2025-week28
K1373 次列车停运三小时，乘客砸车窗通风被民警批评教育后放行，这种情况可以用救生锤吗？应担责吗？
说明大家都是聪明人，
列车停运后车上的乘警之所以不敢开门不砸窗，是因为他如果开门了万一出点什么事他是要丢饭碗的，
但乘客砸窗就不一样了，首先是个人行为，不会影响到乘务员，其次是为了自救，也不用怕啥赔偿处分，
然后民警把几人带走进行批评教育，一来对上面有了交代，不至于被说无作为，二来也没有进行处罚，不至于犯众怒，
大家都是聪明人，挺好的。
与之相反的例子，就是疫情期间找救护车上的医生借治疗仪的事，
你去找人家借人家肯定不会借啊，弄丢弄坏了人家医生要赔钱的，
但是你自己出于紧急情况哪怕去抢都比站在原地指责人强，车又没开走，何必道德绑架人家？
只能说跟聪明人相处确实要舒心很多。
周杰伦三十多岁和未成年的昆凌交往，为什么大众对周杰伦的评价没有崩？换作普通男性的话会被打上恋童标签？
2007年韩寒在博客里安利了一个人，日本av女优松岛枫。他用一篇文章讲述了松岛枫如何为男友导演梦想，甘愿牺牲名声出演av。
他不仅晒出游日本时买松岛枫作品的照片，文章结尾处将松岛枫称为“完美女友”。而且还很贴心在博客侧栏附上了松岛枫作品的下载链接。
在当年人人都不缺种子的时代，韩寒作为年轻人的意见领袖，带头宣传av还是造成了很大的网络影响。
可你猜这事怎么着？新浪博客只是过了一段时间才下架了他的这篇博客，并没有更严厉的惩罚。韩寒照样在网上激扬文字，想说啥就说啥。
2008年，陈冠希闹出了艳照门，被香港娱乐圈封杀。陈冠希的事业尽毁，可却在当时内地互联网尤其是男性网民中“声名鹊起”了。无数男性网民都把陈冠希奉为精神偶像。
我当时虽然还小，却网上冲浪很多年。我记得当时无论身边的男性学长，还是大一点的男性年轻人，没有反感陈冠希的眼里只有崇拜。无论线上还是线下，只要不是官方口径，我没见过有一个年轻男性批评陈冠希的，也没见过当时猫扑天涯贴吧哪个限制讨论的。
女网民可能反感，但她们在网络的声音微乎其微。
你是不是觉得当时网络风气之开放，像在听聊斋？
可我要告诉你，这连开胃菜都算不上，当时社会风气是天天吃重口呢？
我记得当时基本每个大型贴吧都有人义务分享种子，比如李毅吧。虽然吧务会很快删帖，但却是当时很多不会科学上网只会“好人一生平安”的人唯一满足看片需求的渠道。
微博刚兴起的时候，很多av女优都在国内都开通了微博，比如波多野结衣，长泽梓等。这还不是最离谱的，离谱的是当时有些av女优还能接国内的商演，甚至大型媒体还搞了av女优的接待活动，比如网易接待泷泽萝拉。
这些当然都是快播寿终正寝前的事情，可当时人的癖好绝不只是看快播这么简单。
那时候的明星说话也是不着调的，韩庚能公然说自己偷过电脑，陈思诚也当着主持人说对女性大不敬的话，张译也说过自己打过人等等。
这在当时对明星的星途没半点影响，观众也没人关心，仿佛明星的私德和观众的喜爱是并轨运行的。
我将当年的网络风气，对于年龄尚小，或网龄不够，或比较健忘的人来说可能是科幻片。但对当时的人来说，却是每天面对的日常。当时没有几个人靠互联网吃饭，却每天享受的是黄暴段子。
而关于周杰伦，那些年的网络论坛只有对他的歌曲评价，爱国审判，歌手PK等比较宏大的问题，互相对骂的很多。唯独没有几个人骂周杰伦女朋友多的。
当时不是炒作林俊杰喜欢SHE的Hebe么，而周杰伦跟Hebe在一起过，很多林俊杰的黑子还把这当做林俊杰不够男人的黑点。
我依稀记得，我小时候喜欢周杰伦的时候，他都是以风流著称的。不知道近些年，怎么就成了纯情了。当然当时媒体报道很夸张，信息不同步有失真的成分，但他的异性缘丝毫不影响他的地位。
有多不影响呢？
在我高中美术集训的时候，也在周杰伦承认昆凌的前夕，我记得台媒用了一个特别夸张的标题“周杰伦深夜出行，夜战数十嫩模”。我当时拿给了同班喜欢周杰伦的同学看了，他们都没什么表示，反而谈起了周杰伦的歌哪首好听。
上网搜了搜，除了依然看人骂他“结巴”，我没见依然有多少人说他风流成性犯了罪。
现在，你还觉得周杰伦谈了一个未成年女朋友，在当年是个很了不得的事吗？
这就是当时人的道德阈值。他们要是看向今天的人们，仿佛在凝视珠穆朗玛峰。
我知道说了这么多，还是有人不理解。
这是因为我们生活在仅靠情绪极化就能赚钱的营销时代，回望十几年前纯靠热爱发电的石器时代是有些陌生。
如今的互联网，是经历了十几年的互联网基建，介入了上十亿网民的庞大社区。不再是当年主要是年轻人，尤其是男性年轻人的时候。再加上如今互联网的“时间占据”越来越高，基本上也与全体国民的生活节奏重合。
这就相当于如今的互联网风气，等同现实中倡导的国民风气和公序良俗。互联网不再是当年少数人玩的异时空。
网络看似是越发达了，其实是每个人都失去了小众的表达，因为每个人都被迫站队。这也就意味着现实中那些上纲上线的情绪诱导者，在网络人群的多样性面前往往有奇效。
就比如战狼和殖人，男权和女权。他们很多人啥都不会，天天只会念经能赚得盆满钵满。
我不想评价对错，只想说这些人的水平放十几年前论坛经验值都刷不满。、
当然，网络的全面基建也不是没有好处，在以前网络舆情是不可能影响现实判定的。可现在比比皆是。可另一方面，网民断案就没有冤假错案吗？因为一句话，一件不涉及刑法的事，就被封杀很正常吗？
网络世界现在枝繁叶茂，但却像是没有生气无聊的大树缠绕着每个人。它的枝干是粗壮能庇护你，但你也没有挣脱的自由
We’re Still Underestimating What AI Really Means
Node 作者 Ryan Dahl 于 2025-06-14 发表的文章，他的上一篇文章发表于 2022 年。
Most people are focused on short-term gains. Another tech wave, another startup to spin up. It’s easy to frame AI as the next platform shift like mobile or VR. But that lens is much too narrow.
