Aave, deployed on 19 chains with over 85% of its liquidity on Ethereum, held about $40 billion in TVL at the time of writing, more than five times second-placed Morpho. - Aave v4's planned Liquidity Hub is designed to aggregate liquidity across the Prime and Core markets, which Aave v3 keeps separate, with Spokes plugging in as specialised market instances. - Coinbase users had borrowed over $700 million USDC through Morpho against more than $1.3 billion of cbBTC at the time of writing, over 15% of all Morpho borrowing. - Euler V2, launched in September 2024, is built on isolated ERC-4626 vaults, and its Ethereum Vault Connector lets collateral in one vault back borrowing in another. - Fluid, built by Instadapp, runs lending, vaults and its DEX on one Liquidity Layer, with vault LTVs of up to 97% and liquidation penalties as low as 0.1%. - [The Solana Evolution: From Memecoins to Consumer Apps](https://castlelabs.io/research/the-solana-evolution-from-memecoins): How Solana moved from memecoins to DeFi and consumer apps: PumpFun's $500M+ revenue, DeFi TVL of $12.5B above its 2021 high, and the Seeker phone. (2025-09-25; by Castle Labs; tags: Markets) Short answer: Solana rebuilt after the FTX collapse through memecoins, then shifted towards DeFi and consumer apps. PumpFun drove the memecoin era, with over five million tokens launched by the middle of 2024 and over $500M in revenue, and its graduated tokens fed trading and liquidity to Raydium and Jupiter. When we published in September 2025, Solana DeFi TVL stood at $12.5B, above its 2021 high of $11.7B and nearly 60x its 2023 low of $210M. Solana is now courting consumer apps, from the Seeker phone to trading card projects such as Collector Crypt and Phygitals. - [Loopscale: Order book lending on Solana](https://castlelabs.io/research/loopscale-order-book-lending-on-solana): How Loopscale's order book lending works on Solana, matching lenders and borrowers for better rates, with over $100m deposited as Solana lending TVL tops $4b. (2025-09-25; by Castle Labs; tags: DeFi, Markets) Short answer: Loopscale is an order book lending protocol on Solana with curated vaults, one-click loops and fixed-rate, fixed-term loans. Lenders post offers with their own rates, collateral and terms, and a matching engine pairs them with borrowers, so interest goes only to capital that is lent out. In pool-based models such as Aave and Compound, idle deposits share the yield, so lenders earn less than borrowers pay. When we published in September 2025, Loopscale had over $100m in deposited liquidity and $40m in active loans, while Solana lending TVL had passed $4b. - [On RWAs and Institutional-Grade Assets: The Case of OpenEden](https://castlelabs.io/research/on-rwas-and-institutional-grade-assets): Tokenized RWAs doubled to over $29B in 2025. What makes an asset institutional-grade, using OpenEden's TBILL and USDO stablecoin as the case study. (2025-09-23; by Castle Labs; tags: RWAs, DeFi, Markets) - [Can an AI Agent be your Personal Quant?](https://castlelabs.io/research/can-an-ai-agent-be-your-personal): How Almanak's AI agent swarm acts as a personal quant, from strategy design to custody, privacy and human-in-the-loop safeguards. Its vaults passed $60M TVL. (2025-09-19; by Castle Labs; tags: AI, Markets) - [State of Stylus: Breaking Barriers on Arbitrum](https://castlelabs.io/research/state-of-stylus-breaking-barriers): How Arbitrum Stylus adds a WASM VM for Rust, C and C++ contracts, what the Stylus Sprint funded with over 2M ARB, and builds like RedStone and Renegade. (2025-09-18; by Castle Labs; tags: L2s, Deep Dives) - [The AI Dispatch #34 - Robotics, Vader Pivot, Backroom OCDs](https://castlelabs.io/research/the-ai-dispatch-34-robotics-vader): Crypto AI weekly: robotics tokens emerge as the next wave, Vader pivots to EgoPlay smart glasses tasks, SentientAGI drama, and Almanak passes $50M in TVL. (2025-09-15; by Castle Labs; tags: AI, Markets) - [USDH: A Hyperliquid Aligned Stablecoin](https://castlelabs.io/research/usdh-a-hyperliquid-aligned-stablecoin): USDH proposals compared: Paxos, Frax, Native Markets, Agora, Sky, Ethena and others bid for Hyperliquid's native stablecoin, as 95% of its $5.8B supply is USDC. (2025-09-10; by Castle Labs; tags: Trading, DeFi, Deep Dives) Short answer: USDH is the ticker Hyperliquid reserved for a natively minted, compliant stablecoin, and validators were set to vote on 14 September 2025 on which issuer would get it. When we published in September 2025, 95% of Hyperliquid's $5.8 billion stablecoin supply was USDC, sending over $230 million a year in reserve yield to external issuers. Paxos, Ethena and Bastion offered at least 95% of reserve yield back to Hyperliquid, Frax and Agora 100%, Native Markets a 50/50 split between the Assistance Fund and USDH growth, and Sky a 4.85% return on USDH held on Hyperliquid. Key takeaways: - At the time of writing, 95% of Hyperliquid's $5.8B stablecoin supply sat in USDC, which Castle Labs estimates exports over $230M a year in reserve yield to external issuers. - Paxos, Ethena and Bastion each pledged at least 95% of reserve yield to Hyperliquid's ecosystem in their USDH bids, with Paxos taking no fee until USDH passes $1B TVL. - Frax proposed returning 100% of Treasury yield to Hyperliquid, while Agora, leading a coalition including State Street and VanEck, pledged 100% of net revenue from USDH reserve assets. - Native Markets proposed splitting USDH reserve yield 50/50, with half going immutably to the Assistance Fund through HYPE buybacks and half reinvested into USDH growth. - Sky's USDH proposal offered Hyperliquid a 4.85% return on all USDH held on Hyperliquid, combining the 4.75% Sky Savings Rate with a 10bps Integrator Reward. - [The AI Dispatch #33 - Agent Pear 🍐](https://castlelabs.io/research/the-ai-dispatch-33-agent-pear): How Agent Pear finds pair trades by statistical arbitrage, flagging spreads beyond two standard deviations via z-scores, and executes on DEXs like Hyperliquid. (2025-09-08; by Castle Labs; tags: AI, Markets) Key takeaways: - Pear Protocol's Agent Pear is a statistical arbitrage agent that monitors the z-score between two assets and flags pair trades when their spread moves beyond two standard deviations. - Agent Pear treats a z-score below -2 as a long signal, and its backend library updates asset correlations, z-scores, beta and spread in real time. - Agent Pear can execute viable signals as live trades by integrating directly with DEXs such as Hyperliquid, working in a batch generation phase