很多人还很迟钝。不知道大趋势是什么。 以后聪明人用AI说话，还不如少说。 用AI说只会显得人很low，而且是随着用AI的人越来越多，形象就越来越low。 物以稀为贵，语言也是如此。AI时代行动和结果又比表达贵。你表达还选择用最廉价的表达。等于宣告你这个人很廉价。好蠢。

我看了一些写杂文写的好的博主。不管是微博、公众号，还是现在的推主。他们的写作风格都有自成一体的特性，不需要写成结构分明的长文，不需要刻意缩短句段长度，不需要反复设问、凹金句。 很多就只是一大坨论述堆在一起，几大坨论述，用点、洞察、怒骂。一气呵成。 没有什么AI写作者写得出这样的东西。

GLM-5.2中文写作测试。 我的提示只有【定位是自由职业写作者分享自己的心路历程。用你认为最自然且有感染力的方式写】 你们觉得这个像谁家的啊？ 文章内容别当真，我不完全认同（

冰之城墙真好看啊

我很早就说过dario和a社不是虚伪，不是蔫坏。那种商业奸雄画像更符合奥特曼。a社就是实诚，或者说蠢。更准确的说是典型的科技nerd学了一套白左叙事。这也就罢了，然后他忽然因为产品的潜力，觉得自己有资格对全人类放话了。 于是沟通障碍的白左nerd开始随口胡扯，这反而并非出于商业利益的理性考量。

虽然我觉得5.2不错，但只要稍微不那么傻白甜的人都应该清楚，国产模型保持开放的唯二理由是 第一，整体处于落后。 第二，能力不足以威胁国家主权和国家机器的运行。 如果Mythos指向的未来AI能力为真，而它今后可以被国模超越，那么会发生什么是不言自明的事情。 论造墙，中美不过是邪恶双生子罢了

如果说大模型将来必然动摇国家安全，上升到核武器/国之重器的级别，被主权国家圈地圈养。 那未来真正的开源之光必不可能在中国，也很有可能不在美国。因为很显然，如果大模型的能力这么强。big brother长期来看必然不可能允许这种杀器人人拥有。连菜刀都要登记呢。 所以在骂美国和A社的时候，国人也不用乐呵呵地给自家AI自吹自擂，你就是牛上天了，也还是走不出天庭的猪圈。

你们都在说A社禁令怎么怎么样，fable怎么怎么样 而我要说是，GPT-5.6怕是这半个月内甚至一个月内推出都成问题了 因为很显然OpenAI的下一代模型在绝对性能上不可能和fable有代际的差异。且不说真正的Mythos它能不能赶上，至少没有道理fable要禁止，而5.6不会被禁止。

