# Bài viết Liên quan Engineering

Trung tâm Tin tức HTX cung cấp những bài viết mới nhất và phân tích chuyên sâu về "Engineering", bao gồm xu hướng thị trường, cập nhật dự án, phát triển công nghệ và chính sách quản lý trong ngành tiền kỹ thuật số.

Agents Have Entered the Harness-Driven Era

The article discusses the significance of the leaked Claude Code from Anthropic, highlighting its revelation of advanced Agent engineering practices centered on "Harness" design. Rather than relying solely on model capabilities, modern AI systems now depend on a structured engineering framework—the Harness—to maximize performance. This framework includes six core components: multi-layered System Prompts, Tool Schema, Tool Call Loop (with Plan and Execute modes), Context Manager, Sub-Agent coordination, and Verification Hooks. The Harness enables tighter integration between training and inference, supports long-chain tool execution, and improves reliability through objective verification. It also drives six key training directions: behavior alignment via System Prompt, end-to-end tool-use training, integrated plan-execute training, memory compression, sub-agent orchestration, and multi-objective reinforcement learning. The shift to Harness-driven development reduces the emphasis on pure prompt engineering, favoring instead multidisciplinary talent with skills in AI, backend engineering, and infrastructure. The market is evolving toward more secure, private, and vertically integrated Agent deployments, with "model shell" companies needing either strong infrastructure or deep domain expertise to compete. Claude Code’s leak underscores that future AI advancements will be shaped by engineering architecture as much as by algorithmic innovation.

marsbitHôm qua 10:11

Agents Have Entered the Harness-Driven Era

marsbitHôm qua 10:11

a16z: Behind the Palantir-ization of Everything, a Doomed Imitation Show

In the article "a16z: The Palantir-ization of Everything and the Doomed Imitation Game," Marc Andrusko critiques the trend of AI startups adopting Palantir's model—deploying engineers on-site, offering highly customized solutions, and securing large contracts. While appealing for landing seven-figure deals, this approach often fails to replicate Palantir’s success. Palantir’s uniqueness lies in its integrated platform, elite talent, and focus on mission-critical sectors like defense and intelligence, where high stakes justify deep customization. Most imitators lack these conditions, risking evolution into service-based consultancies with unsustainable economics. Andrusko advises founders to use embedded engineering as scaffolding for product development, prioritize reusable platform primitives, and honestly assess scalability and margins. True Palantir-like success requires a blend of product vision, strategic market selection, and patient capital—not mere imitation.

marsbit01/19 03:18

a16z: Behind the Palantir-ization of Everything, a Doomed Imitation Show

marsbit01/19 03:18

Manus Joins Meta, Achieving 100x Company Value Growth in One Year: What Did They Do Right?

Meta has acquired AI startup Manus in a deal reportedly valued between $4–5 billion, marking a staggering 100x increase in the company’s valuation in under a year. Founded by Xiao Hong, Manus had previously turned down a multimillion-dollar acquisition offer from another tech giant to pursue its vision of building a general-purpose AI agent. Despite early domestic skepticism—with critics dismissing it as a mere “shell" built atop existing AI models—Manus gained significant traction internationally. It attracted serious attention from major players like Google, Microsoft, and OpenAI, with Google even embedding engineers to help integrate its Gemini models. The company reached nearly $100 million in annual recurring revenue (ARR) prior to the acquisition. Manus succeeded by adopting an “incremental mindset,” positioning itself not as a competitor to foundational model developers but as an application-layer innovator that drives token consumption and expands use cases for AI models. Its strategy focused on solving high-frequency user tasks through engineering-heavy, user-centric product development, creating what insiders describe as a “smartphone-like” platform for AI agents. The acquisition underscores the value of focused execution and first-mover advantage in the emerging AI agent space. It also signals a broader shift: in the AI era, success may depend less on owning core models and more on delivering superior user experiences and capturing early user workflows.

深潮12/30 01:44

Manus Joins Meta, Achieving 100x Company Value Growth in One Year: What Did They Do Right?

深潮12/30 01:44

The Full Story of USDe's Depegging on October 11: A $19 Billion Lesson in Crypto Financial Engineering

On October 11, 2025, USDe, a major yield-bearing stablecoin (YBS), depegged on Binance, triggering a cascade of liquidations and resulting in a record-breaking $19 billion liquidation event in crypto history. While mainstream media termed it a "crypto crash," the incident was fundamentally a massive exposure of tail risks in complex financial engineering. USDe, created by Ethena, is a synthetic dollar protocol that maintains delta-neutral positions by hedging spot assets with perpetual futures contracts, capturing returns from funding rates, staking yields, and basis trades. At its peak, USDe reached a $14 billion market cap, offering APYs as high as 27%, and was touted as an "Internet Bond." However, a significant portion of its growth was driven by leveraged lending on external platforms. Binance’s launch of a 12% APY incentive program encouraged users to employ recursive lending with up to 5x leverage, using Binance’s own USDe/USDT pair as the sole price oracle. This created $8.4 billion in highly leveraged exposure outside Ethena’s core delta-neutral system. The collapse began when Trump announced a 100% tariff on Chinese goods, causing a sharp market downturn. As crypto assets fell, perpetual funding rates turned negative. Large USDe holders sold on Binance, driving its price down. Once it fell below $0.82, it triggered mass liquidations of leveraged positions. In just 23 minutes, USDe plummeted to $0.65 on Binance due to cascading liquidations and liquidity failure. In contrast, on-chain DEXs like Uniswap saw only a brief 2% depeg, and DeFi lending protocols like Aave experienced minimal liquidations due to robust oracle mechanisms. Ethena’s core protocol remained solvent and operational throughout, indicating the failure was specific to Binance’s market structure. The event underscores critical lessons: the dangers of excessive leverage, reliance on single-point price oracles, and the misperception of complex yield products as risk-free savings. It highlights that stability in crypto depends on robust mechanisms, deep liquidity, and sustained confidence—not just financial engineering.

marsbit12/29 09:07