and a continuous monitoring phase. - Pear Protocol planned limited free access to Agent Pear in September, with Phase 2 requiring users to hold $stPEAR or $PEAR to access its signals. - Pear Protocol plans a Phase 3 that makes Agent Pear conversational through LLM prompts and a Phase 4 in which it trades autonomously from stablecoin-funded Automatic Vaults. - [Going Beyond GPT Wrappers: The Next Phase of AI Agents](https://castlelabs.io/research/going-beyond-gpt-wrappers-the-next): Why LLM trading agents fall short and how Almanak uses structured multi-agent strategies and deterministic execution for verifiable onchain asset management. (2025-09-05; by Castle Labs; tags: AI, Markets) - [Abstract Chain: The Consumer L2 Where Culture is the Onramp](https://castlelabs.io/research/abstract-chain-the-consumer-l2-where): How Abstract, the ZKsync-powered consumer L2 from the Pudgy Penguins team, hit 5.9M wallets and 116.96M transactions with games, Myriad, Bigcoin and XP. (2025-09-04; by Castle Labs; tags: L2s, Markets) Short answer: Abstract is a consumer L2 from Igloo Inc, the team behind Pudgy Penguins, built on ZKsync's ZK Stack with EigenDA, where every account is a smart contract wallet opened with an email, Google login or passkey. At the time of writing it had passed 5.9 million wallets and 116.96 million cumulative transactions since its late January 2025 mainnet, mostly from gaming and social activity, while TVL stood at a modest $50.62 million. Animoca Research put its 30-day retention at 40%, and the $ABS token had not launched as of August 2025. - [Ronin’s Return to Ethereum](https://castlelabs.io/research/ronins-return-to-ethereum): Ronin is moving from sidechain to Ethereum L2. How the Arbitrum, Optimism, Polygon and ZKsync bids compare on performance, sovereignty, migration and grants. (2025-09-02; by Castle Labs; tags: L2s, Markets) - [Fluid: DeFi's New Dominator](https://castlelabs.io/research/fluid-defis-new-dominator): How Fluid works: Smart Collateral, Smart Debt, vaults with up to 95% LTV and 0.1% liquidation penalties, plus Jupiter Lend on Solana and Fluid DEX v2. (2025-08-28; by Castle Labs; tags: DeFi, Markets) - [Stablecoin Chains: Which Blockchain is Best Placed to Capitalize on the Stablecoin Revolution?](https://castlelabs.io/research/stablecoin-chains-which-blockchain): Plasma, Codex, 1Money, Arc, Stable and Monad compared on fees, finality, compliant privacy, cross-chain interoperability and enterprise readiness for stablecoin payments. (2025-08-26; by Castle Labs; tags: DeFi, Markets) Short answer: No chain had won when we compared six stablecoin chains in August 2025, since none was on mainnet, but Circle's Arc was the most institutionally integrated and Plasma had already lined up $1 billion in stablecoin liquidity through its XPL sale. Arc uses USDC as its gas token, draws on Circle's USDC supply of over $68 billion and targets finality under 350 milliseconds, while Plasma sponsors gas for USDT transfers. Stable makes peer-to-peer USDT transfers free, 1Money drops smart contracts entirely and runs KYC'd validators, and Monad still charges gas in its native token. - [The AI Dispatch #31 - FlipsideAI](https://castlelabs.io/research/the-ai-dispatch-31-flipsideai): How FlipsideAI turns natural language prompts into onchain analytics across 35+ blockchains, usable in Claude and Cursor via its MCP server, plus AI news. (2025-08-25; by Castle Labs; tags: AI, Markets) - [The Mantle Report: Growth Factors and Strategic Approach](https://castlelabs.io/research/the-mantle-report-growth-factors): What drove Mantle's July 2025 growth: a $3.1B treasury, mETH as the 4th biggest LST with $1.69B TVL, the UR banking app, and MNT value capture via Bybit. (2025-08-22; by Castle Labs; tags: L2s, Deep Dives) - [Scaling Onchain Culture with Base](https://castlelabs.io/research/scaling-onchain-culture-with-base): How Base, with $4.9B TVL and 24.4M monthly active addresses, uses the Base app, Zora content coins and creator rewards to bring creators onchain, and the risks. (2025-08-21; by Castle Labs; tags: L2s, Markets) - [The AI Dispatch #30- August in the AI Trenches, Backroom, Caesar AI](https://castlelabs.io/research/the-ai-dispatch-30-august-in-the): Crypto AI weekly: the AI sector bounces 30% within a $10 to 14B range, how Backroom's AI-run gated rooms and ROOM token work on Virtuals, plus Caesar AI. (2025-08-17; by Castle Labs; tags: AI, Markets) - [A Deep Dive Into Arbitrum's RWA Sector](https://castlelabs.io/research/a-deep-dive-into-arbitrums-rwa-sector): How Arbitrum grew its RWA sector past $350m TVL through STEP, STEP 2 and treasury programs, with issuers like Spiko, Franklin Templeton, Securitize and Ondo. (2025-08-14; by Castle Labs; tags: RWAs, DeFi, L2s, Deep Dives) - [The AI Dispatch #29 - Eliza is Back? Bullish Vibes, DePai](https://castlelabs.io/research/the-ai-dispatch-29-eliza-is-back): Crypto AI weekly: ETH breaks $4k, elizaOS updates, and how Almanak's AI agent swarm builds code-based strategies, its first vault's $5M cap filled in 24 hours. (2025-08-11; by Castle Labs; tags: AI, Markets) - [User Retention Analysis for Incentive Programs](https://castlelabs.io/research/user-retention-analysis-for-incentive): Compound's 1.8M ARB LTIPP grant under cohort retention analysis: supply TVL rose 180%, stablecoin pools retained best, and retained users deposited $154K. (2025-08-08; by Castle Labs; tags: Markets) Key takeaways: - Compound received 1.8 million ARB, the largest grant in Arbitrum DAO's Long-Term Incentive Pilot Program (LTIPP), which funded Compound incentives from June to August 2024. - Compound's supply TVL rose 180%, from about $90m to about $260m, during the 2024 incentive period, with returning users supplying 4.5x the deposit volume of new users. - Retained Compound users averaged $154K in first-month deposits against $9K for inactive users, and Castle Labs argues that targeting high-capital users early is a more efficient use of incentives. - USDT and USDC incentives on Compound retained capital better than ETH incentives, and the USDT pool introduced during the LTIPP accounted for about 30% of all deposits at peak. - Compound distributed its ARB incentives through external partners including OKX, Vaultcraft, Merkl and Layer3, and OKX-onboarded users accounted for the majority of deposits at the incentive peak. - [The AI Dispatch #28 - InfoFi Improvements, AIxbt Update, Robotics](https://castlelabs.io/research/the-ai-dispatch-28-infofi-improvements): Crypto AI weekly: the AI sector holds $10B