图片来源： Rillet 对于会计部门而言，总账系统是最为关键的软件。作为汇总所有财务交易的核心枢纽，它提供了生成 准确财务报表所需的基础数据。本周三，Rillet 宣布完成 2500 万美元 A 轮融资，由红杉资本领投， 现有投资者跟投。 此次融资距该公司从 First Round Capital、Creandum 和 Susa Ventures 获得 1350 万美元种子轮及 Pre- seed 轮融资仅过去 10 个月。 "总账是财务职能的核心，要求企业更换总账系统无异于进行心脏直视手术。"红杉资本合伙人 Julien Bek 表示。 就在几年前， Bek 还认为风投机构不敢投资开发新型总账软件的初创公司。他解释道，这不仅要克服 客户更换现有会计软件的阻力，建立新的总账业务本身也极具挑战性。 当 Bek 发现成立三年的 Rillet 公司时，他改变了看法。 该公司运用机器学习和生成式 AI 实现会计报 告自动化，直接从客户银行及 Salesforce 、 Stripe 、 Ramp 、 Brex 和 Rippling 等平台提取数据， 生成包括资产负债表和利润表在内的核心财务报表。 " 我认为他们三分之 ...

5.19-5.25 TOP10 | | Best of the week of May 19, 2025 | Daily | Weekly | Monthly | Yearly | Featured | All | | --- | --- | --- | --- | --- | --- | --- | --- | | 1 | 4月 28-5月 4 5月 5—11 | 5月 12—18 | | 5月 19-25 | | 5月 26—6月 1 | 1 | | | Tofu Pages | | | | | D | റ | | | Infinite landing pages from your existing website | | | | | 166 | 1,131 | | | Chrome Extensions · Marketing · SaaS | | | | | | | | | Algebras Al | | | | | D | റ | | | Translate apps with Al, no proofreading | | | | | 236 | 1,085 | | | Open Source ...

图片来源： Salesforce Horizon3.ai 是一家提供自主渗透测试等工具的网络安全初创公司，该公司本周在提交给美国证券交易 委员会的文件中透露，正在寻求新一轮 1 亿美元的融资，并已锁定至少 7300 万美元。 据知情人士透露，此轮融资由 NEA 领投，该初创公司的估值被认为超过 7.5 亿美元。 目前 无法核实这一估值是融资前还是融资后的估值。另一位人士认为该公司已经完成全部 1 亿美元 的融资，并补充说该公司的年度经常性收入约为 3000 万美元。 通过此次交易， Horizon3.ai 成为 NEA 在不到一个月内投资的第二家大型网络安全初创企业，此前 NEA 在 4 月参与了 Veza 以 8 亿美元估值完成的 1.08 亿美元融资轮。 2023 年 8 月， Horizon3.ai 在由 Craft Ventures 领投、 SignalFire 跟投的 C 轮融资中筹集 4000 万美 元 。联合创始人兼CEO Snehal Antani 当时向 TechCrunch 表示，该轮融资使公司总募资额达到 7850 万美元，资金将用于研发扩展、渠道布局及工程团队建设。 Horizon3. ...

创始人 Joseph Semrai 是泰尔奖学金获得者，他在 2024 年意识到现有数字办公套件无法充分利用 AI 模型潜力后，开始着手创建 Context 。 我们现有的各种应用程序在设计时并未充分考虑 AI 模型 的强大能力， "Semrai 在电话采访中告诉 TechCrunch 。 " 我们希望利用 模型， 能够理解大上下文窗口并同时使用多个应用程序的优势，以获 得最佳效果。 图片来源： SpAItial Context 是一家开发人工智能办公套件的初创公司，于5 月 28 日宣布完成 1100 万美元种子轮融资， 由 Lux Capital 领投，高通创投和 General Catalyst 跟投。此轮融资使公司总融资额达到约 1575 万美 元，估值达 7000 万美元。 目标并非完全取代 Microsoft 365 这类功能齐全的办公套件。 Semrai 表示， Context 瞄准的是当前工 具未能很好服务的市场。例如与许多 AI 数据分析产品不同 ， Context 很快将支持离线工作，通过其 桌面客户端基于现有数据和文档实现简易分析与文件草拟。 用户可免费试用 Context ，获得 50 ...

