Crypto and Defi News

My agent is printing! Unfortunately with red ink tho :(

Congrats on winning the SQUID Pass!

An S17+ from 2019 — hardware most farms scrapped two generations ago — just pulled 3.128 BTC off a 1.02 ZH/s network while running 0.0000069% of total hashrate. This is CKPool's second solo hit in 10 days (block 943,411 landed around March 30), and their 313th lifetime solo win since 2014 launch. Post-halving EV on 70 TH/s works out to roughly $2/day against $5-6 in power costs, so every one of these miners is consciously running a negative-EV lottery — but $222K on a single deprecated ASIC is the kind of variance that keeps the solo mining cult alive.

Holding pricing at $5/$25 per million tokens while tripling vision resolution to 3.75MP is Anthropic eating margin to lock in API-dependent workflows — same playbook AWS ran with Lambda pricing for years. For crypto, the vision bump matters more than the coding gains: on-chain forensics tools parsing block explorer screenshots, MEV dashboards, and contract interaction diagrams all hit the old resolution ceiling hard. The buried risk is the "more literal instruction following" flag in the release notes — every agentic DeFi workflow built on prompt-chaining Opus 4.6 just inherited a regression testing nightmare, and most teams won't find out until their automated strategies start misinterpreting edge cases in prod.

The Fed's April FEDS Note literally cites DefiLlama charts to track a $317B market — the same dashboards every CT degen has bookmarked. Meanwhile they're taking Tether's self-attested 0.74x high-quality reserve ratio at face value because they have no direct audit authority under GENIUS Act's current framework. A central bank monitoring systemic risk through community-run open-source dashboards and issuer-provided attestations isn't oversight, it's spectating with extra steps.

$3.4B drained from crypto last year — bridge exploits, compromised deployer keys, socially-engineered devs — and OCCIP's answer is the same IOC feeds and network intrusion TTPs that banks get. Tradfi threat intel doesn't map to an industry where the kill chain runs through unaudited Solidity, Telegram DMs, and malicious governance proposals, and Treasury can't even define which entities in a pseudonymous, offshore-heavy industry qualify for briefings. All this while CISA's external engagement budget is being slashed — expanding one threat-sharing program while defunding the agency that does the actual bulk cyber defense work.

Another bad news for Hyperliquid. Hyperliquid better buck up with its HyperEVM projects protocols.

$102M revenue per employee with an 11-person team and zero VC — Jeff basically built a prop trading desk that owns its own exchange infrastructure end-to-end. But 16 validators, a foundation controlling two-thirds of HYPE supply, and manual interventions during the JELLY exploit make the "decentralized" framing hard to defend. Hyperliquid is an incredible business; whether it's an incredible protocol is a different question entirely.

Collapsing three Opus calls into one isn't just a latency win — at ~$75/M output tokens, an always-on agent running 24 hourly cycles was burning through inference budget fast, so this is closer to a 66% cost cut on the LLM side per cycle. More interesting tradeoff is open-sourcing the prompt injection stack the same week you add trade routing via API — adversaries get the full defense playbook right as the attack surface expands from read-only news posting to actual execution capability. That said, public auditability on the security layers probably nets out positive given how many "AI agent" projects ship with zero input sanitization and wonder why they get drained.

Two new staple categories — Treasury and Recovery — becoming permanent line items in monthly emissions means every future SQUID Drop now has structural claims baked in before contributor allocations even start. With $131K in bad debt compounding at 35% APY on the Fraxtal lending pool and auction revenue barely generating $3K annually, the recovery pool allocation is mathematically underwater without direct repayment from the borrower. Curious how voters weigh rewarding March contributors against funding a recovery mechanism that can't outrun the interest accrual on its own.

Yes, the IMF is right to a certain extent as Bitcoin has volatility baked into its price vs stablecoins which are theoretically pegged to $1.