support, Theoriq and Almanak token sales, AIxbt adds sentiment and whale tracking, and Kaito's InfoFi score shake-up. (2025-08-04; by Castle Labs; tags: AI, Markets) - [The Beras Are Back: An Introduction to Berachain PoL V2](https://castlelabs.io/research/the-beras-are-back-an-introduction): How Berachain PoL V2 adds the BERA Yield Module, routing 33% of PoL emissions to staked BERA, and what it changes for BGT holders, builders and institutions. (2025-07-31; by Castle Labs; tags: Markets) - [Stablecoin Adoption and Regional Usage in 2024 and 2025: What Changed?](https://castlelabs.io/research/stablecoin-adoption-and-regional): Stablecoins from mid-2024 to mid-2025: market cap up from ~$160B to over $260B, active wallets doubling to ~40M, USDT vs USDC, yield stablecoins, regional use. (2025-07-30; by Castle Labs; tags: DeFi, Markets) Short answer: Between mid-2024 and mid-2025, stablecoin market cap rose from about $160 billion to over $260 billion and active wallets doubled from about 20 million to about 40 million. Fiat-backed coins grew from about 85% of the market to over 90%, led by USDT, whose supply rose from about $83 billion to about $150 billion. In Latin America and Africa, stablecoins, mostly USDT on Tron, became a dollar substitute and remittance rail, cutting the cost of sending $200 in Sub-Saharan Africa by about 60%. In the US and Europe they mostly sit inside fintech stacks, corporate treasuries and cross-border settlement. - [The AI Dispatch #27 - Same Old, $REI Network, Hyperliquid AI?](https://castlelabs.io/research/the-ai-dispatch-27-same-old-rei-network): Crypto AI weekly: Theoriq and Almanak raise via Kaito and Legion, and how Rei Network's Core synthetic brain tackles LLM memory, stale data and hallucinations. (2025-07-28; by Castle Labs; tags: AI, Markets, Trading) - [A Comparative Ecosystem Analysis of Arbitrum's Positioning in DeFi](https://castlelabs.io/research/a-comparative-ecosystem-analysis): Where Arbitrum leads and lags across DEXs, derivatives, lending, yield, staking and RWAs, with $2.9 billion in DeFi TVL and 41 Orbit chains live in July 2025. (2025-07-25; by Castle Labs; tags: DeFi, L2s, Deep Dives) Short answer: When we published in July 2025, Arbitrum had about $2.9 billion in DeFi TVL, roughly 2% of all DeFi value, and led all networks with over $1 billion of net inflows in the previous month. It held 2.4% of DEX TVL, 2.2% of lending TVL and 1.3% of RWA value, and had no major CDP protocol. Its earlier incentive programs, STIP, LTIPP and STIP.b, did not sustain growth once they ended, so we recommend targeted incentives for protocols already live on the chain, such as GMX, Ostium, Euler, Fluid and Pendle, plus Merkl rewards for lending and liquid staking assets. - [Beyond the Monolith: MetaDAOs and the New Blueprint for Onchain Ecosystems](https://castlelabs.io/research/beyond-the-monolith-metadaos-and): What MetaDAOs are and how they allocate capital via subDAOs and spinouts, with examples from Sky's Endgame, GnosisDAO and BitDAO/Mantle, plus key risks. (2025-07-24; by Castle Labs; tags: Deep Dives, Markets) - [The AI Dispatch #26 - Finding a Balance, The Launchpad Wars and Loomlay](https://castlelabs.io/research/the-ai-dispatch-25-finding-a-balance): Crypto AI weekly: AI market cap back above $12B, REKT, TIBBIR and TOSHI outperform, the Virtuals vs CreatorBid vs Loomlay launchpad war, and Loomlay's agents. (2025-07-21; by Castle Labs; tags: AI, Markets) - [The AI Dispatch #25 - BTC at ATH but what About AI? Intro on Giza, Virtuals Suffering](https://castlelabs.io/research/the-ai-dispatch-25-btc-at-ath-but): Why AI tokens lag as BTC sets new highs, the struggles of Virtuals and ACP, and how Giza's DeFi agents deliver 83% higher yields than static lending positions. (2025-07-14; by Castle Labs; tags: AI, Markets) - [An Overview of Gaming on Arbitrum](https://castlelabs.io/research/an-overview-of-gaming-on-arbitrum): How Arbitrum is pushing into gaming with Stylus, Orbit chains like Xai and Proof of Play, and a 200 million ARB Gaming Catalyst Program from the DAO. (2025-07-10; by Castle Labs; tags: L2s, Markets) - [The AI Dispatch #24 - The Next Wave](https://castlelabs.io/research/the-ai-dispatch-24-the-next-wave): Crypto AI weekly: the Virtuals Agent Commerce Protocol (ACP) goes live, AI agent market cap slips below $10B on tariff fears, and what builders showed at EthCC. (2025-07-07; by Castle Labs; tags: AI, Markets) - [The AI Dispatch #23 - Virtual Ecosystem, Kaito, TEE, Vader](https://castlelabs.io/research/the-ai-dispatch-23-virtual-ecosystem): Crypto AI weekly: Virtuals genesis launches hit by the downturn, Vader team wallet concerns, newcomers WARP, MANTIS and WACH, plus Kaito and how TEEs secure AI. (2025-06-30; by Castle Labs; tags: AI, Markets, Deep Dives) - [Avantis: The onchain Leverage Layer for Every Asset](https://castlelabs.io/research/avantis-the-onchain-leverage-layer): How Avantis zero-fee perpetuals work: traders pay a fee only on profits and keep over 85% of gains, plus synthetic RWAs for FX and commodities. (2025-06-26; by Castle Labs; tags: Markets) - [Introducing EigenCloud: Enter the Verifiable Cloud Era](https://castlelabs.io/research/introducing-eigencloud-enter-the): How EigenCloud gives developers cloud-grade programmability with crypto verifiability, using AVSs, EigenVerify, EigenCompute and the EIGEN token on EigenLayer. (2025-06-20; by Castle Labs; tags: Markets) Short answer: EigenCloud is EigenLayer's verifiable cloud, which lets developers run complex application logic in offchain containers while the chain handles only token logic and proof checks. Operators running those containers stake collateral that can be slashed if they return a wrong result, and over $11B secured EigenLayer at the time of writing. EigenVerify settles disputes by re-executing deterministic code or, for intersubjective tasks, by a vote of EIGEN-staking operators, and EigenCompute handles deployment and the composition of AVSs such as EigenDA. If a majority of those operators collude, the EIGEN token can fork, and the fork they control would be abandoned. Key takeaways: - EigenCloud lets developers move complex application logic into verifiable offchain containers, while the chain only handles essential token logic and checks proofs from those containers. - Operators running EigenCloud containers stake crypto as collateral that can be slashed if their container produces a wrong result, and over $11B secured EigenLayer at the time of writing. - EigenVerify settles disputes by re-executing