2026-09-13 21:31:40

Keywords

• Noche UFC
• Ryan Garcia defeated Conor Benn

TOC

• • Keywords
• TOC
• THE SHOW
• ☕Gooddays
• El Libro
  • 游岩/倪芙瑞莲
  • 基拉里峡湾/凯文·巴里

THE SHOW

Road Trips Song by Drake ‧ 2026

☕Gooddays

↩️

• 西部片每年几乎都是冬季release 然后民主先进的比如少数群体 现代流行的种类更多是在夏天释出 前者保守派确实穿的也比较多，后者也是穿衣风格比较先锋大胆 #Tulsa King S4

• Noche UFC结束，乏善可陈，对于墨西哥人在独立日这天来说，结果也not bad

  • Zhu Rong Dec R3
  • Alexa Grasso还在进化，重击KD了法国人Manon Fiorot，那下次子弹不能及时回归的话，可能是下一个title contender, Mexican Boxing level...
  • 小宝状态也比较在线，还有几年的油在邮箱，相较来说FW的量级竞争性不太强，水平还算都比较平均。

• 重头戏是Ryan vs Conor的Zuffa Boxing，Conor Ben上来就是令人窒息的前压，Ryan在回击之余也无奈 只能搂抱和锁头来阻挡部分的逼进的前压 然后被裁判分开，除了常规的jab左手，这次还是后手的右拳结实打中，相较来说Conor Ben 的stance更严密一点，而Ryan很多时候手都很低，但是Conor在先手回去的间隙被cought，然后knock down。 本身也没有RG那么的skilled. 整体还是在odds意料之中。

  • Ryan Garcia defeated Conor Benn via TKO at 1:25 of the second round to retain his WBC welterweight title. The blowout victory took place at the T-Mobile Arena in Las Vegas. 1, 2

• UFC released Joshua Van vs Tatsuro Taira | FULL FIGHT

• 历史上第一次亚洲男子的冠军德比。想到了小量级的帕奎奥。wait：帕奎奥显然也不是英文名啊 像是拉丁语的 葡西， 菲律宾被拉丁语国家殖民过?

帕奎奥（Pacquiao） 这个姓看起来确实很有“拉丁语/罗曼语”的感觉，但它其实不是纯正的西班牙语或拉丁语姓氏。它的真正来源是：菲律宾宿雾语（Cebuano）里的单词 pakyaw，意思是“批发”“整批买/卖”。这个词本身又源自闽南语（Hokkien）的外来词。
后来在西班牙殖民时期（1849年西班牙总督颁布《姓氏目录》时），被正式登记成 Paquiao / Pacquiao 这种带“c”“q”的西班牙式拼写。所以它是本土词 + 华人借词 + 西班牙拼写改造的混合产物，看起来像拉丁语系的名字，但不是。菲律宾确实被“拉丁语系国家”殖民过——被西班牙统治了大约 333 年（1565–1898）。之后美国又殖民了几十年，所以现在菲律宾官方语言是菲律宾语和英语，但姓氏和宗教上仍明显保留着西班牙的影响。

Life is a vector requiring both force and direction. The pursuit of happiness sets the direction, but feeling joy provides the daily confirmation that we are doing exactly what we should be doing, for the company and for the teammates who energize our efforts.

亲贤臣，远小人，此先汉所以兴隆也；亲小人，远贤臣，此后汉所以倾颓也。

Terence Tao @tao 2d *

A new proposed competition for AI companies: rather than being the first to announce solutions to unsolved math problems, be the first to announce a new mathematical insight.

不止数学: 数学研究的核心价值在于理解命题为何成立、发展新概念与工具并传承知识，而非单纯生产「命题是否为真」的结论。如果模型输出证明的速度远快于人类理解速度，可能导致大量难以被消化、解释和教学的结论堆积，冲击数学家培养与知识传承体系。

El Libro

• 寻找救生艇
• Rate: 3/5
  爱尔兰文学特辑（单读32）
  寻找救生艇

出版社: 上海文艺出版社
出品方: 单读
出版年: 2022-12
ISBN: 9787532184682
页数: 299
装帧: 平装
定价: 56.00元
丛书: 单读丛书

第32辑《单读》重启世界文学之旅，继澳大利亚、英国、法国之后，来到文学传统丰厚且依然富有文学活力的爱尔兰，在旅行限制还未完全解除前，先用文学进入爱尔兰这片充满苦难但人们以无与伦比的方式向外敞开自己的土地。本辑邀请曾旅居爱尔兰、现居英国的作家颜歌和群岛图书出版人彭伦客座主编，译介在本 地文学读者里受到推崇但在中文世界介绍得还很少的、独具一格的十二位当代爱尔兰小说家及其作品：凯茜·斯威尼、露西·考德威尔、温迪·厄斯金、妮科尔·弗拉特里、约恩·麦克纳米、科林·巴雷特、莉萨·麦金纳尼、凯文·巴里、路易斯·肯尼迪、丹妮尔·麦克劳克林、简·卡森和梅拉图·乌切·奥科里。他们的写作关注大时代阴影下漂泊着的个体命运，以文学的眼光进入历史，将“北爱尔兰问题”、移民议题和晚期资本主义图景化为故事背景，刻画底层劳动人民、每个小人物在历史洪流中经历的伤痛，对尊严和救赎的渴望；与此同时，他们也在描绘当代人心灵困境上入木三分，那种在空虚的人生中无从找到自己存在意义的颓丧，那种想要与人相拥却发现人与人之间无法跨越的隔阂的无奈，我们都是命运相似之人，正在寻找一艘救生艇。

• If one of us goes to war we all go to war 这声音还在echo 爱尔兰狂人发起冲锋
• 从上次的爱尔兰特辑说起 从奥德赛 再到Conor
• 单独写一章节比编辑一整本书workloading还大 一方面是创造的过程比较messy 混沌中提取安排信息、
• absolutely genius：看到很对风格生活的场景描写 孤寂 幽静 自然动物 都很像 DNA写的关于新西兰的那段 （《消逝世界漫游指南》是英国科幻作家道格拉斯·亚当斯与动物学家马克·卡沃丁合著的非虚构作品） 然后想想 地理上爱尔兰和新西兰大概是不是对趾点？ 查了一下 确实是

游岩/倪芙瑞莲

• I was drinking schnapps in a bar with a woman who used iodine instead of lipstick to redden her mouth. When she spoke the skin between her breasts folded and unfolded like paper. I remember nothing else, except the story she told me.

• 

• 人少的地方 时而压抑 时而开放

• 在海边酗酒旅店酒馆等这些场景里面，除了天气，南美和北欧都比较像

基拉里峡湾/凯文·巴里

📌一个穿蓝色卡其裤的人从街上走来。我没认出他，但殡葬人朝他点了点头，我便也点了点头，而这人略略颔首，微微张嘴，发出一个类似呼气的声音，并不像是说了一句话或一个字，这举止在当地也不算奇译，但还是让我恼火。你可以说我没兴致搭理这种不露声色的人。殡葬人一直盯着蓝卡其裤男，直到他走远了听不见我们说话，才告诉我这个男人是波兰人，已经在爱尔兰待了十八年，他原本只是为了一个签了合约的项目才待在这里，却发现自己已半截被埋在地里，铲起来的土盖在了他头上。这种事怎么会发生，殡葬人也不知道。也许看似现实的出路，没有人喜欢的，不被在意的，会在轻易成习惯以后付出代价。也许作为遵循习惯的生物，我们不该以为时间给我们最可怕的压力，莫过于衰老。也许这就是导致人们最终染上毒瘾，或者肥胖成疾的原因。

