AI 时代的想象力正被逐步释放。 从最初的文本生成，到文件与工具调用，再到以自然语言驱动的小程序构建，人类与 AI 的交互形式不断扩展。而最近，这条路径开始指向一个更具冲击 力的方向——可用自然语言直接生成一个可供进入、探索与改变的世界。 北京时间 1 月 30 日凌晨，Google DeepMind 向外部开放了 Project Genie。这是其世界模型（World Model）研究体系中，首次以可交互形态对公众开放的 实验性原型，也被视为 Genie 系列的阶段性成果。 如果说过去的生成式 AI 主要解决的是「内容如何被生成」，那么世界模型开始触及的，是一个更底层的问题：当视频不再只是内容，而成为空间，我们 该如何重新理解「媒介」本身？ 行业前瞻：视频从「观看」变为「进入」的空间 在 Andreessen Horowitz（a16z）发布的 2026 年前瞻观点中，视频被反复提及。但这里的「视频」，已经不再等同于短视频或长视频，而是一种可被进 入、可被操控、可持续演化的空间媒介。 a16z 合伙人 Yoko Li 说，「到 2026 年，视频将不再只是被动观看的内容，而会变成一个我们可以真正'进入'的空 ...

AI 造 AI，顺便接管你的电脑 就在刚刚，硅谷 AI 圈上演了一出「火星撞地球」。 OpenAI 和 Anthropic 像约好了一样，同时甩出了自家的重磅更新：Claude Opus 4.6 和 GPT-5.3-Codex。 如果说昨晚之前，我们还在讨论怎么写好 Prompt 辅助工作；那么今天过后，我们可能需要学会如何作为老板去管理 AI 员工。 就在昨天，Sam Altman 刚在 X 平台上凡尔赛了一把 Codex 的「百万活跃用户」里程碑。短短一天后，OpenAI 再次乘胜追击，扔出王炸—— GPT-5.3-Codex。 技术文档里藏着一句极具分量的话：「这是我们第一个在创造自己的过程中，发挥了关键作用的模型。」 说人话就是：AI 已经学会了自己写代码、自己找 Bug，甚至开始自己训练下一代的 AI 了。这种自我进化能力，也直接体现在了一连串跑分数据上。 还记得那个模拟人类操作电脑的 OSWorld-Verified 基准测试吗？前代模型只有 38.2% 的准确率，连及格线都够不上。但这次，GPT-5.3-Codex 直接跳涨到 了 64.7%。 要知道，人类的平均水平也就 72%。这意味着，AI ...

朱慧卿作（新华社发） 从AI训练师、AI产品经理、AI伦理审核员等新职业涌现，到"一人公司"等创业新范式兴起，人工智能 正在加速重塑职业图景。 近期，越来越多的AI平台开始招聘文科毕业生加入AI训练。 在北京市海淀区的一家AI大模型研发企业里，陈雅正对着屏幕逐条查看对话数据。与人们常说的"码 农"不同，她并非计算机相关专业出身，而是管理学专业毕业。 在陈雅看来，AI能提供海量信息，却常常在人机交互时缺乏"人味儿"和分寸感。将人文判断引入技术规 则之中，正是像她这样的AI人文训练师的日常工作。陈雅在沟通协调、团队管理等方面的学习和经 历，让她在判断用户语气、理解用户情绪方面得心应手。 当下，不少AI平台开始增加文科类训练师岗位。业内人士表示，理科类AI训练岗位侧重于增强AI的逻 辑推理、数学计算和问题解决能力，文科类AI训练岗位则侧重于提升AI在语言理解、表达和人文素养 方面的能力。 在招聘平台上，许多企业在"AI人文训练师""AI数据运营"等岗位招聘要求中，将文学、心理学、社会 学、人类学、新闻传播学等文科专业列为优先方向。 训练师给AI上"情商课" AI人文训练师是如何给AI上"情商课"的？ 当前，AI正加 ...

Core Insights - The AI industry is heavily investing in data centers, with the belief that the company with the most compute power will dominate the market in the future [1][6] Company Projections - Amazon projects $200 billion in capital expenditures through 2026, a significant increase from $131.8 billion in 2025, with investments in AI, chips, robotics, and satellites [3] - Google anticipates capital expenditures between $175 billion and $185 billion for 2026, up from $91.4 billion the previous year, indicating a substantial increase in fixed asset spending [4] - Meta forecasts $115 to $135 billion in capex for 2026, while Oracle projects $50 billion, and Microsoft’s recent quarterly figure suggests a potential $150 billion if trends continue [5] Investor Sentiment - Despite the high capital expenditures, investor confidence is wavering, with stock prices of these companies dropping as they commit hundreds of billions to AI [6][10] - The pressure is not only on companies like Meta, which are still developing their AI strategies, but also on established players like Microsoft and Amazon, as the overall spending levels are causing investor discomfort [7] Industry Dynamics - The tech industry views high-end compute as a critical resource for future survival, leading to aggressive spending strategies among major companies [6] - There is a growing concern that big tech firms will need to manage perceptions of their AI investments to maintain investor confidence, despite the belief in AI's transformative potential [10]

As an investor, it’s challenging to stay constructive when markets feel noisy, crowded, or overdue for a pullback. The instinct to brace for what might go wrong is natural.However, the biggest opportunities usually don’t appear when optimism is obvious — they appear when confidence is quietly rebuilding beneath the surface.That’s why I’m handing today’s Digest over to InvestorPlace Senior Analyst Louis Navellier. Louis argues that while headlines continue to fixate on risks, several important things are goi ...