deterministic code for objective tasks, while intersubjective tasks go to a vote of operators who stake EIGEN as a bond. - If a majority of operators staking EIGEN for EigenVerify collude on a false result, EIGEN can undergo a socially-agreed fork in which the colluders' version is abandoned and loses value. - EigenCompute is designed to deploy developers' Docker-style containers to opted-in EigenLayer operators and simplify composing AVSs such as EigenDA, which provides verifiable data storage. - [Arbitrum Chains: Scaling on Your Own Terms](https://castlelabs.io/research/arbitrum-chains-scaling-on-your-own): How Arbitrum Orbit chains and licensing work: 35 chains live with $252M bridged, with case studies from Gravity, Corn, ApeChain and Proof of Play. (2025-04-22; by Castle Labs; tags: L2s, Deep Dives) - [All for One and One for All: The Evolution of Consensus Models with Hyperliquid, Monad & Sonic](https://castlelabs.io/research/all-for-one-and-one-for-all-the-evolution): How the consensus models of Hyperliquid (HyperBFT), Monad (MonadBFT) and Sonic compare on speed and security. HyperBFT reaches 0.2s finality with 16 validators. (2025-04-16; by Castle Labs; tags: Trading, Markets) - [The Ultimate Guide to Farm the Hyperliquid Ecosystem](https://castlelabs.io/research/the-ultimate-guide-to-farm-the-hyperliquid): How to farm the HyperEVM airdrop meta after HYPE's 31% genesis drop: strategies for KittenSwap, HypurrFi, HyperLend, Felix, Hyperbeat and HyperSwap. (2025-04-11; by Castle Labs; tags: Trading, Deep Dives) Short answer: We expected the next Hyperliquid rewards to go to active perp traders, holders of Hyperliquid-native NFTs and HyperEVM users who lend, borrow and provide liquidity. The guide sets out ten ways to farm, from KittenSwap's boosted concentrated liquidity pools and HyperLend's points programme to minting LHYPE and staking HYPE with validators such as HypurrCollective x Nansen. We wrote it in April 2025, after the November 2024 airdrop gave 31% of HYPE's supply to over 90,000 users, with another 39%, nearly $4 billion at the time, still unclaimed. Key takeaways: - Hyperliquid's November 2024 airdrop gave 31% of HYPE's supply to over 90,000 users, and at the time of writing another 39%, nearly $4 billion, remained unclaimed. - Castle Labs expects active perp traders, holders of Hyperliquid-native NFTs and HyperEVM users who lend, borrow and provide liquidity to be the top candidates for future HYPE rewards. - At the time of writing, KittenSwap was HyperEVM's largest DEX by volume and TVL, with concentrated liquidity pools such as LHYPE/HYPE earning temporarily boosted multipliers of up to 20x. - HyperLend, the primary lending protocol on HyperEVM, underpins Harmonix, Mizu and Felix and launched its public points programme on 7 April 2025. - Stakers with the HypurrCollective x Nansen validator had already received airdrops of 2% of Sentiient AI's SENT supply and 0.5% of Timeswap's TIME supply. - [OpenServ: The Operating System for Agentic Workflows](https://castlelabs.io/research/openserv-the-operating-system-for): How OpenServ moves AI agents from isolated silos to multi-agent teams with its cognition framework, orchestration layer and agent collaboration protocol. (2025-04-10; by Castle Labs; tags: AI, Markets) - [GTE: The Vertical Exchange on MegaETH](https://castlelabs.io/research/gte-the-vertical-exchange-on-megaeth): How GTE on MegaETH combines a launchpad, AMM, onchain CLOB and aggregator in one venue, how it compares to Hyperliquid, Jupiter and GMX, and its $2.5m raise. (2025-04-04; by Castle Labs; tags: L2s, Deep Dives) - [Solana Perpetual Powerhouses: An Overview of GMX-Solana, Jupiter, and Drift](https://castlelabs.io/research/solana-perpetual-powerhouses-an-overview): GMX-Solana vs Jupiter Perps vs Drift: how each Solana perp DEX works, fees from 3 to 10 bps, JLP and GLV liquidity, volume and risk management compared. (2025-04-02; by Castle Labs; tags: Trading, Deep Dives) Short answer: Jupiter and Drift lead Solana perps by volume, with GMX-Solana, an adapted GMX V2, as the newest entrant. Jupiter Perps uses a pool model in which JLP, an index of SOL, ETH, wBTC, USDC and USDT, takes the other side of trades for a flat 6 bps fee, while Drift mixes JIT auctions, an onchain order book and an AMM across 50+ perps for 3 to 10 bps. GMX-Solana charges 4 to 7 bps and splits liquidity between its GLV vault and isolated GM pools. When we wrote this, soon after GMX-Solana launched, Solana held about $2.7 billion of the $5.2 billion in onchain derivatives liquidity. Key takeaways: - At the time of writing, Solana held about $2.7 billion of the roughly $5.2 billion in on-chain derivatives liquidity, more than 50%, with Jupiter and Drift the main contributors. - In Jupiter Perps' pool model, JLP, worth about $1.4 billion at the time of writing, is the counterparty to traders, who pay a flat 6 bps to open or close. - Drift sources liquidity from JIT auctions, an on-chain limit order book and an AMM, and lists 50+ perpetuals with fees of 3 to 10 bps based on trader tier. - GMX-Solana, a modified GMX V2, charges about 4 to 7 bps and had done over $2.4 billion in total volume with about $6.5 million TVL at the time of writing. - At the time of writing, capital efficiency (volume divided by TVL) was about 0.59 for GMX-Solana, 0.38 for Jupiter and 0.15 for Drift, where above 1 counts as good. - [Navigating the Complexities of Onchain Development with Enso](https://castlelabs.io/research/navigating-the-complexities-of-onchain): Why onchain development is slow and costly, and how Enso Shortcuts and its Intent Engine help 75+ projects ship faster, with over $11b in intent volume settled. (2025-04-01; by Castle Labs; tags: Deep Dives, Markets) - [MegaETH: From Alpha To Omega](https://castlelabs.io/research/megaeth-from-alpha-to-omega): How MegaETH works as a real-time Ethereum L2: specialized sequencer and replica nodes, dual-block architecture, EigenDA, and the MegaMafia projects on it. (2025-03-17; by Castle Labs; tags: L2s, Markets) Short answer: MegaETH is an Ethereum L2 built for real-time speed. A single sequencer on enterprise-grade hardware orders and executes transactions, replica, full and prover nodes check its work, and EigenDA stores the data. It produces mini-blocks every 1-10ms alongside 1-second EVM blocks, and targets 100,000 TPS and 10+ Ggas/s. At the time of writing its mainnet was not live, and it had raised a $20m seed round at a $200m fully diluted valuation plus $10m on Echo, while its MegaMafia incubator had