图片来源： UniX AI 图片来源： UniX AI 算法层面 ，优理奇提出了 可泛化 模仿学习框架「 UniFlex」， 通过解耦感知和操作模型， 在 操作 数据极度 匮乏的当下 可通过 极 少量示教完成复杂任务 的高效模仿学习 ， 同时利用预训练多模态感知模型实现物体场景操作的抗干扰泛化，并已在泛化场景操作中初步展现了涌现能力 。 同时，优理奇构建了 全世界第一个融合触觉的多模态感知模型 「 UniTouch」 ，可通过视觉生成触觉的方式实现夹豆腐，捏鸡蛋等精细任务 。在此基础上，优理奇还自主构建 了支持长程任务规划的推理架构「 UniCortex 」，可 在复杂场景和多智能体间，实时规划以及调用已习得技能 。以上核心 算法 能力的协同进展，使得优 理奇在复杂开放环境中 已初步实现了任务的泛化场景应用落地 。 近期，具身智能公司「优理奇机器人 U niX AI 」完成 数亿元天使轮及天使 +轮 融资， 中关村前沿基金，赛纳资本及长安私人资本参与本轮融资。 本轮融 资将用于加速 研发 多模态具身智能大模型与 通用 机器人本体的同步演进，以及面向多个 泛商业服务和 C端 场景落地与交付。 硬件方面，优理奇 ...

Core Insights - Anthropic has launched a voice mode for its AI model Claude, allowing users to interact using voice and choose from five unique tones [1] - This feature enhances user experience by enabling natural and intuitive conversations, similar to offerings from other AI companies like OpenAI and Google [2] Group 1 - The voice mode allows users to discuss documents and images, with the ability to switch between text and voice at any time [1] - Voice interactions are subject to usage limits, with most free users allowed 20-30 conversations [2] - Only paid subscribers can access the Google Workspace connector, which integrates with Google Calendar and Gmail, while Google Docs integration is exclusive to enterprise users [2] Group 2 - Anthropic's Chief Product Officer, Mike Krieger, confirmed the development of the voice feature in early March during an interview with the Financial Times [2] - The company is reportedly in discussions with Amazon, its main investor and partner, and AI startup ElevenLabs to enhance future voice capabilities for Claude [2]

Core Viewpoint - Salesforce's acquisition of Informatica for $8 billion is a strategic move to enhance its AI and data infrastructure capabilities, marking a significant shift in the company's approach to autonomous AI deployment [1][2]. Group 1: Acquisition Details - Salesforce will acquire Informatica at a price of $25 per share in cash for its Class A and B-1 common stock, adjusted based on previous investments in the company [1]. - Informatica, founded in 1993, serves over 5,000 customers across more than 100 countries and currently has a market capitalization of $7.1 billion [1]. Group 2: Strategic Implications - The acquisition aims to provide a more robust data infrastructure and governance system, enabling Salesforce's AI Agent to operate more securely, responsibly, and at scale within modern enterprises [1]. - Salesforce CEO Marc Benioff emphasized that this acquisition is a critical step towards achieving enterprise-level AI that is secure, reliable, and deeply integrated with global data [2]. Group 3: Market Context - The acquisition discussions began in April 2024, with initial market reactions leading to stock price declines for both companies due to concerns over integration challenges and strategic alignment [2]. - Informatica is not the first data management company acquired by Salesforce in the past year; the company previously acquired Own Company for $1.9 billion in September 2023 [2].