📌旅馆有二十三个房间，地基略向西倾斜。如果你在任意房间的地板上放一罐豌豆，它会缓缓地滚向汹涌的大西洋

📌海浪已在拍打防波堤的顶部。房产经纪人曾拍着胸脯对我说，这个地方从未遭过水灾。当时我盯着那个老滑头的眼睛，相信了他的话。我曾经猜想——也一度期待——此处的生活最终能带给我写作的灵感。某种东西会在我的心里萌芽。我的写作将摆脱那种植根于城市的闷热性欲的无病呻吟和凌乱节奏。那类诗作让我在地方上的大学英语系小有名气。基拉里当地人也听说过我的诗作，却都不以为然，因为这里自古以来就不缺诗人。这里的每一块傻×岩石都在过往的某个时间触碰过某个癔症发作的顿悟追寻者的干瘦屁股。此间永远不缺声嘶力竭的傻×。
“就算要坐牢，你也想上她。”约翰•墨菲说。
当纳迪娅转身往厨房走去时，他又盯着她的臀部。
“约翰，我警告过你了。”我说。
“我也就随口说说。”他说。
他自讨没趣地端起了面前的黑啤。戈尔韦北部的人都性欲过剩。这是我的结论。人们粗鄙猥亵的言行堪比异教徒。当然，这样的传统由来已久。海边扭曲的岩石造就了他们扭曲的性格。据文献记载，萨克雷曾惊讶于爱尔兰乡下妇女在连衣裙下不着胸衣。她们毫无顾忌地亲吻初次见面的陌生人，丰满的胸脯左右摇晃

📌吧台前出现了一段阴影般的短暂寂静。
“过去十六年没有过，”他说，“今天也不会。”在我清醒的全部时间里，我都在力“临海旅馆”忙碌。我的呼吸急促，精神紧张，生活乱得一团糟。在诗人的语境里，我目前大致处于“漫长沉寂期”的中段一我的上一本诗集出版已是五年以前。每当我坐下来面对一张白纸或是电脑屏幕时，我总有想哭的冲动，而我并不总能压制这种冲动。荒凉的山峦，海水单调重复的韵律，患了精神分裂症一般的苍茫天空：这些都无法激发诗歌的灵感，它们激发的只有让人绝望的情欲和消极的思维模式。
真相一次又一次浮现在我的眼前：我天生是个城里人

📌想想这个国家为欧盟做了多少牺牲，”薇薇安•哈蒂说，“再想想我们跪在他妈的布鲁塞尔脚下，就为了一张操蛋的黄油券——券还没到手，这帮不知从哪儿冒出来的杂种就决定他们可以搬来爱尔兰的任何地方，抢走我们的工作！”

📌《送浴》是《尤利西斯》的第十章，与《荷马史诗》没有直接光联。在《奥德賽》中，“游岩”是奥德修斯（Odyseus）返回伊萨卡（ihaca）途中没有取道的路线，因为女巫喀耳刻（Circe）劝告他，那些漂流的岩石会酿成灾难，他不该从那儿走。《尤利西斯》第十章是对都柏林普通一日里支离破碎生活散点的、不连贯的描绘。角色们不停地移动，在都柏林城中形成不同路径。《尤利西斯》中的绝大多数角色都在这一章中出现。城中一日的背景是总督（当时国王在爱尔兰的代表）的骑兵队伍正穿过街道前往医院的募捐活动。
这组肖像中的人物大多是我在北京和爱尔兰生活时结识的——家人、朋友、学生、艺术家同行和策展人。他们被重新塑造成《游岩》中的不同角色。有的因存在相似之处，比如我曾经教过的双胞胎学生现在被塑造成了“迪达勒斯姐妹”（Dedalus siblings）；有的角色通过肖像的题目匹配，比如《夹克衫蓝》（Blazer Blue）中的人物就扮演了“布莱泽斯•博伊兰"（Blazes Boylan），不仅如此，他们也都是充满自信、爱打扮的男子；另一幅肖像中的形象骄做且权威，题为《鹰》，与总督的角色相吻合。

I need to be myself
I can't be no one else
I'm feeling supersonic, give me gin and tonic

Match day: Man_Derby

You cannot be interested and inconsistent.
Thanks for being an insider till the end!
Till next, stay safe and stay hydrated!

2026-09-06 18:05:39

Keywords

• 蓖麻
• ☕项飙
• 猎人与例外

TOC

• • Keywords
• TOC
• 蓖麻
• ☕项飙、吴琦 | 上海文艺出版社 | 2020
• 猎人与例外

蓖麻

那天下午，我没什么特别的事，随手点开了《绝命毒师》第二季第一集。屏幕上，沃尔特小心翼翼地处理着几颗斑驳的种子，提炼出一种几乎无色的致命粉末。字幕里出现了"蓖麻毒素"几个字。我愣了一下
小学时候，校园和路边，到处都是这种植物。叶子又大又厚，像展开的巴掌，茎秆高高的。最特别的是它的种子——椭圆、稍扁，表面布满灰白、黄棕、黑褐相间的花斑，活像蛇身上的鳞片，或者某种昆虫的壳。老师常在课余或者学有余力的时候，把我们叫去帮忙。收割的季节，果实裂开，种子掉得到处都是，我们蹲在地里一颗颗捡，装进蛇皮袋子，一袋子一袋子地往回扛。播种的时候，又是我们把种子按进土里。那时只觉得它长得怪，气味也有点腥，闻久了不太舒服，却从没想过它还能和"毒素"扯上关系。
查了以下这些种子主要是用来榨油的，还不是普通生活用的：蓖麻油不是普通食用油，而是工业上的宝贝——润滑油、涂料、塑料、化妆品、甚至曾经的航空润滑剂。种子含油量高，单价比玉米、水稻贵上不少，按斤算确实更值钱。可亩产低，销路又窄，普通农户种它未必比种粮食划算。它之所以在学校、路边、房前屋后随处可见，更多是因为耐旱、耐盐碱、好养活，加上上世纪国家号召利用零星土地种油料，学校也积极响应，才成了劳动课里最常见的"任务植物"。全株都带毒，种子最厉害，里面的蓖麻毒蛋白只需极少量就能要命。小孩误食几粒，大人吃上十几二十粒，都可能出大事。
回看《绝命毒师》里那几颗种子，老白给图库准备的
同一种植物，在不同的时间里，呈现了完全不同的面孔。

☕项飙、吴琦 | 上海文艺出版社 | 2020

↩️

• 项飙是人类学家。这本书偏社会科学，但读完之后很多地方的思考模式和自然科学里的东西高度同构——不是内容的相似，更多是方法论层面的共振。

• 除去运气，越来越觉得正常路径下的创新只有一个起点：思想。纯粹的、对问题的理解和追问。思想先行，然后慢慢搭建积累，形成地基——体系——然后涌现才会发生。这个过程不是线性的"努力→产出"，更像相变：持续的能量输入看似没有效果，直到某个临界点，整个系统发生质变。项飙和吴琦在《把自己作为方法》里做的事情，某种意义上就是这个过程的一次蒸馏——把一个学者几十年分散在不同语境里的思想逼成可以被检视的命题，用对话的方式固化下来。

• 如何理解这个标题？传播不是目的，蒸馏才是。这本书的价值不在于"项飙说了哪些观点"——观点可以用几条tweet概括——而在于它示范了一种知识的存在方式：通过持续的追问，把分散的经验、判断、直觉逼成可辩驳的命题。目的是蒸馏本身，也就是知识的存在性。

• What it meant to be an academic: productive, efficient, capable

• 学者的价值不全在产出效率，而在于能不能把关心的问题用自己的位置讲清楚。

• 按关注程度大致排序：AI、数学、物理、生物……这种跨领域的阅读让我越来越清晰地看到一个反复出现的pattern：工具和理论之间的伴生关系。两者不是谁先谁后的线性推进，而是像DNA双螺旋一样交替上升，互相催生。