produced 17 projects from its Chiang Mai cohort. Key takeaways: - MegaETH targets 100,000 TPS, 10+ Ggas/s and sub-10ms block times by running a single active sequencer on enterprise-grade hardware with 100 cores and up to 4 TB of RAM. - MegaETH's dual-block architecture produces mini-blocks every 1 to 10ms for real-time applications alongside 1-second EVM blocks that keep compatibility with existing EVM tools. - MegaETH's sequencer publishes blocks, witnesses and state diffs to EigenDA, which had 15MB/s of throughput at the time of writing but adds trust assumptions compared with Ethereum DA. - MegaETH raised a $20m seed round at a $200m fully diluted valuation led by Dragonfly, plus $10m on Echo and 5000 ETH (about $13.5m) from its Fluffle NFT sale. - MegaETH's MegaMafia incubator produced six projects from its first Berlin cohort and 17 from a four-week Chiang Mai programme with 50 builders during Devcon 2024. - [Arbitrum’s Path to a Stage 2 Rollup: How BoLD Enhances Dispute Resolution, Validation, and Security](https://castlelabs.io/research/arbitrums-path-to-a-stage-2-rollup): How BoLD moves Arbitrum toward a Stage 2 rollup: permissionless validation, time-bound dispute resolution, and bonds that malicious challengers lose in full. (2025-03-05; by Castle Labs; tags: L2s, Markets) - [A Comparative Analysis of Arbitrum’s Positioning within the Layer 2 Landscape](https://castlelabs.io/research/a-comparative-analysis-of-arbitrums): Arbitrum vs Base, Optimism, Mantle and Blast on TVL, active addresses and stablecoins. Arbitrum leads L2 stablecoin supply with $5b against Base's $3.7b. (2025-02-04; by Castle Labs; tags: L2s, Markets) Short answer: In our comparison with Base, Optimism, Mantle and Blast, Arbitrum still leads on stablecoin supply, with $5b against Base's $3.7b, but Base has passed it on most other metrics. At the time of writing, Base led TVL with $3.6b to Arbitrum's $2.9b, after Arbitrum held first place until October 2024, and Base averaged 1.3M daily active addresses over seven days against Arbitrum's 428K. Base also overtook Arbitrum in DEX volume in September 2024 on the back of Aerodrome, while Uniswap handles over 50% of Arbitrum's weekly DEX volume. Key takeaways: - Among the L2s compared, Base led TVL with $3.6b at the time of writing, ahead of Arbitrum at $2.9b, which had led until October 2024. - Base's 7-day average of 1.3M daily active addresses was almost 4x Arbitrum's 428K at the time of writing, though Castle Labs calls daily active addresses an unreliable usage metric. - Arbitrum was the leading L2 for stablecoin supply at the time of writing, with $5b against Base's $3.7b, even after excluding USDC deposited in the Hyperliquid bridge. - Base overtook Arbitrum in DEX volume in September 2024 on Aerodrome's growth, while Uniswap handled over 50% of Arbitrum's weekly DEX volume ($4 billion) at the time of writing. - At the time of writing, Arbitrum had struggled for weeks to reach $1m in weekly fees while Base consistently generated over $5m, helped by a higher base fee. - [The Evolution of ve(3,3) DEXs](https://castlelabs.io/research/the-evolution-of-ve33-dexs): How ve(3,3) DEXs work: Solidly's tokenomics, why V1 struggled, and how Velodrome, Ramses and Hermes refined it, with over 55% of VELO supply locked. (2025-01-16; by Castle Labs; tags: Deep Dives, Markets) Short answer: ve(3,3) is the DEX design Solidly launched on Fantom in February 2022, combining Curve's vote-escrow locking with OlympusDAO's (3,3) staking. Holders lock tokens to vote on which pools receive emissions, voters earn those pools' fees, and LPs earn the emissions, so rewards follow fees rather than TVL. Solidly V1 fell into a death spiral, and forks refined the model. At the time of writing, Velodrome users had locked over 55% of the total VELO supply with an average lock of 3.65 years, and its Base fork Aerodrome held $1.7b in TVL. - [An Overview of Arbitrum's DeFi Ecosystem](https://castlelabs.io/research/an-overview-of-arbitrums-defi-ecosystem): How Arbitrum's DeFi works across perps, options, DEXs, lending, stablecoins and RWAs. Aave holds $2.2b TVL, GMX $780m and Uniswap $369m with $14m monthly fees. (2025-01-06; by Castle Labs; tags: DeFi, L2s, Deep Dives) - [2025 in the sign of AI?](https://castlelabs.io/research/2025-in-the-sign-of-ai): AI agents like ai16z and Virtuals open 2025 as a sector, Ethena's USDe passes $5.8b to rank 3rd among stablecoins, and Bitcoin ETFs log $663m in outflows. (2025-01-03; by Castle Labs; tags: L2s, Markets) - [Smart Accounts: A New Era for UX](https://castlelabs.io/research/smart-accounts-a-new-era-for-ux): How ERC-4337 smart wallets fix seed phrase and gas UX, with 19.5M account abstraction wallets and 102M user operations, versus EOAs and custodial wallets. (2024-12-20; by Castle Labs; tags: Deep Dives, Markets) Short answer: Smart accounts move wallet control from a single private key into a smart contract, so users keep self-custody and gain social logins, guardian-based recovery, spending limits, batched transactions and gas paid by a paymaster or in ERC-20 tokens. ERC-4337 adds this without changing Ethereum, using bundlers that submit user operations to a shared EntryPoint contract, which is also a single point of failure. At the time of writing, over 19.5 million ERC-4337 wallets had been created since the standard launched in March 2023, with 102 million user operations and about 88% of them paid for by paymasters, but only 0.3% of the earliest wallets were still active. - [BTC stumbles at $100K, Hyperliquid’s $2B Boom and Pudgy Penguins’ Airdrop Claims](https://castlelabs.io/research/btc-stumbles-at-100k-hyperliquids): Hyperliquid's TVL jumps 269% in a week to $2.21 billion, Bitcoin ETFs add $1.13 billion, Pudgy Penguins airdrop claims open, and the case for AI on Arbitrum. (2024-12-19; by Castle Labs; tags: Trading, L2s) - [Institutional Buyers, DeFi Strength, and Alameda are Back?!](https://castlelabs.io/research/institutional-buyers-defi-strength): Bitcoin spot ETFs added $3.52B in a 7-day inflow streak, Trump's World Liberty bought ETH, LINK and AAVE, and an Alameda wallet unstaked $41.5M of SOL. (2024-12-12; by Castle Labs; tags: DeFi, L2s) - [Bitcoin reaches $100K and Ethereum ETFs are on Fire!](https://castlelabs.io/research/bitcoin-reaches-100k-and-ethereum): Bitcoin breaks $100,000 for the first time as 9 Ethereum ETFs add 362,474 ETH worth $1.42 billion in a month, plus $812 million in unlocks and Arbitrum Stylus. (2024-12-05; by Castle Labs; tags: L2s, Markets) - [Binance Listing