Cloudflare and MetaMask introduced new wallet frameworks for AI agents in the same week, moving the conversation beyond simple automation and toward direct economic action. On Aug. 4, Cloudflare unveiled Cloudflare Wallets, giving agents virtual wallets that can use USDC and other stablecoins within preset budgets and permissions to buy APIs, data, content, and computing services. Two days later, MetaMask launched Agent Wallet, allowing agents to connect to onchain wallets and carry out swaps, perpetual futures actions, prediction market participation, and liquidity management under user-defined limits. The shift matters because it changes what an AI agent can do in practice. Instead of stopping at analysis or API calls, an agent can now pay for resources, monitor conditions, and in some cases execute transactions with real assets. That creates a new layer of autonomy, but it also links AI mistakes directly to financial loss. The article argues that the next key battleground is not just smarter models, but permission systems: spend caps, protocol whitelists, policy checks, transaction simulation, threat scanning, 2FA escalation, and the ability to pause or revoke an agent at any time. In that setup, wallets stop being just tools for private key storage and become control layers between humans and software acting on their behalf.

ChainFeeds Research’s PRO issue #154 pulls together several research threads across Bitcoin and Ethereum. The report highlights Fei Wu’s work on Ethereum’s post-Glamsterdam fee market, where EIP-8037 and EIP-7999 would shift gas pricing from a single-meter model to separate pricing for execution, data, and state creation. Using historical gas changes, the study models demand curves for each resource and finds different elasticities: execution is least sensitive to price, data sits in the middle, and state creation is the most sensitive. Under higher block capacity, state creation could become the new fee bottleneck. On Bitcoin, the report covers josh’s HTLC design aimed at removing reliance on mempool-based preimage monitoring and the free-relay assumption, as well as a staletip P2P message proposed by Ram and w0xlt to improve visibility into stale blocks. On Ethereum research, Hudu Yusuf questions whether stake weight truly reflects effective consensus weight and whether continuous oracle output is always safer than silence, proposing DW-BFT and a structured-silence model. The issue also includes Sidistr’s Arcanum, a compiler-layer approach for keeping source code hidden through a TEE-to-ZK roadmap, and a paper from Old Dominion University on what drives contentious Bitcoin soft forks toward orderly upgrades or lasting chain splits.

IOSG argues that the current storage rally is being driven by artificial intelligence, but not in the way traditional IT buyers used to think about storage. In its view, the market is no longer rewarding raw capacity first. It is rewarding the ability to keep GPUs fed, move checkpoints quickly, support retrieval-augmented generation with very low latency, and raise overall compute utilization across tightly coupled infrastructure stacks. That shift, the article says, is why components such as HBM, DRAM, CXL, enterprise SSDs, SSD controllers, NVMe pathways, and performance storage software have become central to the AI investment narrative. The piece draws a sharp distinction between AI storage and decentralized storage. AI storage is framed as an efficiency system built for hot data and commercial output. Decentralized storage, by contrast, is described as a trust system for cold data, focused on permanence, censorship resistance, auditability, and public memory. IOSG uses Filecoin and Arweave as the main examples, outlining how the two networks diverge in architecture and product direction, while also listing persistent problems across the sector, including weak enterprise service layers, retrieval limits, supply-demand incentive mismatches, privacy and compliance tensions, and token economics that can amplify market cycles rather than solve product-market fit.

Circle, the issuer of USDC, has rolled out Arc, a purpose-built Layer 1 blockchain designed for stablecoin applications rather than general-purpose crypto activity. The company says the network is meant to fix several constraints it sees in existing chains, including volatile fees, probabilistic settlement, weak privacy controls for sensitive commercial use, and fragmented liquidity across ecosystems. Arc uses USDC as native gas and can also support other stablecoins through a paymaster model. Circle says its fee system borrows from Ethereum’s EIP-1559 design but replaces block-by-block adjustments with a weighted moving average tied to network demand, with fees denominated in USDC and routed to an on-chain Arc Treasury. The network’s public testnet launched in October 2025, while the public mainnet is scheduled to open on September 16, 2026. Circle says Arc is already in a private mainnet phase with more than 100 ecosystem and institutional builders. CEO Jeremy Allaire said in August 2026 that the testnet had processed more than 500 million transactions across nearly 3 million wallets. The company has also named a validator lineup that includes BlackRock, DTCC, Galaxy, Global Payments, ICE, Mastercard, MoneyGram, SBI Group, Standard Chartered, Sumitomo Corporation, Visa, and Circle itself.