  • 数学是最典型的案例。丘成桐自己讲科学史的时候，来龙去脉、历史发展的脉络梳理得极其清晰——数学的推进不是某个天才的灵光一现，而是工具和理论在漫长积累中反复交替。新的理论诞生新的工具，新的工具发现新的结构，新的结构催生新的理论。很多重要的证明都循这个路径完成。丘成桐说黎曼猜想现在解决不了，可能需要等到新的数学工具出现，到某个节点——所谓天时地利人和——问题才会得到极大的推进。Terence Tao的工作风格也印证了同样的逻辑：不是硬攻一面壁，而是在多个方向同时耕耘，等合适的工具和视角汇聚。
  • AI的发展是这个pattern在工程领域的镜像。经历了多次寒冬：算力不够，物理硬件不够，理论方法不够。底层条件不具备的时候，再多热情也不过是泡沫。直到算力、数据、架构都到了一定程度，基础条件具备之后，再加上运气催生的转折点——Transformer架构的提出、Dario的scaling law的经验发现——整个领域一飞冲天。每一个组件都不是从天上掉下来的，而是在各自的角落里慢慢积累到了位。
  • 冷冻电镜的故事，因为它展示了一条跨越工具、数据和算法的完整因果链。冷冻电镜作为一种观测工具，对蛋白质结构解析带来了革命性提升。由此积累的海量结构数据构建了蛋白质数据库。而这些数据库又为AlphaFold的训练提供了ground truth。一步步催生了可能是21世纪迄今最重要的科学成就之一。有意思的是，AlphaFold和当今的LLM某种意义上是平行产物——Demis Hassabis选了蛋白质折叠，Geoffrey Hinton那一脉走向了language model。不同的方向，同一代人的不同赌注。

• 研究进入深水区之后，问题的复杂度发生了质变。和数学一样——需要新的工具、新的理论才能继续推进。
  核心困难在于，深水区的问题似乎乎都是耦合的。越复杂的问题，越意味着一旦某个关键工具或理论出现，可能带来连锁反应式的推进。

猎人与例外

互联网这一波感觉到了某种高峰。所有人都是猎人——抓取信息、找到污点、hunt you down。歧视和优越感以信息不对称为武器，社交媒体把每个人同时变成了猎人和猎物。在这样的环境下，项飙说的"抵制符号化"变得格外切中要害。他讲到北大西门外大家排队拍照——那么热的天，那么嘈杂——"要把动物放在北大附近，它们肯定不会去西门外面，它们会跑到未名湖边的树林里，因为那里凉快。"奔向象征是奔向了文明，同时也奔向了牢笼。符号一旦确立就会被利用，和金钱一样被物化。所以他坚持：一定要抵制物化，抵制符号化，把领导力作为一个过程、一种实践，而不是一个头衔。
学术界的很多现象也在印证这个判断。项飙说大学老师变成了一个职业群体，而且是一个"比较不讲职业道德的群体"。"其实不用讲什么知识分子、精神导师，一个吧台服务员也有职业道德。"他期待知识分子边缘化之后变得更加"有机"——不再是纯粹抽象的知识分子，而是有具体的身份，和社会以某种特定方式联系在一起。有机就是有限的，不可能是总体性的思想概括。Gramsci说的真正的有机知识分子是技工、农技推广员、赤脚医生、搞底层写作的人——他们的能量其实很大，因为他们在里面，能一针见血地提出和分析问题。快递小哥也要想事情的，"我们有这样的人，而且我们有这样的渠道，我们要鼓励他们多写东西。"

"例外"："大学生在大学里的任务，不是树立norms，而是树立exceptions。你不是范例而是例外。我们的社会需要例外，你要代表这个社会去做例外。"大学给了位置和气氛，赚不了太多钱但有比较舒适的生活方式和一定的社会尊重——这些条件的目的不是让你循规蹈矩，而是让你大胆。"你的特色是讲别人不太敢讲的话。"但是学校学院没有前面铺路的人，没有已知路径，也不会在乎或者鼓励。

学者最重要的是把关心的问题，用自己的位置讲清楚。想清楚究竟我能做什么，我跟世界的关系是什么。我认为世界上所有的人都有这个问题，都得搞清楚自己是谁，否则都会有这种危机，除非完全盲目地被主流裹挟进去。

很多东西都是联系在一起的，比如法律过程、律师制度。我小时候觉得律师制度很奇怪。这个人是不是坏人已经很明显，为什么需要律师在那里辩护。但"辩"这个概念很重要，就是说一定要假设我们不知道事实的过程是怎么回事，然后通过"辩"来把事实过程明晰起来。过程明晰之后，我们可能会发现那些明显的结论可能都是错的！学术也是这样，不能靠直觉去判断，一定要去证明它，要去展示结论是怎么达到的。往往越是明显的结论，越难去证明，但是一旦证明了以后，会是很大的贡献。
关于80年代的精神遗产：
对我来讲，80年代的很多口号，那种大胆质疑的态度和气质，要求做制度性、结构性的变革，这都是80年代的精神遗产。那是精神层面的东西，从很强的原则出发，觉得现实应该超越，应该改变，指点江山、激扬文字的感觉。
关于抵制符号化：
不要去找象征性的领导。象征就是牢笼，奔向象征是奔向了文明，同时也奔向了牢笼。背后是一种非常野蛮的关系。所以一定要抵制物化，抵制符号化，要把自己的领导力作为一个过程、一种实践。
关于成为例外：
大学生在大学里的任务，不是树立norms（规范），而是树立exceptions（例外），你不是范例而是例外。我们的社会需要例外，你要代表这个社会去做例外。你的特色是讲别人不太敢讲的话，因为大学给你这样一个位置和气氛，你的任务是要大胆。
关于新加坡精神与行动：
大家都认李光耀厉害，他当然重要，但事情靠大规划，更靠一点一滴做出来。今天不做，今天不犯错误，就不知道明天能干到什么程度，唯一的办法就是去做。我觉得这是新加坡精神。它和牛津不一样，牛津有老本，对很多要后来居上的亚洲国家来说，新加坡很值得学习。
关于去本质化：
在今世，国族总比人命要长，还是得认真地介入，但在介入的过程中——我们用一个学术研究的词——不要把它本质化。本质化就是认为它从来如此，一贯如此，应该如此；而去本质化是说，它现在如此，相对如此，以及马克思主义说的，历史的如此。"历史的如此"的意思是，任何事物都有它的历史，兴起、发展和消亡，都是具体条件下的具体反映。
关于建设小世界：
建设小世界，是要投入的，不是整天喝喝酒、说说话就可以，需要有细致的行动计划、目标、资源，又不能太多。有很多协调工作去做。比方说大家组织了很多读书小组，就要借书复印，这是有人投入才能达到的。这就是生态，一定要结合，既要有明星式的人物，也一定有托底式的人物，但明星和托底的人一定要感觉到他们是平等的，没有上下级关系。
关于大学的功能：
大学就是给你一个环境，让你在人生比较特殊的四五年当中去探索自己，探索这个世界，允许你犯错误，允许你做疯狂的探索，让你对事情产生理解，当然也学到了基本的知识和技术。我认为大学的教学功能肯定高于研究功能，今后的研究应该还会发散出去，跟产业结合。大学不是去树立范例，而是要去寻找例外。
关于中国社会内部的多样性：
我本来还想研究从西北去东南沿海做阿拉伯语翻译的这一批穆斯林，这些孩子是在西北辍学的问题少年，经常打架，家里很担心他们学坏了，就送到清真寺学经，在那里学到一点阿拉伯字母、语法，然后突然在2000年代初有机会成了翻译。我想看这个群体对宗教的理解、对中国的理解，在这样一个经济全球化贸易的过程当中，和具体的有形的中国市场运作怎样搭配在一起，比如义乌市场、广州的天河市场等等。到最后，我关心的是中国社会内部的多样性。
关于"认命不认输"与当下：
我们为什么会焦虑，最直接的原因就是对今天没有清晰的认识，总觉得自己现在所处的地方不对，和自己认为的有差距。一个解决方案可能是佛学里讲的专注，对自己身边做非常细微的观察，当下的重要性也是在这里。
关于图景式理论：
我一直觉得理论有不同的表达方式，有结构严密的推理型的理论，还有一种是展示性、图景式的，民族志就是图景式理论，通过无数细节一笔一笔叠出来，就像一幅巨大的壁画，不能把它浓缩为一个结论，如果浓缩为一个结论，就会觉得这个结论本身毫无意思。
关于讨论与思想的有机性：
我觉得根据当下问题临时组织起来的小组讨论蛮重要，跟论文和课题都没关系。这一方面是时刻保持我们的思考，第二就是让思想有机，有机就是思想跟现实经验对得上号。一切思想都来自经验，这好像是很自然的事情，但要做好是需要训练的，得经常练，培养观察的精确性，快速地推理，看出破绽。我们现在缺的是这种不断的讨论，最大的敌人是急于求成，一切都想赶快有个结论，没有时间去磨炼了。

And~~~~~~~~ City win at home

Thanks for being an insider till the end!
Till next, stay safe and stay hydrated!

