Software
Search documents
Larry Ellison, Not Elon Musk, Was The Tech Titan Who Defined 2025
MINT· 2025-12-26 10:23
When the year began, a billionaire with close ties to the White House was a lock for the most newsworthy tech titan of 2025. But 12 chaotic months later, Larry Ellison, not Elon Musk, can justifiably lay claim to the title.The 81-year-old Oracle Corp. co-founder and chairman has been omnipresent — playing a role in just about every major business story of the year, from the frenzied artificial intelligence boom to the megadeals that are roiling Hollywood. Oracle even plans to take a stake in TikTok as part ...
Meta公布“超级智能”新进展:无需人类,软件Agent即可自我训练
3 6 Ke· 2025-12-26 10:20
近年来,基于大语言模型(LLMs)的软件工程智能体发展迅速,但其训练数据和训练环境仍高度依赖人类知识和人工策划,本质上是在复现人类开发轨 迹,难以自主发现新的问题结构与解决策略,这从根本上制约了智能体迈向超级智能的能力。 基于此,来自Meta、伊利诺伊大学厄巴纳-香槟分校的研究团队提出 Self-play SWE-RL(SSR),作为软件工程智能体训练范式的第一步。该方法对数据 假设的要求极低,仅需访问包含源代码和已安装依赖项的沙盒化代码仓库,无需任何人工标注的问题或测试用例。 研究表明,智能体可以从真实世界的软件仓库中自主获取学习经验,有望催生在系统理解、解决全新问题以及从零开始自主创建软件等方面超越人类能力 的超级智能系统。 智能体 Bug 注入通过让模型扮演"破坏者"构建起自驱动的进化闭环。 在这一过程中,首先生成包含 Bug 补丁和弱化测试的 Bug 构件,将抽象错误转化为标准化的练习题;随后,运用"删除关键代码"或"回滚历史修复"等复 杂生成策略,从真实工程逻辑中制造出极具挑战的高质量难题;为了确保逻辑严密,系统利用"逆向变异测试"进行严格的一致性验证,剔除无关干扰并确 保错误可复现;最后,通过动态 ...
Here's how much $1,000 invested in Palantir stock at the start of 2025 is worth now
Finbold· 2025-12-26 09:43
Core Viewpoint - Palantir Technologies has experienced significant stock price appreciation in 2025, driven by increasing demand for its AI platforms and broader commercial adoption [1][2]. Stock Performance - The stock price rose from approximately $75.19 on January 2, 2025, to around $194.17 by December 24, 2025, resulting in a year-to-date gain of about 158% [1][2]. - A $1,000 investment at the beginning of the year would now be worth approximately $2,580, reflecting a gain of about $1,580 [2]. Market Context - As the year-end approaches, Palantir's stock enters a period characterized by lighter liquidity due to holiday trading, which may lead to increased sensitivity to market movements and headlines [4]. Technical Analysis - Palantir's stock has recently formed a cup-with-handle breakout pattern, with a notable buy point near $190.40 and a record high of $207.52 achieved in November 2025 [5]. - Despite a bullish long-term trend, the stock is close to key breakout levels, indicating potential vulnerability to pullbacks if market sentiment changes [5]. Fundamental Outlook - Wall Street analysts, including Wedbush's Dan Ives, express optimism about Palantir's future, predicting continued success in commercial AI through its Artificial Intelligence Platform (AIP) [6]. - Ives suggests that Palantir could be on a trajectory toward a $1 trillion valuation within the next two to three years if it maintains its execution [6]. Investor Sentiment - As 2026 approaches, Palantir's strong price momentum, increasing enterprise adoption, and significant exposure to AI keep it on the radar of investors, while also highlighting the inherent volatility in the stock [7].
2500元/月雇个总监级AI数字员工,贵吗?
克而瑞地产研究· 2025-12-26 09:41
Core Viewpoint - A profound transformation in corporate structure is occurring in Silicon Valley, where AI agents are evolving from mere tools to autonomous colleagues, significantly impacting the real estate industry [1][3]. Group 1: AI Transformation in Real Estate - The real estate industry, characterized by high capital intensity and long decision chains, is becoming a breakthrough point for AI applications, with digital employees capable of performing tasks traditionally requiring multiple human roles [3][11]. - Deep Intelligence's "Kerry Digital Employee" has been recognized for its real industry value and scalable application capabilities, winning the "2025 Outstanding AI Product Award" [4]. - The introduction of digital employee teams, such as the "Gold Medal Case Field" team, showcases a collaborative approach to cover the entire process from market analysis to customer service in new housing projects [7][8]. Group 2: Cost Efficiency and Organizational Change - Traditional real estate marketing teams typically require 6-8 personnel with a monthly cost exceeding 150,000 yuan, while digital employees can cover the same functions for around 2,500 yuan, reducing labor costs by over 90% [11]. - The shift towards AI-native organizations emphasizes a model where human experts focus on high-value tasks while digital employees handle standardized, time-consuming tasks, creating a synergistic collaboration [11]. Group 3: Unique Industry Advantages - Deep Intelligence's AI solutions are tailored to the real estate sector, integrating industry knowledge, business processes, and proprietary data to create a specialized AI space that overcomes traditional barriers to high-end capabilities [13][16]. - The company has established four unique advantages: a vast structured database, a knowledge graph from unstructured documents, expert thinking encoding, and a stable multi-agent architecture for collaborative tasks [16]. Group 4: Broader Implications and Future Outlook - The trend of integrating digital employees is not limited to real estate; leading companies across various sectors are adopting similar strategies, with the AI digital human market in China projected to reach 4.12 billion yuan in 2024, growing by 85.3% [19]. - The future competitiveness of enterprises will depend on their ability to leverage top-tier professional capabilities through AI in a cost-effective and sustainable manner [20]. - The introduction of digital employees in real estate represents a significant opportunity for companies to break through the barriers of high-end capability scarcity, positioning them for success in the evolving market landscape [21].
