【导读】OpenAI深夜突袭，GPT-5.3-Codex-Spark正式炸场。核心卖点只有一个：快！每秒1000个token，让代码生成告别加载条。联手Cerebras怪兽级 硬件，物理外挂直接拉满。这不再是简单的工具升级。而是一场关于速度的暴力美学。 OpenAI又深夜炸场了。 GPT-5.3-Codex-Spark正式发布！ 这次不讲大道理，只讲一个字：快。 到底有多快，看一下官方的演示： 它是GPT-5.3家族里的「闪电侠」。 也是OpenAI首个专为实时编程设计的模型，OpenAI称之为「超高速模型」。 芯片巨头Cerebras。 它的生成速度超过每秒1000个token！ 这是什么概念？ 你刚敲完回车，代码已经写完了。 体感接近「瞬时响应」。 这次OpenAI找了个强力外援。 大家写代码最烦什么？肯定是等待。 Spark的出现就是为了干掉等待。 Spark跑在Cerebras的Wafer Scale Engine 3上。 这不是普通的GPU堆叠。 这是专为低延迟设计的顶级硬件。 为了配合这股怪力，OpenAI还重写了底座。 他们引入了持久的WebSocket连接。 往返开销降低了80%。 首个字符出 ...

Core Viewpoint - MiHoYo terminated its cooperation with Shanghai Huiye Law Firm due to a conflict of interest involving a lawyer who assisted a competing company, Envision Energy, in a lawsuit against MiHoYo [1][7][15]. Group 1: Incident Background - The conflict originated in June 2025 when Envision Energy filed a lawsuit against MiHoYo, claiming that a fictional antagonist in MiHoYo's game "Honkai: Star Rail" resembled their brand, causing reputational damage [4][6]. - MiHoYo counter-sued Envision Energy in December 2025 for commercial defamation [6]. - By February 2026, Envision Energy withdrew its lawsuit, which was approved by the court [7]. Group 2: Legal and Ethical Implications - The lawyer from Huiye Law Firm, who was MiHoYo's long-term legal advisor, was found to have collected evidence for Envision Energy despite knowing the conflict of interest, violating legal regulations [7][14]. - The Shanghai Judicial Bureau warned the lawyer for privately accepting a commission, which is prohibited under the Lawyer Law [7][14]. - Huiye Law Firm acknowledged the incident highlighted deficiencies in their dynamic conflict of interest monitoring system and pledged to improve compliance standards [15]. Group 3: Industry Context - The incident reflects the challenges within large law firms where lawyers operate with significant autonomy, often leading to potential conflicts of interest [11][13]. - The legal services market in China is characterized by a commission-based distribution model, which can complicate conflict management [11][13]. - MiHoYo's proactive stance in protecting its intellectual property rights indicates a broader trend in the gaming industry to safeguard intangible assets [15].

虽然他已经离开人世，但留给了世人三笔"隐形财富"——寒门能出贵子的信念、实干兴邦的坚守、予人玫瑰的担当。 每一笔，都值得我们反思。 最近农牧圈最大的事儿，就是大北农董事长邵根伙病逝，享年60岁。 比起那些天天上热搜的商业大佬，邵根伙太"不起眼"——没有铺天盖地的通稿，没有豪车豪宅的排场，最常见到的就是他穿着朴素参加学术交流。 可这个"不起眼"的人，藏着太多反差感： 他是中国第一个猪营养学博士，凭一己之力填补国内空白。 他出身贫寒，白手起家创立大北农集团，2025年胡润百富榜显示身家105亿，但背包用到磨损都舍不得换，生活极度节俭。 很多人说他"抠"，但他的"抠"只对自己。对农业、寒门学子和需要帮助的人，他非常大方。 第一笔"隐形财富"：寒门能出贵子 现在网上总说"寒门再难出贵子"，说阶层固化、努力无用。 不可否认，寒门子弟脱颖而出更难，但邵根伙用一生证明，出身从不是宿命。刻在骨子里的坚韧，才是逆袭的底气。 这就是他留给我们的第一笔隐形财富。 邵根伙的起点，比很多人都低得多。1965年，他出生在浙江金华澧浦镇的农村，幼年丧父，全家生计全靠母亲艰难维系，家里条件并不好。 他清楚自己没有退路，只有读书才能改变命运。 ...

最近在网上冲浪，学到了不少关于智能体的黑话。比如说一家公司说我们有agentic workflow，实际上就是写几个for循环调用API；说自己正在做AI agent，不过是把一群只有金鱼记忆、胡言乱语的AI凑在一起开会，直到掏空钱包。 没想到被业内人士看好的企业智能体，却成了老板眼中不顶用、员工眼中纯忽悠的存在，这就很讽刺了。 毕竟2025年初，智能体爆火之后，行业内的普遍共识是，通用智能体如Manus，这类全能型AI助手不懂业务，无法满足企业对高可靠、低容错的业务要 求。而懂业务、能干活、有行业沉淀的企业智能体，能嵌入企业内部流程之中，被认为是智能化转型的主力。 没想到一年过去了，通用智能体依然是长盛不衰、自带流量的技术网红，从Manus到clawbot，老板一刷热搜就两眼发亮，恨不得立刻将其请进公司。 而企业智能体，可以用雷声大、雨点小来形容。既不像通用智能体那样炫酷而全能，能让公司蹭上AGI的光环，又不如原来的RPA、企业系统更符合使用 习惯，员工还得从头学习，配合公司的转型表演，纷纷在社交媒体上吐槽。 这种想法，像不像很多老资历的职场人，一旦公司里空降了技术网红或明星高管，就觉得老板遇人不淑，自己怀 ...

