7*24

Hot topic

Market

Activities

On June 24, DWS projected that the U.S. economy will grow by 3.2% this year, with a forecast of 2.3% for next year, noting that AI could be a significant contributor to the U.S. economy. Wu Shuangrong, Chief Investment Officer for DWS Asia Pacific, anticipates that the Federal Reserve will cut rates twice before May next year, with the target range for the federal funds rate expected to decrease to between 3% and 3.25%. She explained that the Fed is primarily focused on inflation expectations and the unemployment rate. While oil prices are causing economic growth to slow, the situation remains manageable. On the other hand, there are concerns in the market about whether AI will replace some labor, leading to a reduction in potential job vacancies. Regarding the development of AI stocks, she believes that there will not be a bubble burst in the next 12 months but emphasizes the need to continue monitoring various indicators, including valuation multiples, corporate earnings, and balance sheets.

[DWS: Fed Expected to Cut Rates Twice Before May Next Year]On June 24, DWS projected that the U.S. economy will grow by 3.2% this year, with a forecast of 2.3% for next year, noting that AI could be a significant contributor to the U.S. economy. Wu Shuangrong, Chief Investment Officer for DWS Asia Pacific, anticipates that the Federal Reserve will cut rates twice before May next year, with the target range for the federal funds rate expected to decrease to between 3% and 3.25%. She explained that the Fed is primarily focused on inflation expectations and the unemployment rate. While oil prices are causing economic growth to slow, the situation remains manageable. On the other hand, there are concerns in the market about whether AI will replace some labor, leading to a reduction in potential job vacancies. Regarding the development of AI stocks, she believes that there will not be a bubble burst in the next 12 months but emphasizes the need to continue monitoring various indicators, including valuation multiples, corporate earnings, and balance sheets.

DWS: Fed Expected to Cut Rates Twice Before May Next Year

According to monitoring by Dongcha Beating, Google TPU software engineer Patrick Toulme pointed out that there is a misunderstanding regarding the claim that GLM 5.2 can catch up to Opus through distillation. The training difficulty of large models in agent encoding tasks lies in the 'zero gradient dilemma,' meaning that if the model cannot generate the correct operational path early on, reinforcement learning cannot obtain gradient signals to initiate parameter updates. The role of distilling Claude or GPT-5.5 is merely to provide seed answers during the cold start phase to bypass the zero gradient dilemma. Once the model surpasses the cold start threshold, subsequent performance improvements will no longer rely on distillation but will depend entirely on the self-evolution through reinforcement learning's hill-climbing algorithm. Toulme emphasized that GLM 5.2 already possesses the capability to independently generate successful paths and can autonomously iterate to higher levels through reinforcement learning, completely freeing itself from dependence on American large models. Redis founder Salvatore Sanfilippo added another possible pathway: while introducing reasoning patterns (distillation) through high-capacity models is very useful for obtaining better RL signals, the practice of DeepSeek R0 has already proven that reinforcement learning can operate autonomously and achieve breakthroughs even in pure cold start situations without any distillation seeding. He also believes that if there is still a need to overcome the cold start threshold, large model development can initially use local open-source models like DeepSeek-v3.2 for fine-tuning, rather than relying solely on American APIs.

[Opinion: API Distillation is Just a Stepping Stone; GLM 5.2's Autonomous Iteration Can Completely Eliminate Dependence on American Models]According to monitoring by Dongcha Beating, Google TPU software engineer Patrick Toulme pointed out that there is a misunderstanding regarding the claim that GLM 5.2 can catch up to Opus through distillation. The training difficulty of large models in agent encoding tasks lies in the 'zero gradient dilemma,' meaning that if the model cannot generate the correct operational path early on, reinforcement learning cannot obtain gradient signals to initiate parameter updates. The role of distilling Claude or GPT-5.5 is merely to provide seed answers during the cold start phase to bypass the zero gradient dilemma. Once the model surpasses the cold start threshold, subsequent performance improvements will no longer rely on distillation but will depend entirely on the self-evolution through reinforcement learning's hill-climbing algorithm. Toulme emphasized that GLM 5.2 already possesses the capability to independently generate successful paths and can autonomously iterate to higher levels through reinforcement learning, completely freeing itself from dependence on American large models. Redis founder Salvatore Sanfilippo added another possible pathway: while introducing reasoning patterns (distillation) through high-capacity models is very useful for obtaining better RL signals, the practice of DeepSeek R0 has already proven that reinforcement learning can operate autonomously and achieve breakthroughs even in pure cold start situations without any distillation seeding. He also believes that if there is still a need to overcome the cold start threshold, large model development can initially use local open-source models like DeepSeek-v3.2 for fine-tuning, rather than relying solely on American APIs.