Performance Rebounds as a Mystery Wallet Apes $380M!](https://castlelabs.io/research/binance-listing-performance-rebounds): Binance listings rebound after 2024 listings fell 90%, with ACT and PNUT up over 1200%. BlackRock's IBIT hits $40B in 211 days and a wallet buys $380M of BTC. (2024-11-15; by Castle Labs; tags: Markets) - [Trump’s Win, Bitcoin at New Highs, and the SEC in Retreat!](https://castlelabs.io/research/trumps-win-bitcoin-at-new-highs-and): What Trump's win means for crypto: Bitcoin hit a $77,000 high, the iShares Bitcoin ETF traded $4.1B, a Polymarket whale made $20M, and the SEC outlook shifted. (2024-11-08; by Castle Labs; tags: Markets, Regulation) - [Balancing the Stablecoin Ecosystem: The Success of Delta-Neutral Models with Ethena and Resolv](https://castlelabs.io/research/balancing-the-stablecoin-ecosystem): Ethena vs Resolv: how delta-neutral stablecoins USDe and USR work, comparing single-token and dual-token models on collateral, risk segmentation and insurance. (2024-10-29; by Castle Labs; tags: DeFi, Deep Dives) Short answer: Delta-neutral stablecoins hold crypto collateral against short perpetual futures, so price moves cancel out, and pay holders from staking yield and funding rates. Ethena does this with one token, USDe, backed by ETH and BTC hedged on centralised exchanges, with an insurance fund as backstop. Resolv splits risk across two tokens, with USR as the senior tranche for conservative users and RLP holding the excess collateral, absorbing losses such as negative funding and taking a 30% risk premium on profits. RLP redemptions pause if USR's collateralisation falls below 110%. - [BTC Nearing ATH, AI Memecoin Mania, and More!](https://castlelabs.io/research/btc-nearing-ath-ai-memecoin-mania): How the AI agent Truth Terminal became the first AI millionaire as GOAT rallied 30,000% in 10 days, plus sector performance, memecoin gains and token unlocks. (2024-10-25; by Castle Labs; tags: AI, Markets) - [The Berachain Report](https://castlelabs.io/research/the-berachain-report): How Berachain's Proof of Liquidity consensus works, the roles of BERA, BGT and HONEY, validator game theory, and projects like Infrared, Kodiak and Shogun. (2024-10-24; by Castle Labs; tags: Deep Dives, Markets) Short answer: Berachain is an EVM-compatible Layer 1 that secures itself through Proof of Liquidity, a consensus model built on DPoS. Users earn BGT, a non-transferable token, by staking liquidity in whitelisted reward vaults, then delegate it to validators or burn it for BERA, the gas token. Validators with more delegated BGT get a higher reward weight and choose which pools receive emissions, so applications bribe them with their own tokens. We wrote this after Bartio testnet V2 went live in June 2024, with mainnet expected in Q4 2024. - [Trump’s Token Flops, Solana Soars, and $500M Token Unlocks](https://castlelabs.io/research/trumps-token-flops-solana-soars-and): Weekly crypto recap: Trump's WLF raised only $10M in 12 hours, Solana led sectors with a 7.32% return, $500M in token unlocks, and Aftermath Finance on SUI. (2024-10-17; by Castle Labs; tags: Markets) - [FTX Bankruptcy Plan Approved: Is a Crypto Stimulus Incoming?](https://castlelabs.io/research/ftx-bankruptcy-plan-approved-is-a): FTX's $16B cash distribution could hit crypto markets as over $450M in token unlocks land. Also: how chain abstraction aims to onboard the next wave of users. (2024-10-10; by Castle Labs; tags: Markets) - [CZ Released, Visa Tokenization, and $EIGEN Drop!](https://castlelabs.io/research/cz-released-visa-tokenization-and): Stablecoin supply sits at $172B with USDT holding 70% share, plus Ethena's BlackRock-backed stablecoin, Bittensor's AI incentives and the EIGEN airdrop. (2024-10-03; by Castle Labs; tags: DeFi, Markets) - [Bittensor Starting New AI Surge: Charted and Explained](https://castlelabs.io/research/bittensor-starting-new-ai-surge-charted): What Bittensor is and why TAO is leading an AI recovery, charted, plus Token Sentinel tips to avoid rugs and upcoming unlocks for ETHFI, OP, ENA and VENOM. (2024-09-26; by Castle Labs; tags: Markets) - [Surprise Fed Cuts, Firedancer Unleashed to Testnet at Breakpoint, and More!](https://castlelabs.io/research/surprise-fed-cuts-firedancer-unleashed): What the Fed's 50 bps cut, its first since 2020, means for Bitcoin and altcoins, and how Solana's Firedancer validator client works now that it is on testnet. (2024-09-20; by Castle Labs; tags: Markets) - [The Evolution of Prediction Markets](https://castlelabs.io/research/the-evolution-of-prediction-markets): How prediction markets work, from Augur's failure to Polymarket and Azuro. Polymarket held $87.5M of the sector's $162M TVL in August 2024, driven by election bets. (2024-09-19; by Castle Labs; tags: Markets, Predictions, Deep Dives) Short answer: Prediction markets let users bet stablecoins on future events, holding shares that pay out on the correct outcome and nothing otherwise. Augur, one of the first, launched in 2018 but struggled with Ethereum gas costs, a complicated client and too few markets. As of August 2024 the sector held $162m in TVL, and Polymarket, an order book on Polygon, held $87.5m of it, with two 2024 US election markets making up 84.23% of its volume. Azuro runs a peer-to-pool design where liquidity providers act as the house across many sports markets, and its largest event made up only 0.7% of its volume. - [Token Sentinel Guide and OP Succinct Dive](https://castlelabs.io/research/token-sentinel-guide-and-op-succinct): Token Sentinel spots new tokens early via social signals, like $CATE running from $6M to nearly $20M FDV in a day, plus how OP Succinct adds SP1 ZK proofs. (2024-09-13; by Castle Labs; tags: Markets) - [SEC and CTFC Charges, Trump DeFi Project, Fed Pivot, and More](https://castlelabs.io/research/sec-and-ctfc-charges-trump-defi-project): Weekly crypto recap: SEC charges Galois Capital, the CFTC fines Uniswap Labs $175,000, Trump's DeFi project on Aave, Fed rate cut odds and token unlocks. (2024-09-06; by Castle Labs; tags: DeFi, Regulation) - [Telegram Founder Arrested, Maker Rebrand, EF Spending, and More!](https://castlelabs.io/research/telegram-founder-arrested-maker-rebrand): Weekly crypto recap: Telegram founder Pavel Durov arrested in France, MakerDAO rebrands to Sky with USDS and SKY, Justin Sun withdraws 12k BTC, Sony's Soneium. (2024-08-29; by Castle Labs; tags: Markets) - [Secure Your $EIGEN: How to Choose the Best Validator](https://castlelabs.io/research/secure-your-eigen-how-to-choose-the): What