BLOK's January Winners Point to the Future of ComputingETF Trends is now VettaFi. Read More --Pure- play miners remain tightly tethered to the price of bitcoin. Still, a growing subset of mining companies is beginning to trade on a different narrative: high-performance computing and AI infrastructure. That shift is showing up clearly in [the Amplify Blockchain Technology ETF (BLOK)], where several top holdings are evolving from simple crypto plays into broader computing platforms.As of January 30, the follo ...

Core Insights - The year 2026 is anticipated to be a pivotal moment for the AI agent industry, with significant policy developments, accelerated actions from tech giants, and increased capital market investments [1][3] - There are differing opinions on the current state of competition in the AI agent space, with some believing that leading players like ByteDance and Alibaba have initiated a peak competition, while others argue that the market is still in its infancy due to technological limitations and ecosystem challenges [1][5] Investment Trends - The popularity of AI agents is reflected in capital movements, with five out of the top ten investment sectors in 2025 being directly related to AI agents [3] - Approximately 20% of new unicorns are centered around AI technology, indicating a crowded and competitive landscape with over 120 domestic AI agent platforms [3][4] AI Agent Maturity Levels - AI agents can be categorized into three levels: "basic" agents for simple tasks, "upgraded" agents capable of more complex operations, and future "complete" agents that may operate autonomously like humans [3][4] - Currently, most domestic AI agents are still at the "basic" level, performing tasks such as PPT creation and translation, with only a few, like Doubao and Qianwen, reaching the "upgraded" level [4][5] Competitive Landscape - The competitive hierarchy is emerging, with ByteDance and Alibaba leading as "seed players" capable of providing comprehensive solutions, followed by specialized players and those offering technical tools [7] - The Chinese AI agent market is expected to be dominated by large enterprises, which will hold about 80% market share due to their clear business needs and data foundations [7] Advantages of Leading Players - ByteDance and Alibaba's competitive edge stems from their self-developed large models, extensive ecosystems, and profitable products that can be scaled [8][11] - Alibaba's ecosystem is seen as a "super entrance" for activating its vast commercial network, while ByteDance leverages its traffic advantages through platforms like Douyin and TikTok [11][12] Challenges Faced by Leading Players - Despite their advantages, both companies face challenges: Alibaba's ecosystem is heavily reliant on its own platforms, limiting cross-platform collaboration, while ByteDance's late entry into the market has resulted in a lack of deep industry knowledge [12][14] - The competition is fierce, with predictions that 90% of AI agents may be overshadowed by large models, making it difficult for smaller players to compete [13] Market Dynamics - The AI agent market is characterized by a complex and broad landscape, suggesting that no single company can dominate entirely, leading to a multi-track, differentiated competition [18] - While ByteDance and Alibaba currently lead, there remains room for other players to carve out niches in vertical markets and through technological differentiation [18]

Days after its push into legal automation rattled software stocks, Anthropic has unveiled what it calls its most powerful artificial intelligence model yet, signaling a deeper move into high-value professional work.What Is Claude Opus 4.6 and Why It Matters Amid AI BoomOn Thursday, the AI startup announced the launch of Claude Opus 4.6, a new version of its flagship model designed to handle complex financial research and a wide range of enterprise tasks that would typically take humans days to complete. Th ...

Core Insights - Anthropic has launched a new model, Claude Opus 4.6, which is gaining attention beyond the tech community due to its potential impact on various sectors, particularly legal data firms [1][2]. Company Developments - The recent release of a plug-in for the Claude chatbot, aimed at legal analysis, has caused significant stock declines for major legal data firms, with Thomson Reuters dropping nearly 16% and RELX falling 12% [2]. - Claude Opus 4.6 is designed to assist knowledge workers, enhancing capabilities in coding and addressing challenges faced by non-coders in applications like Excel and PowerPoint [5][6]. - The new model reportedly outperforms its predecessor, Claude 4.5, in key benchmarks, achieving the best results in 38 out of 40 cybersecurity investigations [7]. Market Reactions - Market participants are reacting strongly to advancements in AI, with a focus on identifying potential losers in the sector, as noted by analysts [3][11]. - The perception of Anthropic is still limited, with a poll indicating that less than 5% of the population is aware of the company [3][12]. - The market's reaction may signal a broader trend where only a narrow set of companies will emerge as winners in the AI space, raising concerns about a potential bubble [11][12].

OpenAI首席执行官Sam Altman宣布："GPT-5.3-Codex正式发布！ 编码性能再攀新高：SWE-Bench Pro评测达57%、TerminalBench 2.0达76%、OSWorld达64%； 支持任务中实时调控与动态更新； 运行效率跃升：完成同等任务，令牌消耗量较5.2-Codex减少超一半，单令牌处理速度提升超25%； 具备成熟的计算机实操能力。" 他还表示，"这是我们首款在安全防护框架的网络安全维度获评"高等级"的模型。目前，我们正试点推 出可信访问框架，并投入1000万美元API额度，助力网络安全防御体系的加速建设。" 责任编辑：陈钰嘉 他还表示，"这是我们首款在安全防护框架的网络安全维度获评"高等级"的模型。目前，我们正试点推 出可信访问框架，并投入1000万美元API额度，助力网络安全防御体系的加速建设。" OpenAI首席执行官Sam Altman宣布："GPT-5.3-Codex正式发布！ 编码性能再攀新高：SWE-Bench Pro评测达57%、TerminalBench 2.0达76%、OSWorld达64%； 支持任务中实时调控与动态更新； 运行效率跃升：完成同等任 ...