Code always has bugs. The deep ones used to stay hidden because finding them took scarce, expensive human expertise — the code being clean had nothing to do with it. That cost is collapsing. On OpenAI and Paradigm's EVMbench, the best models could exploit under 20% of critical Code4rena bugs six months ago. GPT-5.3-Codex now clears 70%. But read the second number. Its recall across a full codebase is still around 45%. AI is good at punching through a hole once it's flagged. It's weak at deciding which hole, in a whole protocol, is actually the one that matters. The first is grunt work. The second is judgment. This matters more on-chain because there's no undo button. A hacked bank can freeze and roll back. On-chain, once funds move, there's no clean fix — that would defeat decentralization itself. And 2026 showed where the money actually leaks. DeFi lost over $750M by mid-year. Almost none of it was an AI-found code bug. Kelp DAO lost $292M — not a code bug, but a verifier set to 1-of-1, fed false data while every transaction looked valid. Drift's $285M was a multisig social-engineered open. StablR was a minting key too concentrated. None of those are things you read out of the code. You catch them by understanding the protocol: where the permissions are too broad, how fragile the bridge verifier is, whether the oracle dependency makes sense, how much an attacker actually nets from each path. AI surfaces a hundred suspicious lines. Which one is worth months of social engineering to reach is a human call. Auditing splits into two tiers under this. Checklist auditing — scanning line by line against known patterns — AI does in hours, and that tier depreciates fast. The people who understand a protocol's economic structure, its trust assumptions, its real attack paths, get rarer and more expensive. When bug-finding gets cheap, that kind of judgment is what stays scarce, and scarcity sets the price. The security hire worth paying for in the next cycle can look at what AI found and tell you whether it can actually take the money.
@prlnet is interesting because it flips the usual mining logic. In Bitcoin, mining cost helps determine price. In Pearl, price determines whether useful compute even bothers to join the network. The long-term thesis is simple: A GPU already running inference earns real revenue. If Pearl adds PRL rewards on top with minimal overhead, the same GPU cycle can generate two income streams: - Inference fees - PRL emissions At that point, miners are no longer competing with AI workloads. They are the AI workloads. This creates a different flywheel: Higher PRL price → more inference providers integrate → more useful compute joins the network → stronger organic demand for PRL. Lower PRL price → providers leave → the network falls back to ordinary GPU mining. That’s why the key metric isn’t raw hashrate. It’s the percentage of hashrate coming from real AI workloads rather than dedicated miners. The entire Pearl thesis comes down to one question: Can useful compute grow faster than mining sell pressure?
Something I hadn't realized before: Circle's fix wasn't retracted because one address holds 98.2% of at-risk debt. Fixing the curve = liquidating it. A $14.49B protocol's rates, held hostage by one position. Passively. Governance capture without voting is still governance capture. Full research worth reading!
Crypto's Key Events This Week - US Memorial Day; Q1 GDP + Core PCE - CME launches 24/7 crypto futures & options trading - Unlocks: $H (~$22.2M), $XPL, $SAHARA, $JUP - Nordic Blockchain Conference; ETHGlobal Prague; Warren demands SEC probe WLFI ... Full calendar below 👇 https://t.co/cjdtgO3lnC
Web3's Bootcamps, Ambassador Programs, Growth Fellowships, Campus Leader tracks, and "internships" have stopped being talent pipelines. They're industrial attention factories. The old playbook was KOLs, ads, airdrops. The new one is cheaper: recruit thousands of students who want to break into the industry and have them post, repost, run Discord, host AMAs, ship memes. Manufactured grassroots, at near-zero cost. Framed as education, co-building, builder ecosystems. In practice it's free KPI labor for BD, growth, and ecosystem managers. The students think they're interning. The whole industry now optimizes for DAU, engagement, reposts, ambassador count, content volume. Better numbers, easier funding, listings, exchange deals, TGE. The moment incentives stop, retention, TVL, and volume collapse together. Communities turn into bots and ex-contributors overnight. Large parts of Web3 have stopped optimizing for adoption and started optimizing for the appearance of adoption. Once simulated participation is priced higher than real users, the industry forgets how to ship things people actually want.
