在 2026 年国际消费电子展（ CES ）上， Neurable 宣布与惠普旗下游戏品牌 HyperX 达成战略合作，计划共同开发 业界首款集成神经技术的游戏耳机 。 这 款新型可穿戴设备将结合人工智能与 非侵入式传感器 ，在游戏过程中实时解读脑电信号（ EEG ），并把玩家的神经活动转化为可量化、可用于训练与优化 的表现指标。 与传统脑电监测设备常见的笨重头盔式形态不同，该产品采用 Neurable 的 微型脑机接口（ BCI ）技术 ，可以在不牺牲佩戴体验的前提下嵌入 HyperX 的耳 机结构中，做到 更纤薄、更轻量、外观更 " 隐形 " ，从而更贴近玩家日常使用场景。 Neurable 同时展示了 " 实时脑部活动洞察 " 对提升玩家表现的潜在价值。在一项面向 半职业电竞选手 的初步研究中， Neurable 的神经反馈系统 Prime 在第 一人称射击（ FPS ）训练任务中带来了可观改进：参与者整体表现为 反应更快、命中率更高、击中目标数量更多 。平均来看，反应时间缩短 43 毫秒 ，命中 率提升 0.53% ，并在射击心理测试中 额外击中近 9 个目标 。 在 大学及职业电竞选手 群体中，提升 ...

【1月25日下午，五道口】 与5位AI创业CEO和专家深度交流，4位创业者路演展示，仅设40席位 ♀ 北京五道□ © 1月25日下午 当中国 AI2C 应用公司跻身全球主视野, 下一个百亿美金赛道的 风口正待你我解锁。我们邀请了五位 AI2C 领域的创始人和操盘手 聚焦软件,从模型到应用,从创作到生产力,从国内到Day One出海 他们将带来寻找业务破局点和上升循环的第一视角。 cT 创业者说 五位创业者 × 深度拆解 + 现场Q&A交锋 ◎如何找到AI2C的第一个真需求?◎AI2C的出海打法? 不要客气 直接CHALLENGE >> 2 Speedy Pitch 4位早期创业者 × 5分钟路演展示 n放招募中 Call for speakers ! ! !! 下方报名时请注明是否申请路演展示机会 ct3 圆桌深潜 「特邀主持人 + 5位嘉宾』 Panel现场开放观众互动提问 环 LUU タ | 凹条川北区，胶ガ胆 44 16.H 厂。 | 0] 依托技术积淀推出了 To B 品牌影伙引擎布局智慧视 频创作、AI 短剧及营销视频创作服务。 ACT4 大咖面对面 【2位创业者 + 10位参会者】× 4组平行圆桌 ...

Seb Johnson ： 大家好，欢迎回到《Scaling Europe》节目。我是 Seb Johnson。我和 CJ Simon-Gabriel 一起在这里。CJ 是 Mirelo AI 的联合创始人 之一。 Mirelo AI 刚刚宣布了一个非常夸张的 4100 万美元种子轮 ，由 A16z 和 Index Ventures领投。 这是一笔很大的融资，而且由一些真正的顶级 VC 领投。我觉得特别有意思的是，你们是在欧洲做一个" 基础模型 "。所以对那些不了解的人，你能不能先快 速介绍一下 Mirelo AI ？ CJ： 谢谢你邀请我。 我们主要聚焦在为 视频内容和游戏 做"音频"。所以我们现在做的主要是 音乐和 音效 。 我们的想法其实很简单， 你把你的视频给我，我们告诉你"哪里该用什么声音"，并且把音频生成出来 。你可以生成音效，也可以加上音乐。 Seb Johnson： 你为什么决定做这个业务？ 过去一年，AI 视频生成在模型能力与产品形态上快速迭代，视频产出的边际成本持续下降，生成速度与可控性显著提升。今天不少 AI 创作者都经历过：画面 几分钟出片，真正让人头大的，是后面的音效、配乐、节奏、氛 ...

1 2 月 16 日， Databricks 宣布完成 超过 40 亿美元 融资，投后估值 1340 亿美元 ；本轮由 Insight Partners 、 Fidelity Management & Research 、 J.P. Morgan Asset Management 领投， Andreessen Horowitz 、 BlackRock 、 Blackstone 等参与。公司表示，本轮距离上一轮融资（当时估值 约 1000 亿美元 ）不到半年。经营数据方面，公司披露其第三季度对应的 年化营收规模超过 48 亿美元 ，同比增长 超过 55% ；同时称其 AI 相关 产品与数据仓库业务的 年化收入规模均超过 10 亿美元 ，并在过去 12 个月实现 自由现金流为正 。 02 为什么是 Databricks? 十年前， 真正能拉开差距的往往是资本和规模：前者支撑持续投入，后者沉淀为渠道、产能、销售体系与品牌。 今天这套逻辑仍有效，但不够。越来越多行业的领先来自两类新资源： 数据和人才。数据决定产品迭代、客户理解与运营效率；人才决定数据能否被 转化为决策与业务结果。 现实问题是，很多企业仍用上个周期的数 ...