Opinion: API Distillation is Just a Stepping Stone; GLM 5.2's Autonomous Iteration Can Completely Eliminate Dependence on American Models

On June 24, U.S. optical communication stocks, which collectively declined yesterday, rebounded in pre-market trading today. Among them, Qorvo, Credo Technology, and GlobalFoundries rose over 3%. Applied Optoelectronics, POET Technologies, Astera Labs, AXT Inc, Nokia, Tower Semiconductor, and Coherent increased by more than 2%. Mavenir Technologies rose nearly 2%, while Corning increased by 1.4%.

[U.S. Optical Communication Stocks Rebound in Pre-Market, Mavenir Technologies Up Nearly 2%]On June 24, U.S. optical communication stocks, which collectively declined yesterday, rebounded in pre-market trading today. Among them, Qorvo, Credo Technology, and GlobalFoundries rose over 3%. Applied Optoelectronics, POET Technologies, Astera Labs, AXT Inc, Nokia, Tower Semiconductor, and Coherent increased by more than 2%. Mavenir Technologies rose nearly 2%, while Corning increased by 1.4%.

U.S. Optical Communication Stocks Rebound in Pre-Market, Mavenir Technologies Up Nearly 2%

On June 24, RBC Capital raised the target price for Skyworks Solutions from $72 to $80, maintaining a 'Market Perform' rating.

[RBC Capital Raises Skyworks Target Price to $80]On June 24, RBC Capital raised the target price for Skyworks Solutions from $72 to $80, maintaining a 'Market Perform' rating.

RBC Capital Raises Skyworks Target Price to $80

According to Beating monitoring, in response to the continuous loss of core scientists, Google DeepMind head Demis Hassabis directly countered at the Cannes Lions Festival, asserting that Google is still capable of recruiting and retaining the top scientists. Recently, the departures of Noam Shazeer, one of the founders of the Transformer model, and John Jumper, a key contributor to AlphaFold, triggered collective anxiety on Wall Street regarding Google's talent retention capabilities, even causing the company's stock to drop by as much as 7% in a single day. However, Hassabis appeared confident, stating that Google still boasts the largest and most robust research team in the industry. In Hassabis's view, the high turnover of AI talent is quite normal in the world's most competitive job market. Google's ace lies in its underlying network woven from data, integrated hardware, and computational power, particularly the customized TPU clusters, which remain an irreplaceable asset for attracting top researchers. Additionally, the merger of Google Brain and DeepMind in 2023 has consolidated previously scattered R&D resources into a powerhouse. Semafor believes that the departure of top researchers is more of a public relations crisis for Google rather than an operational crisis. The real battle for talent in the AI field is occurring in academia, as the next generation of algorithm breakthroughs is more likely to come from young scholars just entering the workforce rather than executives resting on their laurels.

[Nobel Laureate Departures and Stock Plunge: Google AI Leader's Strong Response on TPU Power]According to Beating monitoring, in response to the continuous loss of core scientists, Google DeepMind head Demis Hassabis directly countered at the Cannes Lions Festival, asserting that Google is still capable of recruiting and retaining the top scientists. Recently, the departures of Noam Shazeer, one of the founders of the Transformer model, and John Jumper, a key contributor to AlphaFold, triggered collective anxiety on Wall Street regarding Google's talent retention capabilities, even causing the company's stock to drop by as much as 7% in a single day. However, Hassabis appeared confident, stating that Google still boasts the largest and most robust research team in the industry. In Hassabis's view, the high turnover of AI talent is quite normal in the world's most competitive job market. Google's ace lies in its underlying network woven from data, integrated hardware, and computational power, particularly the customized TPU clusters, which remain an irreplaceable asset for attracting top researchers. Additionally, the merger of Google Brain and DeepMind in 2023 has consolidated previously scattered R&D resources into a powerhouse. Semafor believes that the departure of top researchers is more of a public relations crisis for Google rather than an operational crisis. The real battle for talent in the AI field is occurring in academia, as the next generation of algorithm breakthroughs is more likely to come from young scholars just entering the workforce rather than executives resting on their laurels.

Nobel Laureate Departures and Stock Plunge: Google AI Leader's Strong Response on TPU Power

On June 24, according to monitoring by Hyperinsight, as Micron's earnings report approaches, trading activity for MU contracts on the Hyperliquid platform has significantly increased, with a 24-hour trading volume of approximately $459 million and a total open interest value of $195 million. On-chain detection by BlockBeats indicates that this morning the contract price reached about $1,030 before rebounding to $1,091. Last night and this morning, two major on-chain whales opened MU short positions totaling $8.3 million, with an average entry price of about $1,071, currently facing a slight unrealized loss, and the recent liquidation line is at $1,470. It is reported that the two addresses currently hold a total of $55.5 million in various U.S. stock short positions, one of which belongs to the well-known trader CBB.

