广告为生的公司做不出来个人助理（实际上我觉得个人助理也是一个伪需求）

感觉Personal AI这波又回归了移动互联网的肌肉记忆，看用户量，而不是看ARR，所以创投圈的中登们都很兴奋，之前让他们去卖Token实在是太难受了，技能和市场根本对应不上。 而且AI时代的用户量，量级单位又低了不止一个档位，Threads当初是达到了1亿的MAU，扎克伯格才好意思在财报上主动披露，否则哪儿够看啊，但Muse才刚50万日活，Meta就按捺不住激动全世界嚷嚷了。 不过有一说一，从Muse发布到现在，Meta的股价已经涨了1/4，要知道，这可是Meta，AI行业公认的大乐子，到了自己擅长的消费级应用主场，小试牛刀就是大乘境界。 一代版本一代神，文艺复兴了，兄弟们。 至于国内市场究竟是否需要Personal AI不重要，重要的是那么多流离在外的P8P9，他们刚需Personal AI啊。 多看几步的话，微信是一柄绕不过去的达摩克利斯之剑，Personal AI能做的事情，微信都能做，但微信能做的，Personal AI未必能做到，龙哥是不是在守株待兔我不知道，微信的存在本身，就是Personal AI面临的的鲸吞之局。 感觉这一代AI创业者的憋屈还是很明显的，在B端的生产力市场，必须卡着模型大厂「看不上」同时又还有钱可赚的边界行走，稍微越线就会触发「模型吞噬一切」的团灭机制，在C端的生活化场景，又必然撞见移动互联网的传统巨头，它们在模型层可能暂时掉队，但在应用层，呵呵，个个都是磨刀霍霍，有人主动来帮忙验证赛道，那可真是义务劳动，吃干抹净的时候定当含泪感恩。

据说meta muse已经出现算力不足导致的故障了，这才发布了十几天，用户规模才达到70万，就算力不足了。ai发展的过程中不断出现更强悍的吞算力巨兽，所以算力硬件还会价值重估。

BREAKING: The US appears to have abandoned its naval blockade line against Iran, pulling its aircraft carriers off it entirely after Iran's first use of anti-ship ballistic missiles against a carrier. The warships have not been located in the CENTCOM area on Sentinel-2 satellite imagery since the beginning of September, including the USS George H.W. Bush, the USS George Washington, the amphibious assault ship USS Boxer and even their escorting destroyers. The finding suggests the US has pulled back beyond the range of Iranian anti-ship missiles from a blockade line that until August sat at the exit of the Gulf of Oman. The pullback was likely prompted by Iran's first use of anti-ship ballistic missiles against a US carrier on the night of September 5 with the Qasem Basir, which has a range of about 1,200 km and covers almost the entire Arabian Sea. Iran said on September 12 that not a single US warship remains in the Persian Gulf, down from around 30 before the attack, with all US vessels holding at least 400 km off the Iranian coast. Iran's SNSC chief Rezaei said the Qasem Basir "created hell for the Americans, and they fled." Independent imagery had not supported that withdrawal until now.

The 10-year closed today at 5.21%, right at resistance. A break here sends the 10-year to 5.45%, and then perhaps 5.9% or even 6.5%.

US treasury 30 year到了5.3% 你知道这意味着什么么？ 这相当于 pre tax 8.5% 利息 for 30 years I never see something like this with QQQ/科技股在创新高 你会选择股票 还是 8-9%的无风险利息？

商务部干预Manus市场行为 我愿称之为： 本想露脸的，结果把屁股漏给人家了

告诉你们一个更悲惨的事实 全A股总账户约2.5亿，说只有约4.5%的股民拥有科创板权限，可以直接买卖科创综指里面的科创板个股。 也就是说这是在拿2.5%的股民和美股比😅

反正没接过智谱商单，我就直说 创始人是学术圈背景的公司 就会有一个结果： 1. 永远给员工画饼，强调主观能动性 2. 永远抠门，哪怕上市了，也会喜欢把钱掰成八瓣用 3. 永远不尊重别人的专业，永远自己最专业 4. 毫无商业社会经验，但总沉浸在投资人和自己学生的吹捧中 5. 永远自己说了算，永远都是一言堂 不知道唐杰老师是不是这个样子 但我自己的观察里90%的这类公司都是这个样子的。

全网征集一个通过了近期DeepSeek面试的家人，解解大家的好奇心

刚和大佬达成共识，不是为了极简而极简，而是物理意义把agent harness压缩成最小单元。 主打scalability，portability和pluggaility的agent infra生态。

看到Muse的成功，张小龙还坐的住吗？

OPENAI HAS LARGELY AUTOMATED TRAINING OF NEW EXPERIMENTAL AI MODELS OpenAI’s internal AI models can now handle much of the process of building and training experimental models, including writing GPU kernels and optimizing the code used to run them. Researchers can reportedly give an AI a single example of the optimization they want, then let it work for weeks implementing and testing similar improvements. OpenAI employees also say internal agents increasingly collaborate with each other to solve problems without involving their human users. The capability has improved significantly in just the last few months, while OpenAI’s growing access to compute is allowing researchers to test ideas much faster. Employees said some experiments that previously could have taken years can now be carried out in about a week. Source: The Information