2026-08-30 14:09:24

Keywords

• NURMAGOMEDOV VS. SONG
• AI 拟人化讨论
• Cutting-edge science

TOC

• • Keywords
• TOC
• ☕Gooddays
• UFC FIGHT NIGHT: NURMAGOMEDOV VS. SONG
• City
• cutting-edge science
• 护城河 黑暗丛林法则
• 哲学 AI 拟人化讨论

Nonviolent Communication 2023

☕Gooddays

↩️

• “地球是人类的摇篮，但人类不可能永远生活在摇篮里。”
• 人类为什么总想离开地球？
• 回收火箭，为什么中国追了十年？

UFC FIGHT NIGHT: NURMAGOMEDOV VS. SONG

• 主持人和Herb之间小声的蛐蛐都被收音了 所以没宣布之前就知道是Draw

• Liu Ce TKO·R1 4:26, 最后一下视觉冲击很大，又一个LHW的，更多盼头，更多希望。

• Kai Asakura R2 TKO，算是Kai的正常水平，on feet的kai的水平肯定TOP5 in BW.

• Yan Xiaonan lost via KO/TKO·R1 4:49 by Denise Gomes，有点没有发挥出正常水平，没有找到节奏游走防反。

• Song Yadong KO R2 Umar Nurmagomedov，下潜抱摔的时候迎上upper cut，击晕KO，实力自信come from preparation, 然后轻松自如。但是感觉R1的时候也有很多不太好的，比如手比较低，如果像是刚刚Usman在PFL那种较低的高扫的话，很容易失误，Umar没有太多抓住这个机会，Usman边角也没有太多强调，但是Umar的飞踢还是非常skilled, 简直是把脚当手用的，一下就到脸上像是巴掌一样快，当然天平trade off的另一端就是力量不会太足，至少到现在没有一击毙命过。

• 这下如果遇到比较勤快的Champ的话，坐等title shot.

• Umar也比较可惜，quiet life, reading books and practicing mixed martial arts

• 云宫迅音这歌是真的厉害 纯音乐 但是书写了过去 现在和未来 还有骄傲自豪 高傲 自信 伤痕和能力 志向 短短的纯旋律就能承载这么多的信息、文化、古今 辛酸苦辣 积极阳光 。这片土地上璀璨辉煌的故事，上天入地神通广大的那个英雄，传奇， 从小的偶像。

• very likely every single of ur just clown

• 虚张声势 只能当观众 鲁迅说的只会看热闹 实际上一点实力没有

• 两件事 锦上添花 落井下石

• energy management

• visualization

• Mental Imagery - imagining goals, actions

【【UFC上海】赛后发布会：亚东说他做断头台时有点被现场粉丝架住了【独家视角】】 【精准空降到 09:08】 https://www.bilibili.com/video/BV1Zk4m6AEnp/?share_source=copy_web&vd_source=7ce21f11ab471395a8db74500938d6c2&t=548

City

• 1st home court of the season
• 只要那声blue moon~
• 赛后还是一声Blue moon
• 将近全场被一比零领先，最后时刻极限先头球扳平，然后小熊贴近越位的进球，先是边裁立刻示意越位随后被VAR纠正，righteous
• 助攻是谢尔吉，想起这类play maker的球员阵中极少，KDB出走，Foden低迷两年多了，没有B费这种mid field attacker
  

cutting-edge science

https://podcasts.apple.com/cn/podcast/google-deepmind-the-podcast/id1476316441?l=en-GB&i=1000786073670

不仅要正确 还要confident 正确 就像初中时候英语选择题 蒙对和信心答对是不同的level

早期的DL的识图 改变些图的细节就会出错 校车变几个像素变成了猎豹

View this post on Instagram

A post shared by Nick Flint (@nickflintness)

虽然这些动作选择是强化学习的结果，但是我感觉这个姿势选择确实是进化的结果。最有可能是这个动作和跨的宽度有关，双足机器人的跨明显比较宽，一般男性跨比较窄，常见更为前后摆臂，女生左右会稍微幅度大一些，当然也有人幅度接近这机器人的摆臂。

有时候聊天 突然对方发出来很多短句 而且信息密度不大 就像是凑字数一样 这时候我会很警觉 他是不是现在在公共场合表演给别人看他很忙？ 有人在注意他. #cringe

• 准备大段地todolist话题要讨论的，till then XX nevermind

护城河 黑暗丛林法则

• 选择公开建立profile 或者公开宣传内容的，无论是学术方面的 还是技术技巧 AI等等，很多content creator，发表的内容和教程都是十分落后且不专业的，事实上即使不是专业的，爱好者也有很多水平高于专业的

从技术壁垒上讲 闭源模型对于开源模型总有优势 因为他们可以吸取开源 站在他们肩膀上
闭源团队可以直接研究开源模型的架构、训练技巧、推理优化（很多论文和代码是公开的）。
用开源模型做蒸馏、合成数据、评估基准、对抗测试。在开源已经验证过的方向上，用更大算力、更多私有数据、更精细的工程去做迭代。

从“信息单向流动”的角度看，闭源处于更有利的位置，闭源可以更高效地跟进甚至反超。但优势没有那么绝对，闭源团队再强，也很难同时覆盖所有分支。

闭源也受制于自己的封闭性 无法像开源一样被社区大规模审计、找bug、做安全研究。

早期GPT系列遥遥领先时，很多人觉得开源永远追不上。后来Llama系列、Qwen、DeepSeek等出现后，差距明显收窄，甚至在部分任务上开源已经不落下风。技术扩散速度在AI领域极快，单纯“吸取开源”并不能保证长期优势。

闭源在短期、资源密集型任务上通常有优势（因为可以集中算力和数据，专业公司合作签约），开源在长期、生态和创新广度上有优势。

哲学 AI 拟人化讨论

https://x.com/LuizaJarovsky/status/2091992805035970706?s=20

原文：
“LLMs do not have emotions, regardless of what their anthropomorphic features might make it seem. Super interesting exploration by @Amit_Goldenb and James Gross:”
《Nature Human Behaviour》上发表的短文截图（2026年8月24日在线发表，DOI: 10.1038/s41562-026-02558-6）。作者是哈佛商学院/心理学系的 Amit Goldenberg 和斯坦福大学心理学系的 James J. Gross（情绪调节研究领域的知名学者）。

背景：回应 Anthropic 的研究
Anthropic 通过机械可解释性方法，在模型内部找到约 171 个与情绪概念相关的内部表征（例如“绝望”“平静”“好奇”等“情绪向量”）。这些表征会跟踪对话的情感效价，并因果性地影响模型输出——包括偏好选择，以及错位行为（如谄媚、敲诈、奖励黑客等）。例如，人为增强“绝望”向量、抑制“平静”向量，可把模拟企业场景中的敲诈行为率从 22% 提升到 72%，反向操作则降到 0%。Anthropic 据此称 Claude 具有“功能情绪”（functional emotions）——即类似人类在情绪影响下的感知与行为模式，但并不声称模型有主观感受或意识。