Over the past six months, crypto has produced a wave of stablecoin card projects. Most people call this a payment boom. I don’t. Most of these new cards don’t have real daily users. They have behavior subsidies: referrals, KOL pushes, first swipes, cashback screenshots, and GMV-shaped data. That is not payment adoption. That is adoption cosplay. A real payment product has boring, repeated, unprompted usage. Most new crypto cards are subsidized usage theatre: - KYC for access - Top up for rewards - Swipe for points - Refer for cashback - Post screenshots for social proof - Wait for a token The card isn’t the product. The activity around it is. This explains why so many DeFi teams suddenly want a card. Users don’t need another way to pay. Projects need a way to turn weak consumer intent into fintech-looking proof: KYC users, deposits, GMV, referral growth, repeat-usage signals. Retention can wait — the deck already looks like consumer fintech. At its core, the stablecoin card wave is narrative refinancing. TVL becomes deposits. A token becomes rewards. Points become loyalty. Off-ramping becomes “real-world adoption.” Tired crypto metrics get laundered into metrics outsiders still reward. This is why Infini-style outcomes don’t stop the cycle. Infini generated attention and usage signals, then became the cautionary tale for weak retention once incentives faded. From a product lens, that path looks bad. From a project lens, it still worked: attention, metrics, fintech narrative, one more cycle. Many teams aren’t trying to build Revolut. They only need to extend the story. A strong exchange or wallet uses a card to monetize existing distribution. A weak DeFi team uses the same form to manufacture the appearance of distribution. One is a monetization layer. The other is a growth prosthetic. Supply keeps coming because consumer crypto has very few believable growth templates left. Wallets are saturated. Yield vaults are tired. SocialFi died. Perps are dominated. AI agents are vague. Cards are concrete, familiar, outsourceable, KOL-friendly, and easy to put in a deck. The stablecoin card boom reveals crypto’s latest adoption machine: Find a product form that creates adoption-shaped data. Attach behavior subsidies. Push through KOLs. Show growth. Raise or relaunch. Let retention decide later. The next wrapper may change. The machine will remain.
1/@HyperliquidX just got Coinbase and Circle to stake HYPE and hand over ~$180M/year in USDC reserve yield. The 8-month story of how it happened is more interesting than the headline. 2/First, the business model that makes this matter. When you hold 1 USDC, Circle holds 1 real dollar somewhere. Circle parks that dollar in US Treasury bills earning ~4%. You get the stablecoin. Circle keeps the 4%. That is how stablecoin issuers make money. 3/Hyperliquid earns trading fees and uses ~97% of them to buy back HYPE from the open market. Every additional dollar of protocol revenue compresses HYPE supply. The token's price action is mechanically linked to how much money the protocol pulls in. 4/Now connect the two. ~$5B of USDC sits on Hyperliquid as trader collateral. At ~4% T-bill yields, that USDC generates roughly $200M a year in interest. Pre-today, every dollar of that went to Circle and Coinbase. Hyperliquid earned zero on the float its own users posted. 5/In September 2025, Hyperliquid moved to fix that. It opened a public bid (RFP) for the right to issue USDH, its own official stablecoin. Whoever won would replace USDC as the venue's default and share the reserve yield with the protocol. Six teams submitted. 6/The offers, ranked by generosity: – Paxos: 95-100% + $20M incentives – Frax: 100%, "zero Frax take" – Agora: 100% of net revenue – Ethena: 95% + ≥$75M (withdrew Sept 11) – Sky: 4.85% direct savings rate + $25M fund – Native Markets: 50% to Assistance Fund 7/Native Markets — the least generous bidder — won with 71.47% of the validator vote. A few facts about that win: – The company was incorporated in early September 2025, days before the RFP – 3 founders, no disclosed funding round – Its onchain wallet was funded hours before the RFP was publicly posted – Its proposal was submitted 90 minutes after the RFP went live 8/Dragonfly partner Haseeb Qureshi called the process "basically custom made for Native Markets" at the time. Agora's founder asked publicly what the point of the bidding even was. The vote happened anyway. USDH launched September 24, 2025. 