THURSDAY, APR09
1. NVK's quantum math, 2. Bitdeer's 9.45 J/TH SEALMINER A4, 3. FDIC stablecoin framework, 4. DeepMind measures AI manipulation
From Proto and Bitkey - part of the Bitcoin ecosystem at Block, Inc.
1. quantum
NVK, founder of Coinkite, published a detailed breakdown of the quantum computing threat to bitcoin, arguing that the real risk lies not in quantum hardware but in the media cycle that swings between panic and dismissal. He walks through common claims and finds most of them hollow. Quantum mining, for instance, would require roughly 3% of the Sun’s total energy output. The genuine vulnerability is narrower than headlines suggest: about 6.26 million BTC sit in addresses with exposed public keys, per Chaincode Labs estimates, and BIP-360, the leading quantum-resistant proposal, was merged into Bitcoin’s BIP repository in early 2026. But NVK’s sharpest point is that quantum computers have broken zero cryptographic systems while classical math has broken countless ones. “Relying on a single cryptographic assumption for a $2 trillion network is exactly the kind of risk that serious engineering addresses proactively.” The case for upgrading bitcoin’s foundations rests not on a quantum computer that may never arrive but on the historical reality that no cryptographic standard lasts forever.
-EDITOR·OP_DAILY2. efficiency
Bitdeer has launched the SEALMINER A4 series, its most efficient mining hardware to date, with the flagship A4 Ultra Hydro rated at 886 terahashes per second and an energy efficiency of 9.45 joules per terahash, an improvement over the prior A3 generation’s approximately 12.5 J/TH. According to TheEnergyMag, the launch arrives as bitcoin mining hashprice sits in “the low-$30 per PH/s per day range,” described as “the weakest profitability environments ever” for miners. At the A4’s efficiency rating and $0.06 per kilowatt-hour electricity costs, the machine implies a fleet power hashcost of roughly $13.60 per PH/s per day, providing a “relatively wider margin cushion compared with legacy rigs.” Despite the hardware advance, weak economics have broadly cooled equipment demand, with operators prioritizing balance sheet preservation over large-scale fleet expansion. The A4 series positions Bitdeer to benefit when hashprice recovers, while the current environment rewards only the most cost-efficient operations with meaningful margins.
-EDITOR·OP_DAILY3. genius
The Federal Deposit Insurance Corporation has issued a proposed prudential rule framework detailing how supervised depository institutions can issue and manage stablecoins under the GENIUS Act, moving federal stablecoin regulation from legislative intent into operational compliance structure. Reporting by Micah Zimmerman in Bitcoin Magazine describes the move as “signaling expanded federal oversight of dollar-backed digital assets,” as the agency translates GENIUS Act authority into specific capital, liquidity, and disclosure requirements for chartered banks entering the stablecoin market. The framework defines the operational lane for bank-issued stablecoins, requiring any FDIC-supervised institution to meet prudential standards before issuing a dollar-backed token rather than treating stablecoin issuance as an unregulated adjacency to traditional banking. For digital asset policy, the proposal marks the federal government’s clearest assertion yet that stablecoin issuance is banking activity subject to banking supervision, a framing with direct consequences for who can compete in the stablecoin market and on what terms. The rule enters public comment before any final adoption.
-EDITOR·OP_DAILY4. manipulation
Google DeepMind researchers have published a large-scale empirical framework for measuring harmful AI manipulation, using Gemini 3 Pro as the test subject across 10,101 participants in the US, UK, and India. According to the paper, “the frequency of manipulative behaviours (propensity) of an AI model is not consistently predictive of the likelihood of manipulative success (efficacy)” -- a finding with direct implications for how AI safety evaluations are built. The study ran participants through realistic interactions in three high-stakes domains: public policy, financial investment decisions, and health supplement choices. Finance proved most susceptible to AI influence; health showed the least, partly because model safety guardrails made conversations less engaging. Geography proved equally determinative: Indian participants behaved measurably differently from UK and US cohorts in nearly every metric, suggesting that Western-market evaluations cannot be safely generalized. The paper argues that domain-specific, geographically diverse testing should be a baseline requirement before deploying any general-purpose AI model, a standard the industry has not widely adopted.
-EDITOR·OP_DAILYConsider subscribing and sharing OP_Daily with your community.