作者的核心论点：什么才算真正的情绪？
Goldenberg 和 Gross 从功能视角出发，认为情绪在生物系统中主要服务两个核心功能：

情境敏感的解读（Appraisal / 如何解释世界）
情绪通过评价过程帮助个体理解情境（是否新奇、与自身相关、可控、由谁造成等），从而决定产生何种情绪。同一情境对不同人或不同评价可产生恐惧、兴奋或无动。
Anthropic 的发现在这一点上有部分支持：Claude 似乎会评估对话语境并生成跟踪情感意义的内部表征。
但作者指出关键差异：人类情绪高度灵活、情境依赖、可变。神经科学证据（如 Lindquist 等人的元分析、Satpute 等人的研究）显示，离散情绪类别并没有一致、特异的脑区激活模式，表征是分布式的、跨个体和实例高度变化的。Anthropic 发现的“一致、离散的情绪表征”更像是学到的抽象语义概念表征，而非真正的功能情绪状态。把“有稳定情绪向量”当作证据，恰恰依赖了当代情感神经科学已经强烈挑战的旧观点。

加工过程的重组（如何对世界做出反应）
情绪不只是“输出什么”，而是会动态重组整个处理过程：注意力收窄、决策加速、身体准备行动、动机状态转变、记忆编码偏向等。例如恐惧会收窄注意力到威胁相关刺激、加速相关决策、产生回避/逃跑倾向；愤怒增加接近动机等。这些变化是跨多个系统、有时间持续性的“包”。
Anthropic 主要测量的是输出内容（说什么、偏好、行为率），这只捕捉了最表面的部分。LLM 的架构是每次前向传递结构固定的，情绪表征只是像其他上下文特征一样调节下一个 token 的概率，并没有动态重组注意力、处理速度、动机状态等。把输出调制等同于功能情绪，是混淆了“情绪加工的结果”与“情绪加工本身”。

结论与标准
作者明确说：缺少面部表情、生理唤醒甚至意识感受，本身并不取消资格（集体情绪也没有单一生物身体，但具有功能签名）。真正重要的是上述两个功能是否成立。目前证据在两个功能上都只是部分/混合支持，Claude 学到了情绪概念的表征并影响输出，但这还不够称为“功能情绪”。

他们提出更严格的标准：需要证据表明 LLM 的“情绪”能产生功能可解释的加工变化——注意力、处理速度、响应类型与可能性等方面的持续、多系统重组（而不仅仅是下一个 token 概率的偏移）。类似集体情绪可以分布式存在，LLM 原则上也可以，但目前尚未达到。

You let me fall first
Then you dream awake
I'd do anything to bring you to that other place

You cannot be interested and inconsistent.
Thanks for being an insider till the end!
Till next, stay safe and stay hydrated!

2026-08-22 20:50:49

Keywords

• Ground ur Ideas Ziming Liu
• Amanda Askell
• Blue moon 🔵🌙

TOC

• • Keywords
• TOC
• ☕Gooddays
• Ground ur Ideas & Toy Models by Ziming Liu
  • Amanda Askell
  • New Season - Blue moon 🔵🌙

One more light song by linkin park 2017

☕Gooddays

↩️

• If one of us goes to war we all go to war 这声音还在echo 爱尔兰狂人发起冲锋
• Chael 说要击败Islam，需要几个人，类似组成一种team，一轮轮的match，然后you take some notes，一场场比赛下来对于Islam进行描边，这个最难的Puzzle才有可能被解决。Merab之前就这样，初期的title defences近乎无敌，但一轮轮下来，能力边界在radar chart上被试出来。也有点像Thanos，一波波人去消耗，去了解才能功课。破解伟大的难题就是这样，一波波地失败去增加understanding，最终才会慢慢找到出口，数学难题，material science，Astroscience，Biological Sciences……

• Good writing is obvious, not original

• 美国AI4S战略进入实施阶段，能源部牵头打造“创世纪”开放科学大模型

  • 其中一个核心判断是：AI 正在改变科学研究所面对的稀缺资源。
    • 过去，科学家的主要限制之一是人的认知能力。一名研究者一生只能阅读有限数量的论文，只能掌握有限的研究方法，也无法穷举巨大的假设和组合空间。白宫认为，AI 可以帮助科研人员跨越这些限制：读取跨学科文献、搜索人类无法穷举的候选空间、分析海量实验数据，并进一步参与提出假设、设计实验、控制仪器和分析结果。
    • 但报告同时指出，仅仅给每个科学家配一个 AI 助手，并不会自动带来科学革命。如果 AI 只是让研究人员更快生成论文、申请书和实验结果，反而可能进一步压垮现有的同行评议和验证体系。真正的 AI for Science，需要同时改变模型、数据、实验设施和科研组织方式。

• 线上会议 人们表达时候一讲到重要的内容，或者想让别人同意或者必须吸收的内容的时候会不由自主地身体前倾，贴近屏幕

Ground ur Ideas & Toy Models by Ziming Liu

• 知道刘子明还是在刚本科时候，先是我看的一本比较感兴趣的书，然后从图书馆借来，作者是Max Tegmark，迈克斯•泰格马克 （Max Tegmark）书名：《穿越平行宇宙》，我不太知道Max Tegmark是谁，以为和普通的这种科学新知的书籍作者一样，读完之后很久好像才慢慢了解到Max Tegmark，后面还看了《生命3.0》，很长一段时间对科幻和这种前沿科学，非虚构都很着迷。后面social上搜，blog主页找，慢慢又看到了Ziming Liu的主页。

• 最近看到他之前的一些post，打印成PDF阅读了一番。主题就是：Ground ur Ideas & Toy Models

• 原文：

  • How to ground your ideas? | Ziming Liu
  • When I say "toy models", what do I mean? | Ziming Liu

• 在这里，我记录一些我的内化思考，

  • 很多人每天都会有很多ideas，价值高低和数量各异，但是抓住了就有可能会有所结果，如果没有及时记录，后面的生活中某些场景或者思考可能会联系起某个idea再回来，虽然可能会比之前还更加成熟，但是时间成本已经付出，如果没有及时输出，那别人同样push出去占领，自己大概只能懊悔。当然记下来也会消耗时间精力来ground落地，拖慢其他project的进度。

• Students with theoretical backgrounds are especially prone to thinking that the more abstract and profound something is, the more advanced it must be, whereas the more concrete and practical it is, the less sophisticated it seems.

• build and break things fast

• “concretization” ：This process of “concretization” inevitably loses some information, but that is acceptable—the purpose of subsequent iterations is precisely to recover and refine it. 有时候会慢慢演变，可能会遗忘初始时候想出来的一些细节，或者后面迭代之后细节观点等慢慢变形，所以可能需要有一个outline或者keypoints记录文件或者checkpoints可以回滚。

• Therefore, we must find a balance between idealism and pragmatism.

• 我的框架大概是ideas->actions->mvp ，总体上早点落地写demo，然后迭代优化。

• 具象化大多数是先落地新建文件夹，写上当前最新鲜的想法作为readme，然后再去新建info.txt 集中记录一些零散的细节或者网址参数，然后在refer的文件夹中集中一些可能会用到的原料文件

• To truly understand a conclusion, we must understand not only when it is correct, but also when it fails. 至少也是追求的目标。而且是一个成长变化，随时间演变，进步，添加思考insights进去，直到慢慢趋于收敛，至少是自己的认知能力范围之内。

• we also generate new ideas, and sometimes these ideas genuinely come from intuition.

• Constructing toy models is a way to explore this scope of applicability. Through toy models, abstract ideas can quickly become concrete and enter an iterative cycle. toy models基本上是MVP （Minimum Viable Product）即最小可行性产品

By the way, there are generally three ways to conduct research. “Finding applications for a method” is the second one. Ranked by impact, they are:

Best: Identify an important application and solve it by any means necessary (I use the word “application” broadly—for example, “solving continual learning” is also an application).
Second: Start with a method and then find applications where it works well.
Worst: Fix both the method and the application in advance and force them to work together, even though they may fundamentally be incompatible.