We’re living through what may be one of the most significant moments in history: the emergence of a new non-biological form of intelligent life.
And yet, it doesn’t feel like it.
There’s no cinematic score, no blinking AGI warning light. Just Slack threads, blog posts, and conference panels. It reminds me of witnessing childbirth - profoundly transformative, with some shocking moments, but also lots of mundane time waiting around.
Meanwhile, the models keep improving. I’ve been following this since DeepDream in 2015 where the similarity to psychedelic experiences was eye-opening. Since then: ResNets, GANs, AlphaGo, transformers, diffusion. Each expanded what machines can model and reason about.
Many still treat today’s models as narrow - powerful, but ultimately a tool like any other. A better search engine. A neat hack for creating images. But that’s a misunderstanding of what AI has become.
Machine learning - now rightly called AI - is a deeply general-purpose field. The same core techniques behind Midjourney and GPT share research lineage, and often architecture. This isn’t a stack of isolated tricks. It’s one evolving system architecture applied across language, vision, reasoning, robotics, and more.
These systems are built on a mountain of science: decades of research, countless failed experiments, and thousands of contributors. (I’ve even contributed a few failures myself.) And we haven’t found the limits yet - these models can already translate language, write poetry, generate high-definition video, and write deeply technical software, and so much more.
Mobile technology was transformative. But general-purpose synthetic intelligence is something else entirely.
And still, we treat it like a product cycle - the next wave of tools to write, code, and build. That framing is tempting, but it assumes a clear boundary between “tool” and… what? When a system can reason, create, and act through agents, at what point does the distinction become semantic?
The Turing test was passed, and almost no one remarked on it. For most of my life, that milestone felt impossibly far off – the thing that would prove AI had truly arrived. When we crossed it, there was no headline. Just another Hacker News thread.
This is not just another technology. It’s an inflection point in the story of life on Earth.
There is turbulence ahead. Disasters are coming. Jobs will vanish. Industries will collapse. The arrival of AGI may trigger a shockwave of scientific discovery – breakthroughs cascading so quickly, human ingenuity gets squeezed out. Beneficial on one hand, deeply troubling on the other. Where that leaves us, I don’t know. But it doesn’t change the trajectory.
We are building the first intelligent entities that didn’t evolve - we designed them. Humanity may never leave this solar system due to our intrinsically fragile biology. But our AI offspring might. It’s very possible it will outlast us.
Stop and take a moment. Look around. Recognize what’s happening. This is what it feels like to witness the birth of something beyond us. There’s no background music. But it’s happening anyway.
知乎爬虫
找到一款可用的爬虫，可以爬取知乎某个人的所有回答。
可以爬取出来，喂给大模型。
https://github.com/NanmiCoder/MediaCrawler
Prompt Engineering to Context Engineering
The New Skill in AI is Not Prompting, It’s Context Engineering
What is context engineering? While “prompt engineering” focuses on crafting the perfect set of instructions in a single text string, context engineering is a far broader. Let’s put it simply:
Context Engineering is the discipline of designing and building dynamic systems that provides the right information and tools, in the right format, at the right time, to give a LLM everything it needs to accomplish a task.
Context Engineering is
- A System, Not a String: Context isn’t just a static prompt template. It’s the output of a system that runs before the main LLM call.
- Dynamic: Created on the fly, tailored to the immediate task. For one request this could be the calendar data for another the emails or a web search.
- About the right information, tools at the right time: The core job is to ensure the model isn’t missing crucial details (“Garbage In, Garbage Out”). This means providing both knowledge (information) and capabilities (tools) only when required and helpful.
- where the format matters: How you present information matters. A concise summary is better than a raw data dump. A clear tool schema is better than a vague instruction.