2026-08-12 Quarta

Notícias de cripto

Mantenha-se a par do mercado de cripto. Notícias em tempo real, análises, preços, histórias em alta e análise de especialistas — tudo num só lugar.

ArthurHayes新文：押注日元升值，ENA未来几月或涨5至10倍

Arthur Hayes argues that the Japanese Yen is significantly undervalued and posits that its appreciation against the US Dollar is imminent. He outlines three potential mechanisms for this shift, dismissing the first two—the Bank of Japan raising interest rates and domestic institutions selling foreign assets—as politically or economically unfeasible. He identifies the third and preferred method: the Japanese Ministry of Finance (MOF) using its holdings of US Treasuries as collateral in the Fed's FIMA repo facility to borrow US dollars, then selling those dollars to buy Yen in the forex market. Hayes believes US Treasury Secretary Bessant has signaled support for this approach, which requires the Fed's Foreign Currency Subcommittee, led by Chairman Walsh, to remove lending limits on the FIMA tool. Hayes asserts that implementing this "Scheme 3" would lead to a significant expansion of US dollar liquidity. He predicts this surge in liquidity will act as a catalyst, driving up the prices of assets like Bitcoin and physical gold. Within the crypto space, he views Ethereum (ETH) as undervalued and singles out Ethena's ENA token as a speculative play with potential for 5-10x gains in the coming months, contingent on a recovery in Bitcoin basis trades that would boost demand for its USDe stablecoin. He concludes that investors should watch for the Fed's rule change as the key trigger for these market movements.

marsbitHá 14m

ArthurHayes新文：押注日元升值，ENA未来几月或涨5至10倍

marsbitHá 14m

CoreWeave Soars! Q2 Revenue Doubles, Exceeds Expectations, Backlog Reaches $104 Billion

CoreWeave's Q2 revenue reached $2.575 billion, a 112% year-over-year increase, surpassing the expected $2.56 billion. The key metric, revenue backlog, reached approximately $104 billion, further increasing from $99.4 billion at the end of Q1. The CEO stated that the expected profit margins for Q2 signed contracts are 5‑10 percentage points higher than those for new contracts signed in recent quarters. CoreWeave's stock price surged 16% in after-hours trading.

Há 15m

CoreWeave Soars! Q2 Revenue Doubles, Exceeds Expectations, Backlog Reaches $104 Billion

Há 15m

Is It Time to End the Era of Perp DEXes Burning Cash for Growth?

The article questions whether the era of perpetual decentralized exchanges (Perp DEXes) relying heavily on token incentives for growth should end. It argues that while token incentives (points, airdrops, farming) are useful for bootstrapping liquidity and attracting users, they often attract short-term, incentive-driven behavior rather than genuine, long-term traders. The piece analyzes how platforms use future token value to acquire current trading volume and liquidity, but notes this creates data that is difficult to interpret. A key test comes after the Token Generation Event (TGE), when the removal of incentive expectations reveals whether user retention is based on product quality or continued subsidies. Data from platforms like Hyperliquid, Lighter, and edgeX shows varying success in converting airdrop attention into sustained trading relationships post-TGE. Post-TGE, the focus often shifts from distributing tokens to supporting the token's value through mechanisms like fee revenue buybacks and burns. However, this ties value redistribution to token holdings rather than direct trading contributions. Analysis of user behavior on platforms like Hyperliquid reveals that holding a platform's token, participating in its financial products, and engaging in perpetual trading are often distinct activities, indicating that broad token ownership does not equate to genuine trading demand. This creates a tension for platforms, which must manage both a trading market and a token market with potentially conflicting goals. The article concludes by introducing PopDEX, an approach backed by Foresight Ventures, which proposes a different model. Instead of channeling value first through a platform token, PopDEX aims to establish a "100% value return" framework. This system is designed to directly and transparently redistribute the value generated from trading fees back to those who create it—such as active traders, referrers, and affiliates—based on verifiable market contributions. The piece suggests that as the Perp DEX industry matures, exploring alternative incentive structures that move beyond pure token-driven growth is a necessary evolution.

marsbitHá 22m

Is It Time to End the Era of Perp DEXes Burning Cash for Growth?