微软AI CEO苏莱曼:未来5-10年,一家公司需砸千亿美元备战AI竞赛
Sou Hu Cai Jing· 2025-12-26 09:39
苏莱曼在 12 月 17 日更新的"Moonshots with Peter Diamandis"播客中提到,未来 5 年至 10 年内,要在 AI 最高层级展开竞争,所需投入将高达数千亿美元, 涵盖基础设施建设、硬件投入以及专业人才成本。这种规模的资本需求,使大型成熟企业在结构上具备明显优势。 苏莱曼还将微软当前的角色形容为一家"现代建筑公司",数十万名员工正为 AI 建设以千兆瓦计的 CPU 和加速器算力基础。 ▲ 图源微软官网 苏莱曼强调,成本压力不仅来自算力和硬件,还来自对人力资本的激烈争夺,尤其是 AI 研究人员、数据科学家和工程师,单个研究人员或技术人员的薪酬 水平正在急剧攀升。 IT之家 12 月 26 日消息,微软 AI CEO 穆斯塔法・苏莱曼警告称,企业若想在 AI 竞赛中占据领先位置,需要承受极其高昂的投入成本。 除人才争夺外,微软与其他大型科技公司一样,正加速推进 AGI 和超级智能的发展。行业领袖公开承认,这些目标不仅成本惊人,同时也具有极其重要的 战略意义。 IT之家注:AGI(通用人工智能)指能够在大多数任务上达到人类智能水平的系统,而超级智能则是指在整体能力上超越人类的 AI。 ...
用友网络:北京用友科技及其一致行动人累计质押数量约为3.98亿股
Mei Ri Jing Ji Xin Wen· 2025-12-26 09:33
每经AI快讯,用友网络(SH 600588,收盘价:12.88元)12月26日晚间发布公告称,截至2025年12月26 日,北京用友科技及其一致行动人上海用友科技咨询有限公司、用友研究所持有公司股份约14.21亿 股,占公司总股本的41.59%。本次补充质押及展期后,北京用友科技及其一致行动人累计质押数量约 为3.98亿股,占持股数量的27.98%,占公司总股本比例的11.64%。 每经头条(nbdtoutiao)——对话马斯克脑机接口"一号受试者":大脑植入芯片23个月,我正重新夺回 人生的独立 (记者 曾健辉) 2024年1至12月份,用友网络的营业收入构成为:软件行业占比98.19%,其他业务占比1.81%。 截至发稿,用友网络市值为440亿元。 ...
2 Overlooked AI Stocks to Buy Before They Soar Up to 100% in 2026, According to Wall Street Analysts
The Motley Fool· 2025-12-26 09:30
Core Viewpoint - Wall Street analysts believe Upstart Holdings and Atlassian are undervalued, predicting significant gains for shareholders in the coming years, despite their recent stock declines [1]. Group 1: Upstart Holdings - Upstart is an AI lending platform that enhances credit risk assessment by analyzing over 2,500 variables, unlike traditional systems that rely on FICO Scores [4]. - The highest target price for Upstart is set at $80 per share, indicating a potential upside of 70% from its current price of $47 [5]. - Upstart's third-quarter financial results showed a 71% increase in total revenue to $277 million, with non-GAAP net income rising to $0.52 per diluted share from a loss of $0.06 per share last year [7]. - Wall Street anticipates Upstart's adjusted earnings to grow at an annual rate of 51% through 2026, making its current valuation of 33 times earnings appear reasonable [8]. Group 2: Atlassian - Atlassian is known for its work management software, particularly Jira, which is widely used across both technical and non-technical departments [9]. - The highest target price for Atlassian is set at $320 per share, suggesting a potential upside of 100% from its current price of $160 [5]. - Atlassian's first-quarter fiscal 2025 results reported a 21% revenue increase to $1.4 billion, with non-GAAP earnings rising 35% to $1.04 per diluted share [13]. - Wall Street expects Atlassian's adjusted earnings to grow at 22% annually through June 2027, with a current valuation of 41 times earnings considered reasonable [14].