过去一年多里， 大多数权威评测仍然在反复强调同一件事：在最前沿的综合能力上，闭源模型的曲线更陡，开源想在所有维度上追平变得越来越难。 DeepSeek 在技术报告中也承认：开源社区在进步，但 Anthropic 、 Gemini 、 OpenAI 这些闭源模型的性能曲线更陡，差距其实在拉大。在复杂任务上，闭源 系统展现出越来越明显 的优势。 目前开源模型有三个关键问题 ： 1. 首先，在架构层面，当前主流仍高度依赖 Vanilla Attention 机制，这在 长序列场景 下会严重限制计算效率。这种低效对模型的 大规模部署 以及有效的后训 练都构成了实质性障碍。 2. 其次，在资源投入上，开源模型在 后训练 阶段普遍面临 算力投入不足 的问题，从而限制了其在高难度任务上的表现。 3. 最后，在 AI Agent 场景中，相比于闭源系统，开源模型在 泛化能力 与 指令跟随能力 方面存在显著滞后，这削弱了其在真实部署中的有效性。 12月1 号， DeepSeek 发布了两款新模型： DeepSeek V3.2 和 DeepSeek V3.2 Speciale ，针对这三个问题， 提出了三个改进 : 1. 引入了 ...

Core Insights - The article discusses the advancements in AI models, particularly focusing on Opus 4.5, which shows significant improvements in performance and efficiency compared to its predecessors and competitors [1][14][16] Group 1: Performance Comparison - Opus 4.5 outperforms Gemini 3 Pro in generating interactive applications, achieving a high level of completion and usability with minimal prompts [1][3] - In coding tests, Opus 4.5 demonstrates superior efficiency, using significantly fewer tokens while achieving comparable or better results than Sonnet 4.5 [6][7] - The model's ability to utilize tools has improved, allowing it to selectively call only relevant tools, which enhances efficiency and reduces token consumption [8][9] Group 2: Cost Efficiency - The pricing structure for token usage has been reduced to $5 per million input tokens and $25 per million output tokens, approximately one-third of previous costs, leading to a notable increase in cost-effectiveness [7][8] - Opus 4.5's advanced tool usage allows it to complete tasks at a much lower cost compared to Sonnet 4.5, with estimates showing a task cost of about $1 for Opus 4.5 versus $4 for Sonnet 4.5 [8][9] Group 3: Advanced Features - The introduction of the "effort" parameter allows users to customize the model's input intensity, balancing between time and cost efficiency [4][6] - The "infinite chat" feature enables continuous dialogue without hitting context limits, allowing for more seamless long-term projects and collaboration [11][12][13] - The enhanced computer use capability allows the AI to perform tasks directly on a computer interface, including zooming in on elements for precise interactions [9][10] Group 4: Market Positioning - Opus 4.5 is positioned as a tool for professional software developers and knowledge workers, emphasizing its utility in complex project management and collaborative development [16] - The model aims to redefine software production processes by acting as a collaborative developer rather than just a code completion tool [16]

Core Insights - Fal has transformed "real-time video generation" from a flashy demo into a reusable infrastructure, achieving an annual recurring revenue (ARR) growth from approximately $2 million to over $100 million in less than two years, serving over 2 million developers and more than 300 enterprises, including Adobe and Canva [1][3][4] Company Overview - Founded in 2021 and headquartered in San Francisco, Fal is a generative media platform aimed at developers, hosting image, video, and audio models through a high-speed inference engine and unified API [4] - The company has raised multiple rounds of funding, with the latest round in October 2025 amounting to $250 million, leading to a valuation exceeding $4 billion [4] Transition from Data to AI - The initial focus was on data infrastructure, but the emergence of models like DALL-E 2 and ChatGPT prompted a shift towards inference, allowing users to utilize pre-trained models without extensive data preparation [6][9] - The decision to pivot was challenging, as the company had existing paying customers and two products running simultaneously, leading to confusion in communication [7][8] Product and Growth Strategy - Fal identified a significant market opportunity in generative media, particularly in video generation, which is seen as a new blue ocean market with rapid growth potential [11][17] - The company opted for an API-based approach to provide ease of use for developers, optimizing workflows while maintaining control over the code [13] - The focus on video generation has led to increased computational demands, necessitating further optimization of their systems [16] Commercialization and Sales - Fal has transitioned from a pay-as-you-go model to annual contracts to ensure revenue stability, with a focus on long-term commitments from enterprise clients [25][26] - The company actively promotes new model releases as marketing opportunities, aiming to be the first platform to support new models [24] Team and Culture - The company maintains a unique culture with no dedicated engineering managers, promoting a collaborative environment where all engineers contribute to coding [33] - Recruitment focuses on individuals with a passion for optimization and experience in database or system-level work, fostering a strong technical team [35][36]