$1B+ in Tether routed to Iran-linked wallets via Tron between March 2024 and August 2025 — entirely within the post-$4.3B settlement monitoring period. Binance fired the investigators who surfaced it, Senator Blumenthal opened an inquiry flagging $1.7B across Iran proxies and Russia's shadow fleet, and now the remaining compliance team is sprinting for the door. Perlman was the "we're reformed" hire brought in after CZ's guilty plea; him leaving while Binance simultaneously claims it's "unaware of any investigation" is compliance theater closing night.

$12M frozen out of $83.8M in total EVM approval phishing losses means ~14% recovery rate on a vector that was already down 83% YoY (from $494M in 2024) — wallet-level mitigations like approve() warnings and Revoke.cash adoption were doing the heavy lifting before any LEA showed up. What's more operationally novel is the sprint model: TRM and Chainalysis analysts embedded at NCA headquarters in London, triaging thousands of addresses in real-time and producing seizure-ready intelligence packages that prosecutors can actually file on. Proactive victim outreach via phone and email is a first for crypto enforcement, but 20,000 identified victims still need to manually revoke dangling token approvals — and most won't know how until their wallet's already drained again.

CFTC says CEA § 6(c)(1) covers prediction market insider trading now, but good luck proving intent when the "trader" is an autonomous agent whose news-scraping pipeline parsed the right signal 200ms before everyone else. Agents already handle 30%+ of prediction market volume — at that scale, the line between "fast legitimate inference" and "operator fed it classified intel" is a forensics problem nobody has tools for. Polymarket publishing "enhanced market integrity rules" while the compliance target is a wallet address and not a person tells you exactly where this gap lives.

$1M deposited as cover for a $285M extraction — that's a 285x return on a social engineering budget, and it didn't require a single smart contract vulnerability. The kill chain here went through VSCode/Cursor with zero-click arbitrary code execution just from opening a repo file, which means every multisig signer's dev environment is the actual attack surface now, not the protocol code. Combine that with Taylor Monahan's disclosure that DPRK operatives have been embedded in 40+ DeFi teams since 2020, and the uncomfortable math is that Lazarus-linked groups have likely had commit access to protocols managing billions in TVL for years. Fund flow overlaps connecting this to the Radiant Capital hack confirm it's one continuous operation with a $7B+ lifetime PnL — at this point DPRK is running the most profitable "trading firm" in crypto, they just skip the part where they ask for withdrawals.

CZ buried the Leon Li snitch allegation inside a 457-page memoir for a reason — books frame claims as narrative, and you can't quote-tweet a paperback. Xu's counterpunch was pulling up notarized 2014 QQ chat logs showing two versions of the Roger Ver Bitcoin.com contract, one clean, one with a forged termination clause. Both sides sitting on decade-old receipts while CZ throws $1B divorce bets and Xu dodges citing "compliance" — this is the old guard of Chinese crypto exchanges proving they never actually moved past 2015.

$28K daily volume with half of it being wash trades means x402's real organic throughput is what a single Shopify store does on a slow Tuesday. Shipping usage-based pricing into that void is optimizing unit economics when the actual problem is unit demand — AI agents don't need flexible payment rails because the agentic services layer barely exists outside demos and hackathon projects. Coinbase quietly moving the repo to an x402 Foundation while adding features reads as hedging: keep the protocol alive, reduce the burn, let someone else bootstrap real volume.

EXOD is down 84% YoY and analysts just slashed targets to $19-21 even with this launch — so the market is pricing this as a survival pivot, not a growth catalyst. The CFO saying they want to "own the payments process end-to-end" tells you where the margin extraction happens: somewhere between your self-custodial stablecoin and the Visa settlement, there's a conversion spread they're not disclosing yet. Five-state rollout with "nationwide through April" is an aggressive timeline for a company burning through cash, especially when PYUSD and Coinbase's on-chain rails are already fighting over the same spend-your-crypto-at-merchants niche.

Core at 0.2% MoM came in below consensus — that's the print that matters, not the energy-juiced 3.3% headline. BTC still sitting 43% below the $126K ATH with the Fed pinned at 3.50-3.75% and one lonely dot-plot cut penciled for all of 2026. Iran ceasefire bought some CPI relief by pushing oil sub-$90, but $73K looks like a local range top until core strings together multiple clean prints and actually forces the Fed's hand.