marsbitHá 22m

Chinese Data Generation Team Makes Debut in Nature Journal

A Hong Kong-based startup, Weina AI, has become China's first and the world's fourth data-generation technology company to publish in a leading Nature journal (IF>10 in recent three years) with a paper on AI-assisted kidney cancer surgery decision-making. Founded by Professor Liu Qifeng, who previously led the creation of the world's first thousand-card H800 SuperPod cluster at HKUST, the company focuses on enhancing AI's "questioning" ability to generate high-quality reasoning Q&A data, which is key for AI self-learning. The published Nature Communications paper, co-authored with medical experts, addressed a clinical challenge in renal surgery. Using an RDPM model on multi-source data from 1,621 patients, the AI achieved high predictive accuracy for long-term kidney function decline, providing a quantitative basis for surgical decisions. Professor Liu outlines AI development in three stages: from data to models, from models to token generation, and crucially, from tokens back to data—where AI actively generates questions, reasoning steps, and verified answers (cQrA: context, Question, reasoning, Answer). Weina AI's mission is to create this "data → model → token → data" feedback loop, enabling Agentic AI to autonomously evolve in professional domains. This moves beyond costly manual data annotation by using AI agents to generate scalable, chain-of-thought data. The company raised a HK$50 million seed round led by Lenovo Capital. Instead of focusing deeply on one sector, it adopted a "breadth-first" strategy, successfully deploying its technology across four high-accuracy, disparate fields: value-alignment safety, government affairs, insurance, and horse racing, proving cross-industry replicability. Liu argues that the future core of AI is not the model or data alone, but this self-reinforcing closed loop. This logic extends to embodied AI, where training shifts from imitation to autonomous generation of action data (cTrA: context, Task, reasoning, Action) through trial, error, and feedback. Weina AI's vision is to leverage this paradigm to "generate the world."

marsbitHá 22m

Chinese Data Generation Team Makes Debut in Nature Journal

marsbitHá 22m

A New Scaling Variable for Text-to-Image Generation, Discovered by ByteDance's Seed Team

ByteDance's SEED team investigated a crucial but often overlooked scaling variable in text-to-image diffusion models: the amount of image-grounded information in training captions. They found that simply increasing caption length with natural language does not improve model performance, as it often adds redundancy without new, usable visual supervision. The core discovery is that the final training loss of a diffusion model can be predicted by the *information content* of its text condition, measured by two complementary metrics: Grounded Perplexity Gain (GPG) and Effective Detailness (ED). This establishes a scaling relationship for text conditioning. To systematically increase information content, the team proposed **Structured Prompt (SP)**, a JSON-based representation that organizes visual variables (global scene, object attributes, spatial relationships) into clear fields, enhancing **Diffusability**—the model's ability to learn from captions. For inference, an LLM **Prompter** is trained to convert user queries into detailed SP instances, defining **Promptability**. The overall generation quality is viewed as a product of Diffusability and Promptability. A three-stage training strategy (SFT, cold-start reasoning distillation, and verifier-guided reinforcement) significantly improves the prompter's capability. The structured format also enables efficient iterative refinement through a *refine-render-judge* loop. In matched-control experiments using the same Qwen-Image backbone, data, and compute, the SP-based system substantially outperformed its natural-language counterpart, demonstrating that gains stem from the structured information interface, not just more training. The work shows that scaling text-to-image models requires scaling the *usable visual information* in conditions, not just model size or data volume.

marsbitHá 42m

A New Scaling Variable for Text-to-Image Generation, Discovered by ByteDance's Seed Team

marsbitHá 42m

How Does the Overtaken Central Province Plan to Tackle the 'K-Shaped Divergence'?

In the first half of the year, Anhui overtook Hunan in GDP rankings, entering the national top ten. Anhui's growth, driven by high-tech manufacturing and foreign trade, contrasts sharply with Hunan's slowdown, which is attributed to struggles in transitioning from traditional industries like construction machinery to new growth engines. While Hunan boasts strong educational resources, it faces a "strong research, weak industry" mismatch. Recent leadership inspections emphasized dual-track efforts: upgrading traditional sectors through AI and intelligent manufacturing, and fostering future industries like embodied AI and quantum technology. Experts suggest Hunan must avoid a "break then build" approach, instead consolidating its core industrial base while rapidly scaling up new technologies. The province is encouraged to enhance cluster collaboration within its flagship engineering machinery sector and improve the commercialization of research to close the developmental gap with neighbors like Anhui and Hubei.

marsbitHá 1h

How Does the Overtaken Central Province Plan to Tackle the 'K-Shaped Divergence'?

marsbitHá 1h

As the Crypto Market Cools, How is Neutrl Turning Tokenized Stocks into a New 'Yield Farm'?