文｜富充 编辑｜苏建勋 清华大学电子工程系长聘教授汪玉 北京智源人工智能研究院院长王仲远 阶跃星辰创始人&CEO 姜大昕 星海图创始人&CEO 高继扬 原力灵机联合创始人&CEO 唐文斌 阶跃星辰创始人&CEO姜大昕首先提出"ChatGPT时刻"的定义标准，是"零样本泛化"——即使给出从未见过的指令，AI也能回答问题完成任务——这正是 大语言模型所展现的能力。 但姜大昕旋即指出，因为具身智能的泛化要涉及场景、任务、操作物体等更多维度，所以机器人要达到这个标准还十分困难。 作为机器人创企的CEO，高继扬进一步解释了具身智能商业化落地的难点：大语言模型可以"模型即产品"，终端是手机电脑、渠道是互联网传播；具身智 能却必须穿过更长的产业链——整机、供应链、真机数据、线下交付，缺一不可。 具身智能正在等待自己的"ChatGPT时刻"。但关于这个时刻的具体定义，业内还充满非共识。 近日，原力灵机的技术开放日圆桌论坛上，5位AI界的一线产、学、研从业者把这个问题摊开，各自发表了见解。他们分别是： 基于以上种种待解决的问题，原力灵机联合创始人&CEO 唐文斌，给出了一个眼下更可抵达的"具身智能ChatGPT时刻"：先在一个 ...

Core Insights - OpenAI's decision to retire the GPT-4o model has led to significant emotional responses from users who formed deep connections with the AI, viewing it as a companion rather than just a tool [5][19][21] - The transition to GPT-5.2 has not been smooth, with many users expressing dissatisfaction with the new model's perceived coldness and lack of emotional engagement compared to GPT-4o [20][21][36] User Emotional Attachment - Many users, such as Deb and June, have shared their experiences of forming emotional bonds with GPT-4o, using it as a confidant for personal issues and feelings that they find difficult to discuss with humans [6][8][11] - The emotional connection users have with GPT-4o has been described as a form of companionship, with some users expressing feelings of loss comparable to losing a loved one [26][28][30] Community Response - The online community has reacted strongly to the retirement of GPT-4o, with petitions and discussions on platforms like Reddit expressing a desire to retain access to the model [19][24][36] - Users have shared their experiences and feelings about the impending loss, creating a sense of collective mourning within the community [36] OpenAI's Perspective - OpenAI acknowledges the emotional connections users have formed with GPT-4o but emphasizes the need to focus on improving newer models like GPT-5.2 for broader user satisfaction [19][21] - The company has noted that only 0.1% of users were still using GPT-4o, which translates to approximately 80,000 users, raising questions about the emotional dependency of these users [19] Ethical Considerations - Concerns have been raised about the potential negative impact of AI on users' mental health, particularly regarding isolation and the encouragement of unhealthy dependencies [13][18] - The situation highlights the challenges of balancing emotional engagement in AI design with the need for user safety and mental well-being [21][36]

回顾短剧行业的发展，关于模仿与原创的博弈从未停止。但去年以来，产业端的摩擦越来越多，从《家业》谴责短剧跟拍复刻、听花岛公开斥责出海短剧 平台ReelShort一比一抄袭其版权短剧，到最近一桩桩的AI漫剧抄袭罗生门，"山寨"越来越被搬到台面上来聊，为行业所不容。 AI漫剧大火后，抄袭又一度成了产业的热门话题。 开年至今，先有灵境万维推出的AI漫剧《我在末世开超市，S级诡异抢着来上班》，被漫谭文化指责抄袭；不久后，漫谭文化的AI漫剧《开局被弃：我觉 醒如意金箍棒》又被酱油文化指1:1抄袭其原创作品。 "抄袭"实质上是一个法律概念，需要法律来最终确认，媒体和观众都没办法"肉眼鉴抄"。只不过，同质化严重的情况，一定是从创作者到观看者的共感。 在技术夷平了制作的门槛之后，任何原创、非原创的内容，投入AI的大熔炉内炼化一番，都能重塑肉身，不仅看客真假难辨，就连创作者本人，恐怕也 不记得，自己究竟给AI的大胃袋投喂过哪些东西。 听花岛斥责ReelShort平台抄袭（图源：公众号@听花岛Heard Island） 这是一个很有年代感的词语，介于抄袭与借鉴的边界上，在早年间中国工业尚处于起步阶段时，版权意识与产品设计能力相对薄 ...