Bitcoin Lightning Network self-custody wallet Zeus Wallet has taken its infrastructure offline after a cybersecurity incident on Wednesday, according to Cointelegraph. The company said it is conducting a full system audit and will restore services once that work is complete. Founder Evan Kaloudis said the attack was contained within hours, with no customer fund losses identified and no evidence that Lightning node software was affected. Zeus said the scope was limited to its own infrastructure. Users whose LSP channels were forced to close during the incident will receive replacement channels after service returns. The company has not disclosed the nature of the attack or provided a timeline for resuming operations. Zeus added that the incident will accelerate its security work around Trusted Execution Environments, or TEE, and the Validating Lightning Signer, or VLS, project.

ChainFeeds Research’s PRO issue #153 pulls together a broad set of Bitcoin and Ethereum research updates, led by three themes: a more granular circuit breaker design for Gloas builder markets, a discussion of proprietary AMMs on Ethereum, and a proposal to eliminate part of oracle extractable value through synchronized state transitions. Ethereum researcher Potuz argues that failures in payload delivery should not automatically force validators into self-building, and instead proposes a layered fallback system based on builder trust. Ethereum Foundation researchers Mike Neuder and Maryam Bahrani describe PropAMM as a middle ground between traditional AMMs and RFQ systems, preserving onchain execution while letting professional market makers update pricing parameters more actively. Oraclizer CPTO Jay Kim, meanwhile, frames OEV as a structural timing problem and outlines a model where updates and downstream consequences are bound into a single cross-domain atomic transition. The report also covers a Coldcard hardware signer warning tied to weak seed entropy, two denial-of-service bugs in Core Lightning, a zkPoH concept for proving ownership of at least 1 BTC without exposing specific UTXOs, native randomness sourcing for onchain games, selected items from The MEV Letter #148, and a phishing paper that says transaction simulation itself can become an attack surface.

MoonPay has launched PayBox, a payment vault that connects to Claude and ChatGPT and lets AI agents complete transactions after user approval. The system stores crypto wallets and payment cards in a secure vault, lines up transactions inside the chat flow, and uses a passkey on the user’s device to authorize payment. At launch, PayBox supports restaurant reservations, flight bookings, shopping across major online retailers, token swaps, cross-chain fund transfers, and routing assets into DeFi protocols such as Aave. It starts with support for Solana and seven Ethereum-compatible chains: Ethereum, Hyperliquid, Tempo, Base, Robinhood Chain, Arbitrum, and Polygon. MoonPay also said PayBox works with x402, the open payments protocol developed by Coinbase and later donated to the Linux Foundation, now supported by Visa, Mastercard, Stripe, Shopify, AWS, and Anthropic. The infrastructure comes from Sodot, the Israeli key management startup MoonPay acquired in April for roughly $100 million in an all-stock deal. According to MoonPay, wallet keys are split with MPC and stored across TEEs so no single party, including MoonPay or the AI agent, can access a full key or sign on its own.

MoonPay has rolled out PayBox, a new AI payments tool that can be connected to ChatGPT or Claude, according to The Block. The product is designed to let AI assistants prepare transactions for activities such as booking flights, reserving restaurants, and making online purchases. Users can either require passkey approval for every transaction or set spending limits that allow the AI to complete payments on its own. PayBox combines a crypto wallet with a payment card and supports cross-chain transactions based on the x402 agent payment standard. At launch, it works with Solana and a range of EVM-compatible networks, including Ethereum, Hyperliquid, Tempo, Base, Robinhood Chain, Arbitrum, and Polygon. MoonPay said it plans to add support for more AI platforms and DeFi features, including token swaps and perpetual futures trading, within the next month. The company also said wallet private keys are protected through MPC and TEE, while card information is stored in a way that lets AI agents complete purchases without accessing the underlying card details.