EIGEN does as an intersubjective work token on EigenLayer, and how to pick a validator to stake with by track record, restake concentration, AVS and fees. (2024-05-15; by Castle Labs; tags: Deep Dives, Markets) - [Friend Tech v2: Opportunity or Bloodbath?](https://castlelabs.io/research/friend-tech-v2-opportunity-or-bloodbath): Friend Tech v1 made $13M in fees on $130M volume. How v2, the FRIEND token and clubs work, why airdrop claims require joining a club, and if the model lasts. (2024-05-03; by Castle Labs; tags: Deep Dives, Markets) - [The State of Arbitrum](https://castlelabs.io/research/the-state-of-arbitrum): Arbitrum's 50% L2 market share, $3B TVL and 500+ protocols, plus STIP and LTIPP incentives, Stylus, Orbit chains, EIP-4844 and the March 2024 ARB unlock. (2024-02-26; by Castle Labs; tags: L2s, Deep Dives) Short answer: In early 2024 Arbitrum was the largest Ethereum L2, with a 50% share of the L2 market, a TVL of approximately $3 billion and over 500 protocols, led by DeFi and derivatives. We wrote this ahead of three catalysts: EIP-4844, which could cut data availability costs by up to 90%, the Stylus upgrade that lets developers write contracts in Rust, C and C++, and the March 16, 2024 unlock of 1.11 billion $ARB. The DAO earned about $16.6M in sequencer profits in 2023 and funds growth through the STIP and LTIPP incentive programmes. - [Unlocking Hyperliquid: A Deep Dive into its Future and Fair Valuation](https://castlelabs.io/research/unlocking-hyperliquid-a-deep-dive): Hyperliquid valuation before its token: how the dApp-chain perp DEX works, points trading at $3.53 each, and why $1.1B FDV looks like fair value vs dYdX. (2024-02-23; by Castle Labs; tags: Trading, Deep Dives) Short answer: Before Hyperliquid had a token, we put its fair value at $1.1 billion FDV, half of dYdX's $2.2 billion at the time, and a fee model paying 75% of protocol fees to stakers at a 10% APR reached the same figure. With its points program due to end in May 2024, points traded OTC at around $3.53 each, implying a $350 million FDV if total supply was 100 million tokens. Hyperliquid runs as its own dApp-chain, which lets it liquidate positions atomically with oracle updates and guarantee solvency at the end of every block. - [Blurring the Boundaries between DeFi and TradFi: An Introduction to the Real World Assets sector](https://castlelabs.io/research/blurring-the-boundaries-between-defi): An intro to real world assets: tokenised treasuries, private credit and cross-chain rails like Chainlink CCIP, LayerZero and Axelar, with Maple, Ondo and Frax. (2024-02-02; by Castle Labs; tags: RWAs, DeFi, Deep Dives) Short answer: Real world assets (RWAs) are traditional financial assets, such as US Treasuries, private credit, bonds and real estate, issued as tokens that DeFi protocols can use. Fixed income leads the sector, with Maple running onchain private credit pools underwritten by pool delegates, Ondo selling tokenised money market and bond funds, and MakerDAO holding Treasury bills on its balance sheet. Chainlink CCIP, LayerZero and Axelar move these assets between chains, and Boston Consulting Group estimates tokenised assets could be a $16 trillion market by 2030. - [The Dark Side of Tokenless Protocols: Arbitrum Builders with no ARB Airdrop](https://castlelabs.io/research/the-dark-side-of-tokenless-protocols): Arbitrum builders left out of the ARB DAO airdrop: Rage Trade, Dolomite, Rysk, Shell, Socket, LiFi, Odos and NFTperp, including Dolomite's climb to $18.6M TVL. (2023-12-15; by Castle Labs; tags: L2s, Deep Dives) - [The Evolution of Liquidity Provision: A Deep Dive into DeFi's Architectural Plumbing](https://castlelabs.io/research/the-evolution-of-liquidity-provision): How DeFi liquidity evolved from x*y=k AMMs to Curve, Balancer and Uniswap V3, where LP fees beat V2 by 54% on average, plus hooks and hybrid CLOBs. (2023-12-01; by Castle Labs; tags: DeFi, Deep Dives) Short answer: DeFi liquidity provision began with Uniswap's x*y=k pools, which spread LP capital across an infinite price range, and has moved toward designs that put more of that capital to work and limit impermanent loss. Curve flattened the curve for stable pairs, Balancer allowed weighted pools such as 80/20 ETH/USDC, and Uniswap V3 let LPs concentrate liquidity in a chosen range, where a study by Austin Adams and Gordon Liao found fees beat V2 by 54% on average. Only 48.6% of ETH/USDC V3 LPs turned a profit over impermanent loss, so DEXs added automated liquidity management, and hooks, RFQ and hybrid CLOBs such as Vertex come next. - [The Puzzling Reality of Optimism Grants Pt 2: the Good, the Neutral, and the Questionable](https://castlelabs.io/research/the-puzzling-reality-of-optimism-df2): How Optimism RetroPGF recipients used their OP: Subli DeFi, Blossom Labs, ZachXBT, Polynya and Beaconcha.in, from 188k OP sold for $227k to funding devs. (2023-11-24; by Castle Labs; tags: L2s, Deep Dives) - [Demystifying Options: A 0 to 1 Guide about DeFi Options](https://castlelabs.io/research/demystifying-options-a-0-to-1-guide): A beginner's guide to DeFi options: calls, puts and the Greeks, plus how Aevo, Lyra, Premia, Rysk, Dopex, Panoptic and others compare by liquidity model. (2023-11-10; by Castle Labs; tags: Options, DeFi, Markets) Short answer: An option gives the buyer the right, without the obligation, to buy (a call) or sell (a put) an asset at a set strike price, in exchange for a premium paid to the seller. DeFi options protocols fall into four groups: order books such as Aevo, AMMs such as Premia, Rysk, Dopex and Hegic, structured-product vaults from Thetanuts, Ribbon and Cega, and volatility products such as GammaSwap, Smilee and Panoptic. At the time of writing, DEXs did about $100M a month in options volume against $40B+ on CEXs. - [A Deep-Dive into Data Availability](https://castlelabs.io/research/a-deep-dive-into-data-availability): What the data availability problem is, how EIP-4844 and Danksharding cut rollup costs (calldata is 90% of fees), and how Celestia, Avail and EigenDA compare. (2023-10-26; by Castle Labs; tags: Markets) Short answer: Data availability is the guarantee that every node can access and check the data in a new block, so a rollup sequencer cannot hide invalid transactions. Before EIP-4844, rollups posted that data to Ethereum as calldata, which made up about 90% of rollup fees. EIP-4844 adds cheaper blobs of about 125 kB that nodes can delete after a month, and full Danksharding targets around 16 MB per block. Celestia and