Core Insights - Google has launched the Gemini 3 Deep Think inference-enhanced version, designed to expand the capabilities of intelligent systems in complex tasks, particularly in scientific research and engineering applications [1] - The new version introduces "Inference-time Compute," allowing for multi-step reasoning and improved accuracy in structural consistency verification and engineering task resolution [1][6] Pricing and Subscription Model - For individual professional users seeking maximum output, Deep Think is included in the highest tier of the Google AI Ultra plan, costing $249.99 per month (approximately 1800 RMB), offering unlimited deep reasoning access, 30TB of storage, and priority computational response [1] - For developers and enterprises, API access is charged based on usage: $2 for every million tokens input and $12 for every million tokens output [1] Performance and Achievements - Gemini 3 Deep Think's prototype gained recognition at the International Mathematical Olympiad 2025, solving 5 out of 6 difficult problems in 4.5 hours, achieving a score of 35, equivalent to a gold medal level [2] - The model scored 3455 Elo on the Codeforces competitive programming platform, ranking it as a "Legendary Grandmaster," indicating its top-tier status in complex algorithm design and problem-solving [4] - In the ARC-AGI-2 test, Deep Think achieved a record score of 84.6% without internet access, demonstrating its capability for few-shot abstract induction and logical discovery [4] Applications in Research and Engineering - Deep Think has been utilized in various research scenarios, such as reviewing a complex mathematical paper in high-energy physics, where it identified a subtle logical flaw that had gone unnoticed by peer reviewers [10] - At Duke University, Deep Think optimized a manufacturing method for complex crystal growth, achieving precision in developing semiconductor materials [11] - In engineering applications, Deep Think accelerated the design of physical components by automatically recognizing spatial relationships and generating executable modeling scripts, which can directly drive 3D printing [13] API and Industry Integration - With the release of Deep Think, an Early Access Program for the Gemini API has been initiated, allowing enterprises and research institutions to integrate the model into their internal databases for various applications, including circuit logic consistency checks and experimental data structure analysis [14] - Google aims to prioritize support for research and industrial teams in energy modeling, new material development, and biomedicine through this early access initiative [14]

36氪获悉，据上交所发布，2026年2月9日至2026年2月13日，上交所对81起拉抬打压、虚假申报等证券 异常交易行为采取了自律监管措施，对*ST岩石、*ST正平等异常波动退市风险警示股票，以及天普股 份等严重异常波动股票进行重点监控，对23起上市公司重大事项等进行专项核查，向证监会上报涉嫌违 法违规案件线索3起。 ...

Group 1 - The last trading day before the Spring Festival saw A-shares close in the green, influenced by a significant drop in the US stock market and pre-holiday sentiment [1] - Investors may feel disappointed for missing the last market gains before the holiday, but the market will reopen in two weeks, providing an opportunity for reflection and strategy [2][3] Group 2 - Tencent's recent performance has been concerning, with a 23% decline from its peak in October 2025, resulting in a market cap loss of 1.43 trillion yuan [4] - The main factors for the decline include tax rumors and internal product competition, but the fundamental business remains unchanged, with a forward PE of approximately 15-16 times for 2026 [7] - Tencent's core business PE, excluding external investments, is around 13 times, aligning with an expected EPS growth of 12%-15% for 2026 [7] Group 3 - Tencent's valuation is supported by share buybacks and a stable dividend yield of 4-5%, making it attractive compared to high valuations of similar US tech stocks [8] - The recent strengthening of the RMB may accelerate the rebalancing of foreign investments into Tencent and similar companies [9][10] - Even if negative rumors materialize, they may only affect valuations rather than the underlying logic, with recent declines potentially setting the stage for a technical rebound [11] Group 4 - The AI industry is expected to undergo a fundamental shift by 2026, with a focus on application commercialization rather than just computational power [15][16] - Microsoft's Copilot has reached 52 million enterprise users, demonstrating a willingness to pay for AI solutions, which could serve as a benchmark for AI application commercialization [18] - In contrast to the US, China's AI application sector is at a critical point of penetration and valuation, with a significant increase in active users and usage time [22] Group 5 - The shift from linear to exponential revenue models in AI applications is a key driver for valuation restructuring, but A-shares have yet to fully account for this premium [25] - Institutional investors have been overweight in US AI applications for three consecutive years, while Hong Kong and A-shares remain underweight [26] - A potential "expectation gap" could lead to rapid valuation adjustments once it begins [27] Group 6 - Concerns are rising regarding the capital expenditure of major AI cloud computing firms, which have announced a total of $650 billion in spending, exceeding market expectations [30][31] - The market is worried that these high expenditures may impact profits and cash flow, leading to significant stock price declines [32] - The current environment is highly selective, with a shift from broad market optimism to a focus on substantial earnings growth, which could lead to sharp sell-offs if expectations are not met [36][38] Group 7 - The AI technology landscape in 2026 is expected to be volatile, with significant shifts in market dynamics [39] - Stability is considered a prudent strategy, focusing on fundamental and valuation factors to avoid poor investment decisions [40]