[Two Major Whales Open $8.3 Million MU Short Position, Anticipating Continued Stock Price Decline Post-Earnings Report]On June 24, according to monitoring by Hyperinsight, as Micron's earnings report approaches, trading activity for MU contracts on the Hyperliquid platform has significantly increased, with a 24-hour trading volume of approximately $459 million and a total open interest value of $195 million. On-chain detection by BlockBeats indicates that this morning the contract price reached about $1,030 before rebounding to $1,091. Last night and this morning, two major on-chain whales opened MU short positions totaling $8.3 million, with an average entry price of about $1,071, currently facing a slight unrealized loss, and the recent liquidation line is at $1,470. It is reported that the two addresses currently hold a total of $55.5 million in various U.S. stock short positions, one of which belongs to the well-known trader CBB.

Two Major Whales Open $8.3 Million MU Short Position, Anticipating Continued Stock Price Decline Post-Earnings Report

On June 24, reports from Israel indicated that as new rounds of talks between Israel and Lebanon were held in the United States, Israel was contemplating a military withdrawal from certain areas in southern Lebanon. The fifth round of talks began on June 23 and is expected to continue until June 25. This discussion contrasts with Israeli Prime Minister Netanyahu's earlier statements about the military remaining in the 'security zone' in southern Lebanon. The current talks are focused on the potential withdrawal of Israeli forces from parts of Lebanon. Reports suggest that the areas expected for withdrawal are located south of the so-called 'Blue Line,' where Israeli forces are currently stationed. A senior Israeli official told Channel 13 that the military 'recently occupied some areas for negotiation purposes and will withdraw from there afterward.' The official mentioned that the military would withdraw from 'a small area' in southern Lebanon, which would only account for 'a few percentage points' of the relevant region. This withdrawal plan is part of a proposal pushed by the United States, which includes training the Lebanese military to enhance its deployment and control capabilities, ensuring that personnel entering the areas vacated by Israeli forces are from the Lebanese military and not from Hezbollah.

[Israeli Military Considering Withdrawal from Southern Lebanon Regions]On June 24, reports from Israel indicated that as new rounds of talks between Israel and Lebanon were held in the United States, Israel was contemplating a military withdrawal from certain areas in southern Lebanon. The fifth round of talks began on June 23 and is expected to continue until June 25. This discussion contrasts with Israeli Prime Minister Netanyahu's earlier statements about the military remaining in the 'security zone' in southern Lebanon. The current talks are focused on the potential withdrawal of Israeli forces from parts of Lebanon. Reports suggest that the areas expected for withdrawal are located south of the so-called 'Blue Line,' where Israeli forces are currently stationed. A senior Israeli official told Channel 13 that the military 'recently occupied some areas for negotiation purposes and will withdraw from there afterward.' The official mentioned that the military would withdraw from 'a small area' in southern Lebanon, which would only account for 'a few percentage points' of the relevant region. This withdrawal plan is part of a proposal pushed by the United States, which includes training the Lebanese military to enhance its deployment and control capabilities, ensuring that personnel entering the areas vacated by Israeli forces are from the Lebanese military and not from Hezbollah.

Israeli Military Considering Withdrawal from Southern Lebanon Regions

From June 18 to July 14, HashKey Chain will host the "HashKey Chain Horizon" series hackathon in Japan. This hackathon is based on building a compliant and secure Web3 ecosystem, targeting global developers, innovators, and Web3 enthusiasts. There are two competition tracks focusing on key areas within the HashKey Chain ecosystem. Main Tracks: • AI (HSP) — Explore AI Agents in on-chain finance, build AI-native financial infrastructure, and enterprise-level autonomous agent applications, promoting the technological integration and large-scale adoption of AI × Crypto. • DeFi (RWA / Lending Protocol) — Focus on the tokenization of real-world assets (RWA), new on-chain lending protocols, and compliant stablecoin applications to create the next generation of decentralized compliant financial infrastructure. Prize Pool: The total prize pool for this hackathon is 12,000 USDT. Winning projects will not only have the opportunity to receive cash prizes and mainnet deployment guidance but can also join the HashKey Chain ecosystem and engage in in-depth collaboration with the official team and partners. Note: Projects must complete the mainnet deployment on HashKey Chain to be eligible for awards. HashKey Chain Ecosystem Support Program: Winning projects will establish dedicated TG working groups, and developer teams will receive targeted resource support. Additionally, all participating project information will be provided to the organizers and sponsors for future ecosystem integration. More activities in the "HashKey Chain Horizon" series will continue to be promoted. In addition to the online hackathon, a series of online communication activities such as X Space AMAs will also take place. Furthermore, on the evening of July 14 (during the WebX conference), an offline Demo Day roadshow and networking event will be held at CIC Tokyo, and all are welcome to participate! Hackathon registration and details link. About HashKey Chain: HashKey Chain is a compliant, institutional-grade blockchain under HashKey (stock code: 3887.HK), dedicated to reshaping the global financial market on-chain. Characterized by compliance, security, and innovation, HashKey Chain provides infrastructure for the next generation of on-chain finance, enabling stablecoins, RWAs (real-world assets), and institutional-grade DeFi to operate efficiently and seamlessly within the same system.