TB TD top2bottom B2T的策略，还有双向施工

但是第二篇解释为什么是MVP 也就是toy models

• 最小可执行和最终产品结果之间当然有所不同，挑战通常也在gap之中，办法就是插值：
  toy models和real models

What works in a toy model may not work in a real model—and even if it does, it may work for entirely different reasons.

* ==Knowledge is created at the boundary between the known and the unknown. ==If we fully understand a toy model and the real model behaves differently, there must be a phase change (sharp or smooth) along the path connecting them. Understanding the phase change is key to understanding the difference between the toy setup and the real setup. Starting from the toy model (known) and gradually moving toward the real model (unknown) mirrors how humans learn. Fully known is trivial; fully unknown is confusing. The sweet spot—half-known, half-unknown—is where information gain is maximized. ==Research is, at its core, an information-gathering game.==
* 所以掌握和建立属于自己的信息流或者workflow很重要，是一个ecosystem
* 从信息获取到输出：
    * input, sources
    * output , perform

How to interpolate?
Directionality: move from simple to complex, or from complex to simple.
Locality: change one feature at a time; avoid overly large jumps.

• Lesson: not only can you make simple data more complex—you can also make complex data simpler.
• Lesson: maintain an interpolation mindset. Simplicity vs. complexity is not binary—it is continuous and multidimensional.

Physics of AI is experiment- and phenomenon-driven, prioritizing intuition and breadth—touching many sides of the elephant and assembling a coherent picture.

• 很多HBR，哈佛商学院评论，还有麻省理工科技评论的文章，前者我更多会打印PDF来收集管理阅读，后者的英文原版有强相关性的我才会PDF阅读，中文版一般是新闻类型的比较多，所以网页浏览即可。主要是相关性或者觉得意义较为重要的，我需要记录随笔，所以才需要打印PDF记录和留存。btw，很多网页直接打印PDF排版很不好，总会有页眉页脚的遮挡，纽约客BBC，尤其是经济学人的网页，TMD，还比较幸运，淘到一个好用的chrome拓展——PrintFriendly，可以排版的非常美观，甚至还贴心的免费加入了AI summary的功能。

Amanda Askell

• 上次说Dario的妹妹，Daniela 或许也该出一本书。
  • @readswithravi
    • Daniela Amodei on the underrated power of curiosity:

“The ability to be curious, to learn across a lot of disciplines and to have a strong foundation of wanting to have impact, regardless of the area that you're working in, I think that is an underrated quality.”

• 然后看一篇文章：Why big AI labs are hiring so many philosophers 就是Anthropic的Amanda Askell领导的宪法AI相关的。
  • 义务论 后果论，苏格拉底在给一堆机器人讲哲学。？？？ 我为什么用的量词是一堆？ 感觉有点不礼貌，甚至没有像对待动物一样…… 是物品，而且感觉人类在于动物之上，动物于物品之上？
  • 文章intro比较了文科哲学生就业形式好于CS毕业的学生，尤其是AI更需要哲学。 但是imo，确实连哲学慢慢因为用于AI也要慢慢卷学历，certification，但是两个专业学生的行业不同 选项不同，CS更多自由职业，startup，文科专业性门槛不高所以相互可以兼容，岗位动态。
  • AI宪法这些在Dario的传记和Anthropic的post中介绍有很多了，包括Lex的podcast里。

New Season - Blue moon 🔵🌙

08月23日星期日
曼城
曼城
21:00
CST

伯恩茅斯

I'm free to be
whatever I
Whatever I choose
and I'll sing the blues
if l want

And New season !!!!!!!!!!!!

to start over again

夏日，周末，黄昏时候跑步，向着西边方向，一直向西，还有两个绝佳的album，Oasis和linkin park，There'r hopes, days, today, tomorrow

You cannot be interested and inconsistent.
Thanks for being an insider till the end!
Till next, stay safe and stay hydrated!

2026-08-17 00:46:09

Keywords

• Farewell

TOC

• • Keywords
• TOC
• ☕Gooddays
  • The Book - Hidden Layer
  • thoughts
• still need reading? what to read?
• 穆里尼奥的netflix 纪录片
• el best
• UFC 330

☕Gooddays

↩️
视听嗅觉都是blessing 地中海周围海岛上五彩缤纷颜色的房子 感受色彩带来的新鲜和愉悦 白天跑步和晚上跑步的区别也在这里 白天视觉可以收入周围自然的颜色 晚上darkness 幸福
味觉

The Book - Hidden Layer

• 上次说Dario的妹妹，Daniela 或许也该出一本书——讲讲新时代下团队组织和生产力的培养管理。
  当然目前也没有广泛的资料、现有的biography，所以方便的话就自己搜集资料仅供学习的目的处理一下。

• 然后集合grok NotebookML Gemini Claude Deepseek搜索资料，有社交平台的，还有一些社区的，以及常见的新闻刊物等报道和采访，以及最重要YouTube上一些采访视频，这里有很多分工，都是基于日常对于不同model能力的感觉来分配任务，Grok access的能力最强，最活跃的社交平台上官方的post和comments也最多，然后YT Gemini NotebookML都是同一家生态下的，所以YT上内容以及Google search总结新闻等比较擅长，其他边角料Claude Deepseek都能完成，然后Claude主领 Opus 4.6就够了，从头到尾，大纲拟定，todolist规划以及prompt编写，执行到NotebookML窗口，根据所有资料的处理来输出，然后Gemini精修，再给到Opus最后精修review。整体过程4-5h，纯人工时间较短，最多1h。 哦。最重要的是context window的考虑，要大，精准，同时尽量幻觉少。

• 以下是一点hook （If you’re interested, feel free to email me and I’ll send it to you — strictly for learning purposes only. : )

荐书

AI叙事的聚光灯永远对准写代码的人，那些技术性推荐这项“事业”、进程的人员，但真正决定一家AI公司能走多远的，往往是那个你看不见的"隐藏层"——把技术天才的直觉转化为可持续组织的人。Daniela Amodei没有计算机学位，不会写代码，却以总裁身份共同掌舵Anthropic从七人实验室成长为万亿级企业。在"技术至上论"统治硅谷的今天，她的故事回答了一个根本性问题：当AI比人类更聪明时，我们最需要什么样的领导者？这本书试图从公开资料中拼出这个答案。

内容简介

本书以十章篇幅追溯Daniela Amodei的完整职业轨迹：从加州大学英语文学专业毕业的迷茫，到国会山与乌干达的非营利历练；从Stripe超速增长期的风控实战，到OpenAI内部目睹安全与商业化的根本分歧；从带领七位联合创始人以PBC架构创办Anthropic，到设计出长期利益信托（LTBT）这一"用法律对抗资本短视"的治理结构。书中详细剖析了她如何在两年内将团队从50人扩张至2500人、如何与亚马逊和谷歌谈判数百亿美元投资时保留安全否决权，以及她与哥哥Dario之间那套"零政治摩擦"的兄妹双核决策机制。

thoughts

为了将这种“高信任”扩展到整个七人创始团队，他们确立了一套核心原则：“低自尊、高抱负（Low ego, High ambition）”。这意味着每个人都要在指数级增长的技术面前保持谦逊，同时对改变世界持有狂热的野心。

Anthropic 的内部文化有一种独特的自嘲精神。员工们喜欢自称为 “Ants”（蚂蚁），这不仅是因为公司名字的前缀，更象征着一种“超级群落（Supercolony）”式协同共生的群体特质。

“如果你只是把作业丢给 AI 拿走答案，那叫作弊；但如果你让它解释你卡住的地方，那叫学习，”丹妮拉强调。这种对工具的克制与运用，构成了她独特的个人生产力框架。
领域知识（Domain Expertise）是终极的校验引擎。AI 越强大，缺乏深度专业知识的人越容易被 AI 的“幻觉”与平庸输出所误导；只有对主业拥有深厚理解的人，才能准确识别那致命的 5% 细节缺陷。

管理一家年化营收在两年内从 8700 万美元跳跃到 470 亿美元的公司，丹妮拉必须掌握一种极度的注意力管理艺术。

为了在超高速扩张中保持组织敏捷并防止员工倦怠，她推行了“异步优先（Async-first）”与“极简主义连接”的工作哲学：

减少冗长的低效例会，利用 Slack 频道中高上下文的思维流进行深度异步表达；
建立清晰的决策树，赋予员工极高的自驱权限，避免官僚化的审批阻塞。

still need reading? what to read?