2026趋势报告:数据与人工智能
DataArt· 2025-12-26 09:18
Investment Rating - The report emphasizes that the highest return on investment in 2026 will come from modern data infrastructure rather than the latest AI models [11][14]. Core Insights - The gap between AI ambitions and actual operations is widening across industries, with organizations needing to focus on foundational work to achieve transformative wins [6][5]. - Companies are shifting from broad experimentation to specific, high-value use cases, moving AI from proof-of-concept to enterprise-level deployment [19][15]. - Successful organizations are prioritizing data lifecycle management, modernization, and human capabilities to shape their AI-driven transformation strategies [48][59]. Summary by Sections Overview - The report highlights a significant disconnect between organizational expectations of AI and the foundational work required for successful implementation [6][5]. - Many companies are still relying on outdated systems and manual processes, which hinders their ability to leverage AI effectively [6][9]. 2026 AI Trends - AI success in 2026 will be driven by data infrastructure rather than new models, with a focus on creating accessible and real-time data management systems [11][12]. - Organizations are expected to transition from broad AI experiments to targeted applications that deliver measurable business value [15][17]. Industry-Specific Trends - The report outlines predictions for various sectors, including: - Airlines will require rapid experimentation due to competitive pressures [65]. - Retail will see AI operating behind the scenes, influencing pricing and supply chain decisions [66]. - Healthcare will experience regulatory advancements that promote AI-driven innovations [69]. Preparing for 2026 - Companies need to invest in data management and governance to support AI initiatives effectively [48][51]. - A cultural shift is necessary for organizations to embrace AI as a core component of their business strategy rather than a standalone project [30][60]. Conclusion - The foundation laid in data infrastructure and governance will determine the success of AI initiatives in 2026, with companies that prioritize these areas likely to thrive [87][89].
从安全内核到可用系统:星绽NixOS发行版发布,加速OS行业向Rust迁移
Huan Qiu Wang· 2025-12-26 09:01
Core Viewpoint - Asterinas has launched Asterinas NixOS, the first distribution featuring a pure Rust operating system kernel, marking a significant step towards native memory safety in the OS industry [1][2]. Group 1: Asterinas NixOS Overview - Asterinas NixOS integrates a pure Rust kernel with the mature software ecosystem of NixOS, transitioning from a kernel project to a usable system [1]. - The project has been developed over three years, contributing 130,000 lines of Rust code and supporting over 230 Linux system calls [2]. - The release signifies a milestone, enabling Asterinas to support a user space for installation and experience [2]. Group 2: Technical Advantages - Asterinas NixOS is built on the innovative "framekernel" architecture, providing high performance and security compared to traditional OS kernels [2]. - It inherits the "system state stability reproduction" feature from NixOS, allowing developers to automatically construct systems based on specified software and configuration needs, ensuring consistent results across different machines and times [3]. - The integration with NixOS's extensive software ecosystem, which includes over 120,000 packages, allows Asterinas NixOS to avoid starting from scratch, facilitating a quicker transition to a mature system [3]. Group 3: Industry Trends and Future Plans - The trend in the industry is shifting towards native memory safety, with major players like Microsoft planning to migrate core codebases to Rust by 2030 [1][4]. - Asterinas NixOS aims to establish a reproducible and comparable baseline in real application loads, positioning itself at the forefront of memory-safe kernel engineering [4]. - The latest version, Asterinas 0.17.0, meets core user needs for desktop and server environments, with plans for deployment in secure computing and trusted data spaces by 2026 [4].
品茗科技:股东协议转让公司股份完成过户
Mei Ri Jing Ji Xin Wen· 2025-12-26 08:46
截至发稿,品茗科技市值为114亿元。 每经头条(nbdtoutiao)——对话马斯克脑机接口"一号受试者":大脑植入芯片23个月,我正重新夺回 人生的独立 每经AI快讯,品茗科技(SH 688109,收盘价:145元)12月26日晚间发布公告称,2025年9月24日,公 司控股股东、实际控制人莫绪军、股东李军、陶李义、李继刚及新余灵顺灵创业投资合伙企业(有限合 伙)与通智清研(北京)科技产业发展合伙企业(有限合伙)共同签署了《股份转让协议》,转让方拟 将其所持有的公司约1255万股无限售流通股(占公司总股本的约15.92%),通过协议转让的方式转让 给受让方。本次股份协议转让事项已通过上海证券交易所合规性审核,并于2025年12月26日取得中国证 券登记结算有限责任公司上海分公司出具的《证券过户登记确认书》,过户日期为2025年12月25日,过 户数量约为1255万股(占公司股份总数的约15.92%),股份性质为无限售流通股,本次股份协议转让 过户登记手续已办理完毕。 (记者 曾健辉) 2024年1至12月份,品茗科技的营业收入构成为:软件业占比100.0%。 ...