Core Insights - Google has launched the next-generation general intelligence agent SIMA 2, which integrates deeply with Gemini, enabling it to understand and execute commands in virtual worlds, plan actions around objectives, and interact with players while continuously improving through trial and error [1][2] Group 1: SIMA 2 Capabilities - SIMA 2 can understand and execute complex, multi-step commands in games like "Minecraft" and "ASKA," significantly improving upon its predecessor SIMA 1, which struggled with such tasks [1][2] - The agent has been trained using a large dataset of human demonstration videos with language annotations, allowing it to develop initial "conversational collaboration" capabilities, explaining its intentions and next steps to users [2][4] - SIMA 2's task completion success rate has shown significant improvement compared to SIMA 1, demonstrating its enhanced ability to follow detailed instructions and provide feedback, akin to interacting with a real player [5][9] Group 2: Self-Improvement and Learning - SIMA 2 employs a closed-loop system of "trial and error + Gemini feedback evaluation" during training, allowing it to learn and complete more complex tasks over time [11] - The experience data accumulated by SIMA 2 can be used to train future, more powerful agents, establishing a foundation for a "general agent" capable of adapting to any world [13] Group 3: Path to General Intelligence - The combination of Gemini and SIMA 2 offers a compelling approach to achieving embodied intelligence by training agents in controlled, low-cost virtual 3D environments, where they can gather interaction data [14] - SIMA 2's ability to operate in various gaming environments is crucial for developing general embodied intelligence, enabling the agent to master skills, perform complex reasoning, and learn continuously in virtual worlds [15] Group 4: Implications for Robotics - The capabilities developed by SIMA 2, including navigation, tool use, and collaborative task execution, are essential modules for future intelligent agents to achieve "intelligent embodiment" in the real world [16]

Core Insights - Google has officially released Gemini 3.0 Pro, which emphasizes enhanced reasoning and understanding capabilities, allowing users to receive higher quality responses without needing detailed prompts [1] - In authoritative benchmark tests, Gemini 3.0 Pro achieved a leading score of 72.1% in factual accuracy assessments and 23.4% in mathematical tests, indicating its reliability across various disciplines [1] Benchmark Performance - Gemini 3.0 Pro outperformed its predecessor Gemini 2.5 Pro and competitors like Claude Sonnet 4.5 and GPT-5.1 in multiple benchmarks, including: - 91.9% in scientific knowledge (GPQA Diamond) [2] - 95.0% in mathematics (AIME 2025) [2] - 81.0% in multimodal understanding (MMMU-Pro) [2] - 72.1% in parametric knowledge (SimpleQA Verified) [2] User Experience and Practical Applications - The model's capabilities allow users to generate product interfaces solely through natural language prompts, achieving a level comparable to professional UI designers [5][6] - Gemini 3.0 Pro can create interactive games from static images, demonstrating its versatility and ease of use [7][8] - The model significantly reduces the time and effort required for product development, enabling rapid prototyping and deployment [9][10] Future Implications - Gemini 3.0 Pro represents a shift in software production methods, lowering the marginal cost of trial and error in product development [10] - The model is expected to set a new standard in the industry, potentially transforming the capabilities of individual developers and small teams [10]

Core Insights - Google is nearing the release of its Gemini 3.0 model, with the latest checkpoint "Gemini 3.0 Pro Preview" expected to be the final test version before the official launch [1][3] - The anticipated release date is around November 18, 2025, coinciding with the discontinuation of older versions [2][3] Performance Enhancements - Gemini 3.0 Pro shows significant improvements in overall performance, particularly in code generation, front-end interface construction, and multimodal reasoning tasks [5] - The model can generate complex planetary visualization scenes with real-time parameter adjustments, showcasing capabilities that are currently unmatched by other models [5] - It can also create an interactive Rubik's Cube simulation that adheres to real physical rules, indicating a leap towards next-generation interactive intelligent systems [6] Creative Capabilities - The model possesses full "composition + performance" abilities, autonomously creating and playing original music based on user instructions [7] - It can generate a "creative wormhole" simulation that is visually surreal and logically coherent, further emphasizing its creative potential [8][9] - Compared to other models, Gemini 3.0 Pro excels in generating both visual and audio content simultaneously, achieving higher quality and consistency [9] Visual Quality Trade-offs - Recent tests indicate a decline in image and visual modality generation quality, with notable differences in detail and aesthetics compared to previous versions [10] - The decision to prioritize capabilities in code and multimodal reasoning over visual generation is seen as a strategic product trade-off, given the presence of the Nano Banana model for image generation [11] Market Position and User Engagement - Since the launch of ChatGPT, Google has been perceived as a laggard in the AI space, prompting significant internal restructuring to integrate generative AI into core products [13] - Gemini applications have reached 650 million monthly active users, an increase of approximately 200 million since July, indicating a narrowing gap with ChatGPT's 800 million weekly active users [13] - Google's image generation applications, particularly Nano Banana, are performing well among younger demographics, suggesting a shift in user engagement [13] Competitive Landscape - The release of Gemini 3.0 is seen as a critical opportunity for Google to reclaim its position as a leading player in the AI industry, especially following the lukewarm reception of ChatGPT 5 [14] - The success of Gemini 3.0 could establish a generational divide in AI capabilities, particularly in code generation and multimodal creation, which would be detrimental to OpenAI's competitive standing [14]