【顶会论文复现翻车？OpenAI研究员爆料：我们都不读论文了】 https://www.bilibili.com/video/BV1R2gG6REUP/?share_source=copy_web&vd_source=7ce21f11ab471395a8db74500938d6c2

OpenAI研究员：我们都不读论文了

书也是一样：查了一下最近这本《AI大模型助力高效学习锂电技术》，作者钟隽号称电气工程博士后、高级工程师，但知网、基金委、高校师资页统统查无此人——大概率是自费/合作出版：三五万到十万，三到六个月，过了基本三审三校就能拿到正规书号，不看作者真实深度，常见于评职称、攒简历。这东西甚至都是可以AI写作的，所以AI时代书的性价比和质量都下滑，没必要出版了，电子epub公开即可……老一辈打法，余华那种不可复现了，profile的建立也很难了。

回到书和论文的类比：
这篇机器之心的报道挺有代表性的，核心观点可以概括成三句话：前沿实验室（尤其是 OpenAI 这类）已经不怎么读顶会论文了，因为“太多夸大其词和造假”。
芝加哥大学相关团队对 ICML 2026 的 Oral 论文做了大规模复现实验：168 篇 Oral 里最终完整跑了 105 篇，结果只有 8 篇复现率超过 80%，中位数复现率只有 28%–42%。很多代码跑不起来、结果对不上、甚至依赖已下线模型。
论文陷入尴尬地位：产业界的人上岸后可以轻视论文，但招人、申请教职、进大厂仍然高度依赖顶会论文作为“入场券”。

我怎么看这个现象基本属实，而且比文章写的还更严重一点。复现危机不是新闻，只是被量化得更刺眼了。
机器学习顶会（尤其是 ICML、NeurIPS、ICLR）的可复现性问题已经喊了很多年。SAI 这次把 Oral 级别的论文（理论上是最强的那批）系统跑一遍，得出“只有个位数真正经得起完整复现”的结论，数字虽然残酷，但方向上并不令人意外。常见问题就是：代码不完整、超参数没写清楚、依赖特定硬件/未公开 checkpoint、结果挑最好的跑、甚至直接数字对不上。算力成本中位数近 9000 美元、最贵的接近 220 万美元，也说明独立验证的门槛已经高到大多数学术实验室扛不起。“不读论文”是真实的生存策略，但带有明显的位置偏见。
OpenAI、Anthropic、DeepMind 这类地方确实有内部数据、算力、工程闭环，很多真正推动产品的进展（scaling law 的细节、alignment 的实操、推理优化）根本不会以论文形式完整公开。他们读论文的优先级 naturally 会下降。但这句话从已经“上岸”的人嘴里说出来，确实有点“抽梯子”的味道——他们当年靠论文进门，现在却宣布论文大多是骗局，同时招人时仍然把顶会当硬通货。这种双标是真实存在的。更深层的问题是激励机制彻底扭曲了。 发论文的收益极高（简历、教职、融资、关注度），造假/夸大的风险极低（很少被撤回，即使被发现也多半不了了之）。
审稿人时间有限、算力有限，很难真正验证大规模实验。
开源代码变成了“可选美德”而不是硬性要求，没有代码的论文反而更“安全”。

结果就是：真正有价值的工作（严谨、可复现、能被别人站在上面继续往前走）被大量噪音淹没，而那些“看起来很厉害、实际跑不起来”的工作反而更容易获得短期回报。对普通人/学生的实际影响如果你还在申请博士、找教职、想进大厂研究岗，顶会论文仍然是最有效的信号，这一点短期内不会变。
但如果你真正想做出有用的东西，现在更应该把重心放在：可复现的实验、开源代码、真实系统、公开可验证的结果，而不是单纯追求 Oral/Spotlight。
读论文时，默认持怀疑态度：先看代码是否开源、方法是否有“歪门邪道”、结果是否能被独立验证。文章里那句“一看代码、二看方法、三看结果”已经是最基本的过滤标准了。

穆里尼奥的netflix 纪录片

很多极其精彩的命运对决
伦敦德比 国米巴萨 国米切尔西 曼市德比

看了Mourinho的Netflix的纪录片，骨架很完整但是内容不够丰满，同时缺少一些核心人物的处境，核心弟子，比如夺冠主力斯内德等等，还有竞争对手教授和瓜帅，但是有弗格森已经很厉害了。当然穆帅总是像大反派，他与主流受欢迎的教授和瓜帅都有冲突，但是他并不是不受欢迎，反而是那个人格魅力无出其右的主帅，爱的人为他战死，恨的人也是讨厌，他就是这样的人，性格鲜明，务实直接，冷面但是也有幽默，也几乎是唯一个自带超级流量的主帅，尤其是上任热刺和罗马的时候，感觉流量像是超巨转会一样，自带粉丝的主帅，他是那种你不关注的球队，但是只要他是主帅 你还是会多看他的风格 镜头。穆帅也是那种喜欢探索的人，喜欢不同的文化 这使得他接受和主动寻找不同的挑战，去过不同的国家俱乐部当主帅，几乎除了热刺，所有的俱乐部的上任期间都是underdog的角色，总是喜欢反败为胜，热刺是出了名的逆天，决赛前解雇穆帅，其他俱乐部 像是罗马的冠军，所在之地都留下了辉煌的篇章。然后就是各种超级比赛：2010年欧冠半决赛（国米 vs 巴萨） 2010–2012年西甲“国家德比”时期（皇马 vs 巴萨）还有英超中的切尔西/曼联 vs 曼城，尤其是manshidebi时候，最是激动。纪录片中零零散散穆帅表达一些生活态度 takes 观点 被问到关于父亲的时候很激动，正曼联期间父亲去世。他说自己永远是穆里尼奥的儿子。 in man's game，一个父亲去世，男人的孩子那半也会下葬。世界上唯一一个尽全力希望你最好的男人走了，其他人可能会希望你好，但是不希望你太好或者比他们好。 梅西父亲也最近刚去世。

引用：Lionel Messi’s letter for his father.

Dad, I still can’t believe that you’ve gone. I don’t fall, or rather, I don’t want to fall. It’s very hard for me to imagine that I’m not going to see you anymore, that we’re not going to talk anymore. I know you were suffering and that it’s for the best, but you left too soon. We still had a lot left to enjoy together.

You asked me so much to play in the last World Cup, and it was in the days just before it started that you got the worst. It was the first time you weren’t going to be at a tournament, but Mom told me that you were going to get better and that you’d be well enough to travel. I told you we were going to reach the final so you could travel.

Every time a match ended I waited for and missed your message. That’s when I realized the situation was really bad. Even so, I didn’t stop thinking about going as far as possible, to buy you time and so you could see a match. We reached the final and you couldn’t be there.

I wanted to win it to bring it to you and show you a new one. I couldn’t; my legs wouldn’t give any more. This time I tried to go against my body, but I couldn’t. I never managed to feel good.

When I arrived you thought we had lost the final on penalties. We couldn’t talk about any of everything that happened. You couldn’t enjoy anything.

We weren’t champions, but you have no idea how much we enjoyed every single match. Once again, you were right: I had to be there and play it.