As crypto markets cool, Neutrl explores turning tokenized stocks into a new yield source. The protocol focuses on delta-neutral strategies, shifting from crypto-native yields (like funding rates) which thin during downturns, to opportunities in pre-IPO and locked-up shares. The supply of these shares follows corporate calendars (IPOs, employee unlocks) independent of crypto cycles. Neutrl's strategy involves purchasing locked shares at a discount (e.g., 15%) and hedging by shorting equivalent perpetual futures. This creates a market-neutral position. During the lock-up period, the short position can earn significant funding rates from the nascent stock perpetuals market. A representative six-month trade could yield ~20.5%: ~15% from the entry discount and ~5.5% from funding rates, regardless of the stock's price movement. The opportunity is large, with the eligible locked stock pool estimated at 4.5x the size of filtered crypto unlock pipelines. As tokenization expands to assets like commodities and rates, each new market with liquid shorting tools can apply this same arbitrage framework, offering yield uncorrelated to crypto volatility.

marsbitHá 1h

As the Crypto Market Cools, How is Neutrl Turning Tokenized Stocks into a New 'Yield Farm'?

marsbitHá 1h

In Just 6 Months, 4 Rounds of Funding: West Lake University Professor's Venture Takes Off

Westlake Robotics, an embodied artificial intelligence company, has completed its Series A financing round within just six months and a total of four rounds, raising a cumulative 5 billion RMB. The investor lineup includes prominent institutions such as SAIF Partners, Xiaomiao Langcheng, Henan Investment Group Huirong Fund, and Haiyuan Fund, forming a high-quality capital matrix comprising state-owned, industrial, and leading venture capital. The rapid and intensive capital injection reflects strong market confidence in the company's technological approach, product deployment capabilities, and long-term potential. The newly acquired funds will be primarily allocated to the research and development of a unified large model for humanoid robots and the establishment of a talent cultivation base for embodied AI. Founded in 2024, Westlake Robotics originated from the industrial transformation of pioneering achievements in AI and robotics at Westlake University. The founding team is led by Wang Donglin, a leading figure in China's embodied AI and robot learning field, and co-founder Zhang Yue, an expert in natural language processing. The core R&D members hail from top-tier tech companies like Alibaba, ByteDance, Tencent, and Huawei, as well as prestigious global universities. The company follows a fully self-developed strategy integrating a "universal brain + humanoid body-specific cerebellum + proprietary humanoid hardware." It is one of the few domestic enterprises capable of holistically connecting the three core areas of embodied AGI cognitive reasoning, full-body motion control, and humanoid hardware. Its proprietary technologies include the General Motion Model-GAE system for low-latency teleoperation and motion generalization, and a dual pre-trained architecture for general and body-specific processing to bridge cognitive reasoning and physical movement. In 2026, Westlake Robotics launched its self-developed humanoid robot "Westlake o1," completing the full technology chain from underlying algorithms to pre-trained models and hardware. The company has secured nearly 100 million RMB in orders, with applications in scientific research, education, data collection, and power inspection. Future targets include high-risk industrial inspection, post-disaster search and rescue, and remote precision assembly. The company has also partnered with the Longyou County government to establish a county-wide real-scenario training base for humanoid robots, aimed at collecting high-quality motion data and validating technology in authentic environments. With the latest funding, Westlake Robotics plans to further advance its core model development and talent acquisition strategy, accelerating progress toward the "GPT moment" for embodied intelligence in China.

marsbitHá 1h

In Just 6 Months, 4 Rounds of Funding: West Lake University Professor's Venture Takes Off

marsbitHá 1h

Volatility Plummets to Historic Lows, When Will Bitcoin's 'Summer Sideways Move' End?

Bitcoin is experiencing a classic summer of stagnant price action, trapped in a tight range between approximately $62,000 and $66,000. Analysts point to historically low implied volatility and thin summer liquidity as key characteristics of the current market. The consensus among traders is that the catalyst for a decisive breakout will come from macroeconomic factors, not internal crypto dynamics. The immediate focus is on upcoming U.S. CPI data, which could influence Federal Reserve policy expectations. A softer inflation print is seen as potentially supportive for risk assets like Bitcoin. Furthermore, the pending *Clarity Act* legislation is identified as a crucial long-term catalyst that could boost institutional participation by providing regulatory clarity. While U.S. spot Bitcoin ETFs, led by BlackRock's IBIT, have seen their strongest inflows since April, providing underlying support, this buying pressure is being offset by selling from miners and other large holders. This has resulted in continued consolidation even as global crypto trading volumes hit multi-year lows. Market participants expect the range-bound, low-volatility environment to persist for several more weeks, at least until there is clearer progress on macro policy or regulatory fronts. Any sustained break above or below the current range is likely to trigger a significant expansion in volatility. Long-term bullish narratives around adoption, institutional demand, and Bitcoin's unique attributes as collateral remain intact, but the short-term path depends on external macroeconomic catalysts.

marsbitHá 1h

Volatility Plummets to Historic Lows, When Will Bitcoin's 'Summer Sideways Move' End?

marsbitHá 1h

活动图片

Categorias popularesright-arrow

Etiquetas Popularesright-arrow