No, I'm Not Joining the AI Cult. A lot of Chinese AI worshippers these days seem downright eager to picture themselves as a single screw in some grand future machine — a pretty bizarre thing to be excited about, given that China's tech giants are laying off in waves and "niúmǎ" (牛马 — Chinese-internet shorthand for "beast-of-burden tech worker") has become Big Tech's default self-description. The reason they read this as a glorious Utopia instead of a horrifying Dystopia is the fantasy that — sure, I'll be a screw in that future, but AI will spend its long evolution easing us into that world; and AI, agents, robots, they'll all just be screws too. Just look at that tiny AI homunculus sitting at the computer in Marvis (Tencent's hit new Agent product, which renders each agent as a little cartoon worker shuffling around an on-screen office) — isn't he just as much of a workhorse as we are? So even when the grand machine swallows us whole, I — led by the hand into that era by AI — get to stand there as AI's equal. In a narrative like that, maybe they find a kind of hope. But speaking from my soul as a creator: even in the moments when I most worshipped AI, when I most desperately wanted AI to change my fate, I never for one second believed AI could shake a single thread of human subjectivity. From beginning to end, I've learned and used AI under one simple belief: AI exists, entirely and unambiguously, to make creators' lives better. If it doesn't, then all the talk of "fusion," "symbiosis," "mindset shift," "workflow integration," "self-iterating recursive evolution," "human in-or-out-of-the-loop" — every last bit of it is just PUA rhetoric, and every technical or engineering breakthrough on top of it is meaningless. Because I hold that belief, the moment AI fails to meet my needs, I refuse to entertain any wishful thinking. I'd sooner throw AI out altogether and cancel every coding subscription I have than PUA myself with "I just haven't adapted to the AI-era mindset yet! I still need to slowly grow into the new world AI is leading us into!" For ordinary Chinese workers and small-business owners, reality is probably far more brutal than any version of this fantasy. The "worshippers" I'm talking about are mostly the self-media influencers and paid-knowledge hustlers who currently own the AI conversation in China. I'll grant the English-speaking world likely has its own version of this crowd — but in Chinese-language circles, the particular cocktail of ignorance-dressed-up-as-expertise and outright dishonesty makes them especially insufferable. For someone like me, just trying to make a living off words, the rule couldn't be simpler: If AI is useless, it's useless. It really is that simple. I think current LLM writing is dogshit — so unlike those AI-cult bloggers who reflexively post some godawful AI-generated fable or sci-fi vignette the second a new model drops and then perform a little ecstatic swoon over it, I'm not playing along. Stupid is stupid. A year of zero progress means the bit is dead. I'm not sacrificing my taste to lick whatever slop some coding-tech company just cranked out. The fact is: today's LLM-driven AI is nowhere near AGI — and even AGI wouldn't be a god. If AI really were a god, the ultimate productivity revolution, then it ought to guarantee that we who write for a living can sleep in every morning, write our novels till dusk without a single worry, then stroll humming down to the river to wait for a beautiful dinner with the one we love. But it can't do any of that. A life like that, in the end, only comes from your own effort. So I shut the door on the fantasy. Instead of drowning in the Chinese AI self-media bullshit flooding X and WeChat, I sit at home and hammer out my keyboard, one character at a time.

现在很多 AI 崇尚者特别愿意把自己想象成是一个未来的宏伟机械中的一颗螺丝钉。 他们不认为这是可怕的Dystopia而是美妙的Utopia的理由是，他们幻想着虽然在那样的未来里我是一颗螺丝钉，但AI会在漫长的进化中带我们融入那个世界，AI、Agent、robot，未来无非也只是一种螺丝钉罢了。 就像现在Marvis里那个电脑前坐着的迷你 AI 小人，你看他是不是也是和我们差不多的牛马吗！所以未来即使宏伟机器吞噬了我们，被 AI 牵着进入那个时代的我也可以是和 AI 平等的存在。 在这样一种叙事中，他们或许感到了一种希望。 但从我的创作者灵魂来说，我即使是在最为崇拜 AI、最为期待 AI 来改变我命运的时刻，都不认为 AI 能动摇人的一丝一毫的主体性。 从始至终，我是抱着一个单纯的信念去学、去用AI的： AI 就是彻彻底底来给创作者的生活谋福利的。如果不是，那么什么「融合」、什么「共生」、什么「思维转换」，什么工作流融合，什么「自迭代递归进化」、「人在or不在的回路」就统统全都是PUA的话术，它的任何技术和工程的突破与优化，也都没有意义。 正因为有这种理念，在发现某个阶段某个时刻 AI 没有办法满足我的需求的时候，我绝不会心存任何侥幸。我宁可完全扔掉 AI 退订掉我的所有coding plan，也绝对不会PUA自己——是我还没适应AI时代的思维！还需要慢慢融入AI引领的新世界！ AI 如果没有用，它就是没有用，单纯就是这么简单。 就像我认为现在的LLM写作能力就是狗屎，那么我就绝对不会像这些AI崇拜论博主们一样不管出来个什么版本都先发个AI写的烂俗寓言或者科幻小说然后装模作样地陶醉一番。蠢就是蠢，一年毫无进步就是没活儿了，我不会牺牲自己的审美去舔一些编程科技公司造出来的烂活儿。 事实是现在LLM底层驱动的 AI 远未达到 AGI，而即使 AGI 也不是神。 如果AI是神，是一场究极的生产力革命，那么它就应该保证我们这些码字人都衣食无忧，每天睡到自然醒，醒来就可以毫无忧虑地写自己的小说到黄昏，然后哼着歌遛弯到小河边，等着和自己的爱人享受美好的晚餐。 但它其实什么都做不到，实现这种生活终归只有靠人自己的努力，所以我选择把幻想拒之门外。老老实实地在家里一个字一个字敲自己的键盘。