Artículos Relacionados con Architecture

L2 'Recalibration': When L1 Becomes Its Own Rollup, What Is Ethereum's Endgame?

The article discusses the evolving relationship between Ethereum's Layer 1 (L1) and Layer 2 (L2) solutions, moving beyond the initial "L2 for scaling" model. As Ethereum L1 itself scales (increasing Gas Limit, statelessness, zkEVM), the unique value proposition of L2s shifts from merely providing cheap execution to offering differentiated features like application-specific optimization, privacy, and flexible governance. The piece explores three key themes: 1. **L2's New Role:** L2s are transitioning from a pure scaling technology to a spectrum of execution environments with varying degrees of security inheritance from Ethereum L1. 2. **Interoperability as State Trust:** Solving L2 fragmentation is less about cross-chain bridges and more about enabling faster, trust-minimized state verification between environments. This involves initiatives like faster L1 finality, intent-based architectures (Open Intents Framework), and native account abstraction. 3. **Blurring Layers:** With the potential integration of zk-proofs into L1 validation (making L1 akin to its own "Rollup") and the concept of "Native Rollups," the rigid boundary between L1 and L2 may fade. The future could be a unified system with multiple execution domains (for DeFi, gaming, privacy, etc.) sharing a common security, settlement, and state framework. In conclusion, Ethereum's goal is not to abandon L2s or re-centralize everything on L1, but to re-integrate the fragmented user experience—liquidity, accounts, applications—while preserving the scaling benefits of a multi-environment ecosystem. The endgame is a cohesive "one chain" feeling for users, powered by diverse but securely interconnected execution layers.

marsbit07/21 12:54

Visa-Mastercard Stablecoin Debate Puts XRP Ledger Design Back In Focus

A new consortium-backed stablecoin project called Open USD (OUSD), involving major firms like Visa, Mastercard, and Ripple, has reignited discussions about the XRP Ledger's (XRPL) design. Notably, former Ripple engineer Matt Hamilton highlighted that OUSD's multi-lateral reserve-sharing structure resembles concepts present in XRPL's architecture since 2012. This is a technical comparison, not an announcement of OUSD launching on or being built by Ripple. The project's initial rollout is expected on Solana and Tempo. For the XRP community, the significance lies in the validation of XRPL's early design principles for fast settlement and asset issuance, now being revisited by major financial players in the stablecoin space. The debate underscores XRPL's historical foresight but does not guarantee direct integration or usage. The focus should remain on the technical parallels and the broader utility narrative for XRPL, avoiding unwarranted claims of direct involvement.

bitcoinist07/02 23:06

Deforming the Transformer, LLMs Become Smarter

A new research paper proposes "Tapered Language Models (TLMs)," a method that improves large language model performance without adding any parameters. It challenges the standard Transformer design where each layer has the same number of parameters ("feed-forward network" width). Building on evidence that layers are not equally important—earlier layers handle foundational information like grammar, while later layers often reinforce existing judgments—the researchers suggest reallocating model capacity from later to earlier layers. The core idea is to make the layer width taper off monotonically from start to end, keeping total parameters and compute constant. Experiments compared linear, cosine, and sigmoid tapering curves on a 440M parameter model. The cosine curve (e.g., starting width 1.5x baseline, ending 0.5x) achieved the best result, reducing perplexity by 1.84 points compared to the uniform baseline—a significant gain at zero cost. This finding proved robust across four different model architectures (including gated attention and memory-augmented models) and at larger scales (760M and 1.3B parameters), consistently improving performance on commonsense reasoning and language modeling tasks without harming long-context retrieval ability. The work highlights a long-overlooked design dimension: optimal parameter allocation across depth. It offers a "free lever" for efficiency, potentially applicable beyond language models to vision Transformers and diffusion models. The study was conducted by researchers from Mila, Cornell University, and the University of Montreal.

marsbit06/29 12:53

He Kaiming's Team's New Work: After Deleting VAE and Private Data, Text-to-Image Generation Becomes Even Stronger

KaiMing He's team introduces **MiniT2I**, a minimalist text-to-image (T2I) model that challenges the complexity of mainstream approaches. It eliminates components commonly considered essential: the VAE encoder-decoder, AdaLN conditioning mechanisms, auxiliary losses, private training data, and post-training alignment stages like RL/DPO. Instead, it uses a pure flow-matching objective trained directly on RGB pixels. The model employs a simplified **MM-JiT** Transformer architecture. It removes AdaLN blocks for conditioning and instead prepends two lightweight text adapter blocks to a standard pre-norm Transformer, allowing frozen T5 text features to adapt to the denoiser. Training follows a two-stage, LLM-like paradigm using only public datasets: pre-training on LLaVA-recaptioned CC12M for coverage, followed by fine-tuning on ~120k high-quality image-text pairs. With just 258M parameters (B/16), MiniT2I achieves competitive scores (0.87 on GenEval, 84.2 on DPG-Bench), outperforming larger pixel-space models. Scaling to 912M parameters (L/16) yields results comparable to SD3-Medium (~2B parameters) in style, composition, and imagination, though it lags in text rendering and named entities due to public data limitations. Key advantages include lower computational cost (~570 GFLOPs vs. ~1379 for latent models) and architectural simplicity. Acknowledged limitations include patch boundary artifacts in pixel space, side effects of high CFG scales, resolution ceilings for sequences longer than 1024 tokens, and the aforementioned data bottlenecks. The work demonstrates that high-performance T2I generation is possible with a radically simplified, publicly reproducible baseline.