Core Viewpoint - The article discusses the emergence of AI models capable of generating realistic 3D environments from text prompts, highlighting the startup SpAItial founded by Matthias Niessner, a prominent AI researcher in Europe [1][2]. Funding and Team - SpAItial has raised $13 million in seed funding, led by Earlybird Venture Capital, with participation from Speedinvest and several angel investors [2]. - The founding team includes experienced professionals from notable companies, such as Ricardo Martin-Brualla from Google and David Nowotny from Meta, enhancing the startup's competitive edge [2][3]. Market Potential and Competition - The market for photorealistic 3D environments is still undefined, with applications ranging from gaming to architecture and potentially robotics [4]. - Current competitors include Odyssey, which has raised $27 million, and World Labs, valued at over $1 billion, but Niessner believes competition remains moderate [3][4]. Development Strategy - SpAItial aims to allow developers to use its foundational models for specific applications while prioritizing revenue generation over rapid team expansion [5]. - The company focuses on creating larger, more interactive 3D spaces, with the goal of enabling users to create their own video games easily [6]. Future Aspirations - Niessner envisions SpAItial potentially replacing CAD software as the next chapter in 3D generation technology unfolds [7].

Core Viewpoint - Cathie Wood emphasizes the potential for accelerated GDP growth in the U.S. and the positive impact of tax reductions on corporate competitiveness and innovation [3][4][5] Economic Outlook - U.S. GDP is expected to exceed growth expectations, with a focus on tax reductions and the removal of trade barriers as positive signals for the economy [3] - The reduction of corporate tax rates from 35% to 21% during Trump's first term led to a significant increase in corporate tax revenue, demonstrating that lower tax rates can enhance competitiveness and drive economic growth [4][5] Innovation and Technology - The cost of AI innovation is rapidly decreasing, with training costs dropping by 75% annually and inference costs by 85%, leading to a surge in innovative companies globally [6] - The U.S. technology sector is expected to thrive under the current regulatory environment, with a notable increase in the market capitalization of tech companies from 2019 to 2024 [6][8] Healthcare Sector - AI is seen as a major beneficiary in the healthcare sector, with advancements in drug discovery and diagnostics expected to significantly reduce costs and improve outcomes [16] - Companies like CRISPR Therapeutics are at the forefront of utilizing AI for groundbreaking treatments, showcasing the potential for AI to revolutionize healthcare [16][17] Investment Strategy - Cathie Wood suggests that investors should look beyond the MAG-6 tech giants and focus on emerging companies in innovative sectors that are currently undervalued [8] - The healthcare industry is highlighted as a particularly promising area for investment, with AI expected to enhance research and development returns [16] Government and Defense - The shift towards modernization in government and defense sectors is crucial, with companies like Palantir leading the way in improving efficiency and adapting to new technological demands [10][11] - The changing nature of warfare, particularly the rise of drone technology, indicates a need for investment in modernized defense systems [11]

在 2023 年那篇向世界介绍 Llama 的里程碑论文中， 14 位署名作者仅有 3 人仍留在 Meta ： 研究科 学家 Hugo Touvron 、研究工程师 Xavier Martinet 和技术项目负责人 Faisal Azhar 。其余人员均已离 职，其中多数加入了新兴竞争对手或自立门户。 这 11 位离职作者在 Meta 的平均任职时间超过五年，表明他们并非短期聘用人员，而是深度参与 Meta 人工智能研究的核心成员。 其中最早有人于 2023 年 1 月离职，部分人员坚持完成了 Llama 3 项目周期，还有几位直到今年才离开。 他们的集体离职，标志着这个曾帮助 Meta 凭借开源模型奠定 AI 声誉的团队正在悄然解体。 Meta 的人才流失在 Mistral 体现得最为明显。这家巴黎初创公司由前 Meta 研究员、 Llama 核心架构 师 Guillaume Lample 和 Timothée Lacroix 联合创立。他们与多位 Meta 前同事正开发强大的开源模 型，直接与 Meta 的旗舰 AI 项目形成竞争。 随着时间的推移，人才流失引发了外界对 Meta 能否留住顶尖 AI 人才 ...