9/What happened to USDH over the next 8 months: – Peak supply: $97.6M (March 2026) – Supply cap raised to $500M — utilization stayed at 18% – USDC on Hyperliquid grew from ~$2B to ~$5B over the same period – Hyperliquid Labs never deployed a canonical BTC-USDH or ETH-USDH trading pair 10/That last point is the one to sit with. A stablecoin without major trading pairs is unusable. Traders can't post it as collateral against the assets they actually want to trade. USDH was starved by its own parent protocol. The decision that capped USDH was made upstream, not by Native Markets. 11/Today, May 14, 2026: Native Markets agrees to grant Coinbase the right to purchase the USDH brand. Coinbase becomes USDC's "treasury deployer" on Hyperliquid. Circle handles cross-chain plumbing. Both stake 500k HYPE. ~90% of the ~$200M/year USDC yield now goes to Hyperliquid. 12/One reading: USDH was never meant to succeed. It was a credible threat. Three signals: – The least generous bid won: visible enough to threaten Circle, not generous enough to actually flip the market – No canonical USDH trading pair: kept the threat alive, capped its scale – 8-month exit with Coinbase buying the brand: graceful off-ramp 13/Another reading: USDH was a genuine attempt that didn't scale. Native Markets exits with brand-asset compensation. Coinbase becomes deployer because at $5B float there is no other viable USDC partner. Both readings fit the public facts. Pick yours. 14/Either way, the structural result is the same. Bespoke app-stablecoins as a category: dYdX abandoned theirs, GMX never issued, GHO capped at ~$565M, crvUSD at ~$307M, USDH at ~$100M. Only Sky works — because Sky is a stablecoin protocol with apps attached, not the inverse. 15/Yield-sharing went from novel to table-stakes between November 2024 (Paxos USDG) and today (AQAv2). Eighteen months. Coinbase and Circle's commercial agreement renews in August. The next protocol that runs an RFP will know exactly what its quote-asset slot is worth.
1/ Papertrade is the first synthetic perp venue to price entirely off Hyperliquid's order book, and pay Hyperliquid nothing back. A structural read of what's actually being built. ↓ 2 / No order book. No funding. No matching engine. Users post USDC, open a position, and the contract reads Hyperliquid's mid-price. Closing reads it again. The difference is paid out of a single LP pool. That LP is the entire counterparty side of the exchange. 3 / Three cohorts share the casino's economics: — LP depositors earn realized user losses (USDC) — $PAPER stakers earn a share of LP revenue (USDC) — $PAPER holders hold a reflexive claim on more losses arriving Each layer is paid by a different cohort. Each layer prices independently. 4 / The mint rate is front-loaded: 100 PAPER per $1 of user loss while LP < $2M. Decays after. Break-even token price for an early loser, per $1 of realized loss: $0.01.Fundamentals ceiling under reasonable assumptions: ~$0.05–0.10. The early window is real. It's also small. 5 / Papertrade extracts informational value from Hyperliquid's order book (the mid-price) and contributes zero OI back. HIP-3 was Hyperliquid's negotiated answer to exactly this. Papertrade sidesteps it. Loss-backed tokens have a base rate: RLB peaked at $0.26 on a buyback narrative. GLP holders absorbed a $565K oracle exploit. SNX trades 98.9% below ATH. The base rate is not the bull case. The contract will run. The token is a separate trade.
Crypto's Key Events This Week - Senate cloture vote on Warsh as Fed Chair; Powell's term expires Fri - US April CPI; CLARITY Act markup hearing - Unlocks: $AVAX, $APT, $CONX, $STRK, $SEI, $ARB - First prediction market ETF goes effective ... Full calendar below 👇 https://t.co/p2wNoLaSGq