marsbit06/22 10:17

Uniswap v4 Hook Analysis: Architecture Design, Common Vulnerabilities, and Protection Practices

Uniswap v4's Hook mechanism is a major innovation, enabling custom logic injection into liquidity pool lifecycle events like swaps and liquidity provisioning. This transforms the AMM into programmable infrastructure, shifting the security model from protocol-level to pool-level, as each pool's safety now depends on its bound Hook contract. The core architecture revolves around the singleton PoolManager contract, which manages all pools via a flash accounting system. State changes are tracked in transient storage and must be settled by the end of a transaction. Hook contracts are permanently bound to pools via a PoolKey, with their permissions encoded directly into their address via specific low-order bits. This design introduces unique security considerations and challenges for future upgrades. Key vulnerabilities and best practices identified include: - **Access Control Gaps:** Early versions of the BaseHook abstract contract only protect `unlockCallback()`, leaving other lifecycle functions (`beforeSwap`, `afterSwap`, etc.) exposed unless explicitly secured by developers. - **Unrestricted Pool Binding:** The `initialize()` function does not validate if a Hook "consents" to a new pool. Hooks must implement their own whitelisting in `beforeInitialize` to prevent unauthorized pool creation. - **Async/Custom Curve Hooks:** These high-risk Hooks can completely replace Uniswap's swap logic. Their security depends entirely on their own implementation, as they operate outside the native protocol's pricing safeguards. - **Delta Accounting Risks:** The system ensures final balance (NonzeroDeltaCount == 0) but cannot guarantee the *correctness* of intermediate delta states, which attackers could manipulate. - **Token Confusion:** Protocols must implement semantic validation for tokens in user-created markets, not just interface checks, to prevent cross-market confusion attacks. The article emphasizes that Hook auditing requires a "sub-protocol" approach due to extended interaction chains, highlighting a significant shift in security methodology for the v4 ecosystem.

marsbit06/22 08:05

Xpeng and NIO Compete on Computing Power, Li Auto Shifts Architecture

On June 15, 2026, Li Auto unveiled details of its self-developed chip, Mahe M100, for its new L9 Livis model. CTO Xie Yan stated the goal was not just a faster chip, but a fundamentally different one, targeting the chip architecture itself. While competitors like NIO, Xpeng, and Huawei highlight TOPS (computing power) figures for their self-developed chips, Li Auto’s Mahe M100 focuses on redesigning the underlying architecture. It employs a "dynamic data flow architecture" to address memory bandwidth bottlenecks in large model inference, claiming up to 3x the effective computing power of Nvidia's Thor U for its specific workloads and a 40% reduction in latency. The chip's design was peer-reviewed and accepted at ISCA 2026. However, this performance is highly optimized for Li Auto's own VLA2.1 algorithm, meaning it may not generalize as well to other tasks. Li Auto aims to achieve full-stack in-house development with Mahe M100, covering chip, compiler, OS, AI algorithms, and domain controller—a level of vertical integration few competitors match. Beyond the chip, CEO Li Xiang introduced a new strategic narrative: the "embodied intelligent vehicle," defined as an integration of an EV, a professional driver, an AI computer, and a life assistant. This shifts competition from features like large screens to systemic AI capabilities. A key commitment was that Li Auto's Mahe VLA autonomous driving model will match Tesla's FSD V14 by Q4 2026, with specific OTA milestones set for July, September, and December. Financially, Li Auto faces pressure with declining revenue and vehicle gross margins since Q4 2025, while maintaining high R&D investment (approx. ¥12B in 2026, 50% AI-related). Its 2026 sales target is 550,000 vehicles, up from 406,000 in 2025. The new L9 Livis garnered over 10,000 pre-orders in two weeks. The effectiveness of these strategic moves—new products, OTAs, and the novel chip architecture—will begin to show in Q3 2026 financial results, with the year-end FSD V14 benchmark being the ultimate test.

marsbit06/16 04:52

After the Passage of the GENIUS Act and the CLARITY Act, What Is the Correct Architecture for On-Chain Yield?

