2025 Technical Trends: Bitcoin Increases OP_RETURN Data Limit, Ethereum's Dual Upgrades, SOL and BNB Performance Optimization, etc
Author | GaryMa, WuBlockchain
The WuBlockchain summarizes key developments in the blockchain technology sector for the year 2025:
Bitcoin: Programmability Debates, Increasing OP_RETURN Data Limit, and BIP-119 (CTV) and BIP-348 (CSFS) Gaining Grassroots Consensus for New Scripting Methods
Core v30 Default Setting Changes: The reformist faction, led by Antoine Poinsot, pushed for a significant increase in the OP_RETURN data limit from 80 bytes to 4MB, marking the substantial opening of Bitcoin as a “data layer.”
Grassroots Consensus on Covenants: BIP-119 (CTV) and BIP-348 (CSFS) received widespread support from developers, aiming to introduce new scripting methods via soft forks to enable non-interactive withdrawals and Vault functions.
First “Physical Examination Report”: Bitcoin Core completed its first-ever public third-party audit in history. The result of zero high-risk vulnerabilities solidified its robustness as a global financial foundation.
Ethereum: Pectra and Fusaka Dual Upgrades and the ERC-8004 AI Standard
Pectra & Fusaka Hard Forks: The Pectra upgrade was completed in May. Pectra is one of the hard forks involving the largest number of EIPs (11 EIPs) in Ethereum’s history. It further optimized upon the Dencun upgrade, aiming to enhance user experience (UX), optimize validator operations, and drive Layer 2 expansion. The Fusaka upgrade was completed in December, raising the mainnet Gas Limit to 60M. Fusaka is a significant step in Ethereum’s scaling roadmap: enhancing L1 performance, expanding blob capacity, improving rollup cost efficiency, and bringing UX improvements. It also introduced the Blob Parameter Only (BPO) fork mechanism for safely increasing blob capacity when rollup demand rises.
AI Agent Benchmark Protocol: The dAI team was established and implemented ERC-8004 to support AI Agents in achieving seamless transactions within the Ethereum ecosystem, with the goal of building decentralized AI infrastructure.
Privacy Roadmap Iteration: The original team was renamed PSE (Ethereum Privacy Steward), advancing PlasmaFold, an L2 design supporting private transfers, focusing on three major directions: privacy writing, reading, and proving.
Solana: Client Diversification and Consensus Mechanism Redesign, Firedancer Launch, Alpenglow Upgrade, and SIMD-0301 Asynchronous Execution
Firedancer Official Mainnet Launch: The independent validator client developed by Jump Crypto went live, supporting 1 million TPS through a modular parallel architecture and completely resolving the risk of single-client dependency.
Alpenglow Consensus Architecture: Introduced Votor (Voting and Finality) and Rotor (a single-layer relay block propagation protocol), significantly reducing propagation latency and stabilizing block times.
Throughput Limit Increases: Through SIMD-0256 and SIMD-0286, the block computation limit was gradually increased from 60 million to 100 million CUs; simultaneously, SIMD-0301 (Asynchronous Execution) was advanced to remove the replay path from the consensus critical path.
Sponsored by FinTax
BNB Chain: Advancing Toward “Sub-second” Block Generation and Extreme Low Costs, Fermi, Maxwell, and Lorentz Series Hard Forks
Fermi Hard Fork (Nov 2025): The core technical breakthrough lies in reducing the block interval from 750 milliseconds to an extreme 450 milliseconds, achieving industrial-grade throughput through sub-second block generation.
Maxwell and Lorentz Upgrades: Phase-by-phase optimization of BSC block time to 0.75 seconds, with final confirmation time reduced to 1.875 seconds, while simultaneously introducing the EIP-7702 smart contract wallet standard.
Economic Model Fee Reduction: A validator proposal reduced the minimum Gas fee from 0.1 Gwei to 0.05 Gwei, aiming to achieve a cost of $0.001 per single transaction.
Hyperliquid: A Leap from Trading Platform to Programmable Finance L1, CoreWriter Contracts, Read Precompiles, and HIP-3
HyperEVM Bi-directional Interaction: Successfully deployed the CoreWriter system contract, allowing the smart contract layer to directly write to and change the state of HyperCore (the trading engine).
Atomic Data Reading: Launched Read Precompiles, allowing contracts to read on-chain order books, prices, and position data in real-time without the need for oracles.
HIP-3 Permissionless Deployment: Enabled network upgrades allowing developers to create perpetual contract markets permissionlessly via Dutch auctions after staking HYPE.
Base: BaseApp Social Terminal and x402 AI Payment Protocol, Block Gas Target Expanded Five-fold to 112 Mgas/s
BaseApp Ecosystem: Integrated AI Agents, Mini Apps, and Farcaster streaming to create the first “Super App” entry point for Web3.
x402 Protocol V2: An open-source protocol based on the HTTP 402 status code, providing a registration-free, wallet-identity-based native payment layer for AI bots, supporting instant micro-payments with stablecoins.
Performance Scaling: The Block Gas Target was steadily increased from approximately 20 Mgas/s to 112 Mgas/s, supporting high-frequency social and payment activities.
Follow us
Twitter: https://twitter.com/WuBlockchain
Telegram: https://t.me/wublockchainenglish





Superb roundup of what actually shipped in 2025. The Bitcoin OP_RETURN jump to 4MB is wild, basically saying data inscriptions aren't going away so lets just make the infra handle it. Firedancer going live on Solana is probably the biggest single resliency win this year, dunno how the nework survived this long on one client implementation wihtout a serious failure mode.
Solid rundown of what's actually shipping this year instead of roadmap vapor. The BNB Chain push to 450ms block times is bold, tho I'm curious how much real-world latnecy variance there is under load. Fermi's sub-second finality could be a game changer for apps that need near-instant confirmations, something I've seen demand for in payment and gaming contexts. The real test will be whether all this throughput translates to actual user activity or just stays as unused capacity on benchmarks.