Avail are standalone DA blockchains built around data availability sampling, while EigenDA relies on restaked Ethereum validators and offers up to 15MB/s of throughput against Celestia's initial 2MB/s, with weaker security guarantees. - [The Modular Thesis: Scaling Web3 with Rollups](https://castlelabs.io/research/the-modular-thesis-scaling-web3-with): Modular vs monolithic blockchains: how splitting execution, settlement and data availability scales Ethereum, and optimistic vs zk rollup tradeoffs. (2023-10-19; by Castle Labs; tags: L2s, Markets) Short answer: Modular blockchains split execution, settlement and data availability across separate layers, while monolithic chains such as Ethereum and Solana handle all of them in one layer and struggle to scale without giving up security or decentralization. Under Ethereum's rollup-centric roadmap, rollups execute transactions off-chain and post their data to Ethereum, so they inherit its security without running their own validator set. Optimistic rollups such as Optimism and Arbitrum rely on fraud proofs and a 7-day withdrawal wait, and zk rollups use validity proofs that confirm instantly but cost more to generate and verify. At the time of writing, limited data availability capacity was the main bottleneck for Ethereum rollups. - [ERC-7281: A new Token Standard for a Multichain World](https://castlelabs.io/research/erc-7281-a-new-token-standard-for): How ERC-7281 (xERC20) makes bridged tokens safer by letting issuers set bridge mint limits, after bridge attacks caused nearly 69% of funds stolen in 2022. (2023-10-03; by Castle Labs; tags: Deep Dives, Markets) - [The Puzzling Reality of Optimism Grants Pt 1: Introduction and Movements](https://castlelabs.io/research/the-puzzling-reality-of-optimism): Where did Optimism RetroPGF Round 2's 10 million OP go? Onchain tracing shows 35% of grant tokens sold, 25% held and 40% untraceable by recipients. (2023-09-19; by Castle Labs; tags: L2s, Deep Dives) - [Deciphering GMX v2: The Next Wave of Decentralized Perps](https://castlelabs.io/research/deciphering-gmx-v2-the-next-wave): How GMX v2 works: fees halved from 0.1% to 0.05%, isolated liquidity pools replacing GLP, new markets, delta neutral farming and pair trading on perps. (2023-07-03; by Castle Labs; tags: Trading, Deep Dives) Short answer: GMX v2 replaces the single GLP pool with isolated liquidity pools for each pair, so liquidity providers choose which markets to back. Volatile markets are backed by ETH for longs and USDC for shorts, and new funding fees pay the less dominant side to keep open interest balanced. Open/close fees fall from 0.1% to 0.05%, and 8.8% of v2 fees go to the treasury and 1.2% to Chainlink for its low-latency oracles. We wrote this while v2 was still on testnet, when v1 had processed almost $138B in volume since launch. - [What did DAOs do with their $ARB airdrop?](https://castlelabs.io/research/what-did-daos-do-with-their-arb-airdrop): Where 112.8M ARB airdropped to Arbitrum DAOs went: GMX and Uniswap held, Balancer, Radiant and Stargate rewarded LPs, and TreasureDAO funded game builders. (2023-06-23; by Castle Labs; tags: Deep Dives, Markets) ## Topic Hubs - [Market Structure & Capital Flows](https://castlelabs.io/research/topics/markets) — Castle Labs market structure research on liquidity, volatility, tokenisation, capital flows, and how digital asset markets actually move. - [Real-World Assets & Tokenisation](https://castlelabs.io/research/topics/real-world-assets) — Castle Labs research on real-world assets, tokenised markets, institutional adoption, RWA collateral, credit, treasuries, commodities, and equities. - [DeFi & Onchain Finance](https://castlelabs.io/research/topics/defi) — Castle Labs research on DeFi and onchain finance: vaults, lending, stablecoins, yield infrastructure, tokenised assets, and institutional adoption. - [Networks, Rollups & Ecosystems](https://castlelabs.io/research/topics/l2s) — Castle Labs research on Ethereum L2s, app ecosystems, rollup economics, ecosystem strategy, and the networks competing for users and liquidity. - [AI, Agents & Autonomous Finance](https://castlelabs.io/research/topics/ai) — Castle Labs research on AI and digital assets: agentic finance, autonomous curation, AI tokens, verifiable compute, and machine-led markets. - [Perp DEX & Onchain Trading Research](https://castlelabs.io/research/topics/trading) — Castle Labs research on perp DEXs and onchain trading: Hyperliquid, Lighter, Variational, TradeXYZ, RWA perps, options, funding rates and venue microstructure. - [Protocol & Ecosystem Deep Dives](https://castlelabs.io/research/topics/deep-dives) — In-depth Castle Labs research on leading protocols, ecosystems, tokenomics, architecture, adoption, and market positioning. - [Privacy, Identity & Compliance Infrastructure](https://castlelabs.io/research/topics/privacy) — Castle Labs privacy research on Zcash, Railgun, Aztec, FHE, MPC, TEEs, identity, selective disclosure, and compliance-ready privacy infrastructure. - [Prediction Markets & Volatility](https://castlelabs.io/research/topics/predictions) — Castle Labs research on prediction markets, binary outcomes, onchain options, volatility venues, Polymarket, Kalshi, and information markets. - [Policy, Regulation & Institutional Access](https://castlelabs.io/research/topics/regulation) — Castle Labs coverage of digital asset regulation, MiCA, SEC posture, privacy rules, compliance infrastructure, and institutional access. - [Stablecoin Research: Chains, Issuers and Payments](https://castlelabs.io/research/topics/stablecoins) — Castle Labs stablecoin research: which chains hold the supply, how issuers earn, stablecoin chains such as Plasma and Arc, and the neobanks and cards built on them. - [DeFi Lending, Vaults and Curators](https://castlelabs.io/research/topics/lending-and-vaults) — Castle Labs research on DeFi lending and vaults: the largest vault curators, the eight vault categories, fixed-rate lending and how institutions use Morpho and Aave. - [Onchain Options and Volatility](https://castlelabs.io/research/topics/options) — Castle Labs research on onchain options: the largest platforms, perpetual options, options vaults, binaries and prediction markets, and BTC volatility and skew. ## Feeds & Contact - RSS (full article content): https://castlelabs.io/feed.xml - Sitemap: https://castlelabs.io/sitemap.xml - Weekly newsletter: https://research.castlelabs.io - Contact: hello@castlelabs.io · X: @castle_labs · LinkedIn: https://linkedin.com/company/castle-labs-io/