The article discusses the evolution of on-chain credit, distinguishing three markets: overcollateralized crypto lending, unsecured lending (largely unsuccessful), and asset-backed credit (ABC). ABC, backed by identifiable real-world collateral with legal recourse, is identified as the fastest-growing category and the only one credibly addressing adverse selection—the core problem in credit where the riskiest borrowers self-select. Current growth in on-chain Real World Assets (RWAs), particularly tokenized private credit funds (e.g., Maple Finance, Centrifuge), is substantial but often merely "wraps" existing fund structures, inheriting their risks rather than solving adverse selection at the protocol level. The regulatory landscape is a key driver, with the US GENIUS Act (prohibiting stablecoin issuers from paying yield) and the proposed CLARITY Act (closing loopholes on indirect yield) set to redefine permissible yield-bearing products. This makes vaults (like ERC-4626) the critical architecture—they become the primary compliant vehicle for delivering yield, functioning as issuance, disclosure, distribution, and recovery mechanisms. The author's thesis is that the correct post-GENIUS/CLARITY architecture involves building ABC solutions where credit assessment, structure, and recovery are encoded directly into the smart contract vault layer, moving beyond mere tokenized fund wrappers to solve adverse selection fundamentally and ensure regulatory compliance.

Foresight News06/11 11:13

Crossing the 'Memory Wall': The Wafer-Level Revolution and Computing Power Routes in the AI Inference Era

In 2026, a historic shift occurred in AI as major cloud providers' inference spending surpassed training spending for the first time, signaling a move from "building large models" to "using large models." This shifts the core challenge from computing power to the "memory wall"—the bottleneck of data movement (model weights, activations, KV Cache) between external DRAM and processors, where energy and latency from data transfer far exceed computation itself. Companies like Nvidia face GPU idle time due to bandwidth limits. In contrast, Cerebras Systems adopts a radical "wafer-scale" approach with its Wafer-Scale Engine (WSE). Instead of cutting a silicon wafer into many chips, Cerebras uses almost the entire wafer as one massive chip (WSE-3). This design provides 44GB of on-chip SRAM, delivering memory bandwidth thousands of times higher than traditional HBM (e.g., 21 PB/s vs. Nvidia B200). For LLM inference, weights are streamed layer-by-layer from external MemoryX storage to the chip, avoiding HBM bottlenecks. This results in token generation speeds 1.5–5 times faster than Nvidia's B200 in some models and significant advantages in first-token latency and long-context tasks. Additionally, Cerebras's architecture offers much lower interconnect power consumption (0.15 pJ/bit vs. GPU's ~10 pJ/bit). However, Cerebras faces challenges: SRAM scaling has slowed with advanced nodes, limiting future capacity gains; the chip requires specialized liquid cooling and custom software stacks; and its external I/O bandwidth (150 GB/s) is low compared to NVLink, hindering multi-system scaling for very large models. Competition is intensifying. Major players are pursuing three paths: 1) Developing proprietary inference ASICs (e.g., Google TPU, Microsoft Maia), 2) Leveraging advanced packaging (e.g., TSMC's SoW) to democratize wafer-scale-like integration, potentially eroding Cerebras's process advantage within a few years, and 3) Exploring optical interconnects for ultimate bandwidth. Commercially, Cerebras is transitioning from a hardware vendor to a service provider, facing the immense challenge of building high-power, specialized data centers to meet large contracts (e.g., 250MW/year from 2026–2028). In conclusion, the AI inference era presents a fundamental architectural trade-off. Cerebras opts for extreme physical optimization for low-latency, single-task performance, while Nvidia prioritizes versatility and massive cluster throughput. The path forward remains uncertain, with technology and business models still evolving in the race toward advanced AI.

marsbit06/05 11:07

GitHub, Transfixed by AI

On the night of February 9th, GitHub suffered a major outage caused by a simple configuration change—reducing a cache refresh interval from 12 to 2 hours—that triggered a cascade of failures. This was not an isolated event, but part of a broader pattern. In early 2026, GitHub experienced at least 8 major incidents, failing to meet its promised 99.9% availability. These outages stemmed from structural issues: explosive growth in load, tight service coupling, and insufficient protection against abnormal traffic. This unprecedented load is driven by AI Agents. In 2025, GitHub handled ~1 billion commits. By 2026, weekly commits reached 275 million, projecting to ~14 billion for the year—a 14x increase. AI tools like Claude Code now contribute 4.5% of all public repository commits, with weekly submissions surging 25x in just three months. AI-generated pull requests jumped from 4 million to 17 million per month in half a year. Unlike human developers, AI Agents work continuously, generating commits at a scale that overwhelms infrastructure designed for human rhythms. The surge also shattered GitHub's business model. Copilot's flat-rate pricing, based on assisting human developers, became unsustainable as Agentic AI sessions consumed resources worth hundreds of dollars for a few dollars in fees. In response, GitHub imposed usage limits and, by June 1st, shifted to a pay-per-use "AI Credits" system. Facing this new reality, GitHub realized a 10x scaling plan was insufficient. It announced a need to *redesign* its architecture for 30x current scale—decoupling services, adding fault isolation, and improving change management to prevent cascading failures. Other platforms like Stripe and AWS are facing similar challenges with AI Agents. Fundamentally, GitHub is transitioning from a human collaboration platform to an "exhaust pipe" for automated AI workflows. Its detailed post-mortem reports aim to maintain trust during this turbulent rebuild. The February outage was not just a technical glitch, but a signal of the software industry's entry into a new, AI-driven era.

marsbit06/04 10:40

跨链桥主动求变，LI.FI用意图架构成为TradFi机构流动性中枢

marsbit06/03 06:35