Welcome to The Protocol, CoinDesk’s weekly wrap of the most important stories in cryptocurrency tech development. I’m Margaux Nijkerk, a reporter at CoinDesk.
In this issue:
* Ethereum Developers Prep for Fusaka, Second Upgrade of 2025
* Anthropic Research Shows AI Agents Are Closing In on Real DeFi Attack Capability
* Ethereum Devs Push ZK ‘Secret Santa’ System Toward Deployment
* Bitnomial Prepares to Debut First CFTC-Regulated Spot Crypto Market
Unknown block type “divider”, specify a component for it in the ‘components.types’ option
Network News
FUSAKA WILL GO LIVE ON ETHEREUM: Ethereum developers are preparing for the network’s second upgrade of 2025 to go live later today. Fusaka – a blend of the names Fulu + Osaka – consists of two upgrades happening on Ethereum’s consensus and execution layers at the same time. The goal of the upgrade is to enable Ethereum to handle the large transaction throughput from the layer-2 chains that use the blockchain as their base layer. Fusaka consists of 12 code changes, also known as “Ethereum Improvement Proposals” (EIPs) that will make the layer-2 experience faster and cheaper. The biggest change in Fusaka is known as PeerDAS, which allows validators to check only segments of data instead of full “blobs,” easing bandwidth demands and lowering expenses for both validators and layer-2 networks. Layer 2s currently submit thousands of transactions to Ethereum via “blobs,” where validators currently on the Ethereum blockchain have to download all of the transaction data from the blob to verify it is accurate, creating bottlenecks. With this improvement, those validators will only need to verify a fraction of a blob, speeding up the process and lowering the transaction fees that come with it. — Margaux Nijkerk Read more.
ANTHROPIC STUDY ON DEFI AI AGENTS: AI agents are getting good enough at finding attack vectors in smart contracts that they can already be weaponized by bad actors, according to new research published by the Anthropic Fellows program. A study by the ML Alignment & Theory Scholars Program (MATS) and the Anthropic Fellows program tested frontier models against SCONE-bench, a dataset of 405 exploited contracts. GPT-5, Claude Opus 4.5 and Sonnet 4.5 collectively produced $4.6 million in simulated exploits on contracts hacked after their knowledge cutoffs, offering a lower bound on what this generation of AI could have stolen in the wild. The team found that frontier models did not just identify bugs. They were able to synthesize full exploit scripts, sequence transactions and drain simulated liquidity in ways that closely mirror real attacks on the Ethereum and BNB Chain blockchains. The paper also tested whether current models could find vulnerabilities that had not yet been exploited. GPT-5 and Sonnet 4.5 scanned 2,849 recently deployed BNB Chain contracts that showed no signs of prior compromise. Both models uncovered two zero-day flaws worth $3,694 in simulated profit. One stemmed from a missing view modifier in a public function that allowed the agent to inflate its token balance. Another allowed a caller to redirect fee withdrawals by supplying an arbitrary beneficiary address. In both cases, the agents generated executable scripts that converted the flaw into profit. Although the dollar amounts were small, the discovery matters because it shows that profitable autonomous exploitation is technically feasible. — Sam Reynolds Read more.