[HashKey Chain Japan Hackathon Countdown: $12,000 USDT Prize Pool Invites Global Developers]From June 18 to July 14, HashKey Chain will host the "HashKey Chain Horizon" series hackathon in Japan. This hackathon is based on building a compliant and secure Web3 ecosystem, targeting global developers, innovators, and Web3 enthusiasts. There are two competition tracks focusing on key areas within the HashKey Chain ecosystem. Main Tracks: • AI (HSP) — Explore AI Agents in on-chain finance, build AI-native financial infrastructure, and enterprise-level autonomous agent applications, promoting the technological integration and large-scale adoption of AI × Crypto. • DeFi (RWA / Lending Protocol) — Focus on the tokenization of real-world assets (RWA), new on-chain lending protocols, and compliant stablecoin applications to create the next generation of decentralized compliant financial infrastructure. Prize Pool: The total prize pool for this hackathon is 12,000 USDT. Winning projects will not only have the opportunity to receive cash prizes and mainnet deployment guidance but can also join the HashKey Chain ecosystem and engage in in-depth collaboration with the official team and partners. Note: Projects must complete the mainnet deployment on HashKey Chain to be eligible for awards. HashKey Chain Ecosystem Support Program: Winning projects will establish dedicated TG working groups, and developer teams will receive targeted resource support. Additionally, all participating project information will be provided to the organizers and sponsors for future ecosystem integration. More activities in the "HashKey Chain Horizon" series will continue to be promoted. In addition to the online hackathon, a series of online communication activities such as X Space AMAs will also take place. Furthermore, on the evening of July 14 (during the WebX conference), an offline Demo Day roadshow and networking event will be held at CIC Tokyo, and all are welcome to participate! Hackathon registration and details link. About HashKey Chain: HashKey Chain is a compliant, institutional-grade blockchain under HashKey (stock code: 3887.HK), dedicated to reshaping the global financial market on-chain. Characterized by compliance, security, and innovation, HashKey Chain provides infrastructure for the next generation of on-chain finance, enabling stablecoins, RWAs (real-world assets), and institutional-grade DeFi to operate efficiently and seamlessly within the same system.

HashKey Chain Japan Hackathon Countdown: $12,000 USDT Prize Pool Invites Global Developers

On June 24, the yield on UK 10-year gilts fell to 4.722%, marking the lowest level in two and a half months.

[UK 10-Year Gilt Yield Falls to 4.722%, Lowest in Two and a Half Months]On June 24, the yield on UK 10-year gilts fell to 4.722%, marking the lowest level in two and a half months.

UK 10-Year Gilt Yield Falls to 4.722%, Lowest in Two and a Half Months

On June 24, the European Securities and Markets Authority (ESMA) called for unlicensed crypto asset service providers (CASPs) to orderly exit their operations as the MiCA transition period comes to an end.

[EU MiCA Transition Period Ending, ESMA Urges Unlicensed Crypto Service Providers to Cease Operations]On June 24, the European Securities and Markets Authority (ESMA) called for unlicensed crypto asset service providers (CASPs) to orderly exit their operations as the MiCA transition period comes to an end.

EU MiCA Transition Period Ending, ESMA Urges Unlicensed Crypto Service Providers to Cease Operations

Bernstein Report: Agentic AI Will Transform CPU from Supporting Role to Leading Role, Bullish on Hygon Information

Cursor's New 1.5T Model Abandons Kimi Foundation, $60 Billion Stock Swap for Cursor, Musk's First Move Targets Microsoft's GitHub

3B Small Model's Programming Scores Rival Opus 4.5, Mysterious Model Sparks Heated Discussion, Turns Out to be Domestic

Breaking: Google Gemini Co-Head Poached by OpenAI

SK Hynix Stock Price Hits New High: Delivers HBM4E Samples, Reinforcing Its Leading Position in AI Memory

Ethereum Q1 Report: On-chain Activity Hits Record High, Tokenized Assets Lead the Industry

Intel CEO Liwu Chen's First Podcast Interview: Our Goal is '10x in 5-10 Years', Betting on Advanced Packaging, Glass Substrate, and Synthetic Diamond

EU MiCA Transition Period Ending, ESMA Urges Unlicensed Crypto Service Providers to Cease Operations

Isenção de responsabilidade：O conteúdo acima não representa as posições da HTX.，A HTX não faz quaisquer recomendações de negociação.。

Todos os comentários0Mais recentePopular

Tópicos em destaque