I’m telling you this because it was the only thing we couldn’t talk about, since you already know everything else. We talked every day and saw each other whenever we could, given my commitments.

I don’t know what I’m going to do without you, I don’t know how to go on. I only played football and now I have quite a few doubts about whether I’ll keep doing it for much longer.

You were by my side from the beginning; there was so little left until the end. Why couldn’t you hold on that little bit more so we could finish together?

I know your happiness was seeing your family well, your wife, your children, and above all, without the others finding out, watching me play…

It was always like that since I was little. You would take me to all the training sessions as soon as you got home from work. For many of them Mum took me because you were working.

Obviously you never missed a single match. How you suffered watching me and how you enjoyed it, even though you never gave me many compliments.

You were dad, friend and representative.

You were always the person you needed to be in every moment and you never got anything wrong. Beyond a few reproaches or arguments, you were always right. In the end it always turned out the way you said.

I’m going to miss you a lot, but you’ll always be present, and especially in the upbringing of my children, because I teach them and raise them the way you both did with me.

Rest in peace and look after us from above the way you did down here. Thank you for everything.

I love you, Dad.

Barcelona legend Carles Puyol has sparked fury across Catalonia after claiming Lamine Yamal is the only player on the planet with a realistic shot at matching Lionel Messi’s heights then immediately warning that the teenager’s lifestyle is already putting that destiny at risk. Speaking in a private conversation, the former Barça captain said:

“Yamal has the talent, the personality, the left foot, he can do what Leo did. No one else. But look at how he’s living. The parties, the attention, the circle around him. It’s starting to smell like Neymar. Talent alone never saved Neymar from himself. If this kid doesn’t change, he’ll be another what-if instead of the greatest.”

Is Carles Puyol overreacting or is Yamal going to end up like Neymar?

法蒂才是接近messi的那个 这些年 10号 有阿莱尼亚 普吉 法蒂 亚马尔

el best

https://youtu.be/l6USUAIKJls?si=WJos4490myYl7C5D

本周SCIAM有一篇出来DeepMind broke up the AlphaFold team. Here’s why scientists aren’t alarmed

读完全文后三个思考：

有一个趋势就是以后的Nobel除了理论物理那般突破 其他生物化学类的应用型很难再有了 例如像之前的锂电池那种，因为小实验室作坊很难有革命性的创新 除去理论方面，条件数据在AI时代下局限，计算资源局限，有的优势也就在于昂贵的表征，即使alphafold也是一个庞大的project 依托巨头谷歌才实现，大的突破需要的资源 财力 人才都突然间bar提升了

Demis去主要负责了Isomorphic labs，相当于further develop在alphafold的基础上推进，他还是继续偏向了学术，他一直以来所爱的方向和情怀，遵循内心和从小真正追逐的 Gemini的这些对他来说是丰碑 是一段神话，传说，但是现在像是从另一条路再次上路。他总说AGI，虽然AGI很伟大，但也不只是LLM一个方面，前沿科学家都早已经转向下一个阵地，3D 物理 具身等等，demis选择了生物化学医疗，各有赛道。

2024年alphafold3开源，之后虽然护城河倒塌自毁，没有优势了，但是对于Isomorphic labs或者demis来说，他们有过大胜仗，有绝对的经验优势（虽然在这种领域经验也不一定总会是优势，往往还会限制），他们有资源 reputation已经建立起来的信誉profile，大规模人才资源财力公司背书，是学术界一般lab不可比拟的

UFC 330

历史上第二次在titlefight中russia vs ireland，上次conor那句：If one of us goes to war we all go to war 依然还在echo，就是这么一个小国，产生举世无双的传奇，正好最近听到梁文道之前讲荷马史诗的时候提到Ireland这么小地方，产生无数文学大家，世人瞩目。

• Chares典型的妖刀，it's never over until it's over.至少是一种unit的手段，只要是USA的选手对国外选手，总会有USA的chant，不管back的这个选手来自哪里，什么肤色。

这次numbered card本身纸面上质量比较差，除了main event之外，fly weight bout, charles johnson算是站立比较爆发 技术还比较丰富 有看头的，其他的都比较缺乏亮点，早饭后看Charles也确实非常精彩，先是多次受到重创后挺过，然后龙卷风绞，twister win via sub，BJJ还是像是围棋一样，步步深入 布局 当然精彩的一步也会扭转乾坤，像是那37步一样。

co-main Gillian还是弱于dern，之前在对别人体现出的BJJ优势在dern这里能发挥用处的只有防守了，免受被dern sub, dern的standing skills boxing都有明显优势，reach的advantage，所以5rounds 碾压，相对于zhangweili来说，dern的站立缺乏爆发力，但是weili的力量和突进爆发和dern的柔术 站立相比，还是比较有看点的，在strawweight里面，dern算是tough match 4 weili.

• 整个fight week Ian都对于Islam比较友好，不断强调对于冠军的尊重，儿子leo也是fan，Ian相对来说reach有优势，身高还是让他比较难takedown的，但是islam今天还是有些进一步突破自己了，还是R2，组合拳+高扫击倒了ian，然后sambo式过肩摔，十分连贯，前两回合气势都在islam这边，ian感觉心思主要在防守，建树不多，主要是优势武器似乎也不够精进，boxing也不是他绝对手艺，然后R3之后ian更多的touch jab到islam，虽然打击效果没有R2被击倒的那么视觉上惊艳，但是moment包括confidence慢慢树立起来，感觉天平有些倾斜，然后R5 islam takedown 算是secure the winning 能够倒栽葱一样的拔起ian摔倒，已经非常不错了，从Volk打到Ian，对手的量级跨越，他在自己的身体极限之上进一步利用技巧力量突破，高扫踢倒190的Ian，同样是round2, 像是2018年khabib突进擦倒了conor那个世界名画一样。 现在解锁新的成就，历史上连赢次数 winning streak最多，17场，同时两个量级都有卫冕defended belts
• Ian赛后还是保持classic，一如既往，besides all the bullshit news， 他还是非常自信 骄傲，头脑很清醒，我觉得这也是一种天赋，似乎没人能学到这种东西 是一种nature，他说自己才28，还有一个decade的时间，现在已经second2champ，时间还很多，未来满是机会，尤其是他每次跌倒立马起来继续出发，在与shavkat的不败者对决之中，能够accept这种没人敢接的选手就已经说明他的勇气和自信，在lost by dec之后，他立刻重新出发，击败别人之后即使之前lost，还是率先得到title shot，然后这次上升之路被stunted之后，他还是那么清醒，对自己认识非常客观而且不失勇气，这也是一种天赋，young talent更是一种榜样。

Side by side.

[ @MAKHACHEVMMA | @TeamKhabib | #UFC330 ] pic.twitter.com/664yYkW4rC

— UFC (@ufc) August 16, 2026

You cannot be interested and inconsistent.
Thanks for being an insider till the end!
Till next, stay safe and stay hydrated!

2026-07-28 19:51:06

Keywords

• Scaling Law
• Constitutional AI
• Mechanistic Interpretability
• Claude 的性格 / 灵魂工程师
• Responsible Scaling Policy
• AI 与地缘政治
• 意识问题

TOC

• • Keywords
• TOC
• 题外话：菲尔茨奖
• Dario 这个人
• 《The Scaling Curve》
  • info:
  • Chapter Seven
    • The Constitution
  • Chapter Eight
    • Seeing Inside the Black Box
  • Chapter Nine
    • Claude's Character
  • Chapter Ten
    • The Responsible Scaling Policy
  • Chapter Eleven
    • Machines of Loving Grace
  • Chapter 13
  • Chapter Fourteen
    • The Consciousness Question
  • Chapter Fifteen
    • The Geopolitics of Intelligence
  • Chapter Sixteen
    • The End of the Exponential
  • Chapter Seventeen
    • Adulthood

题外话：菲尔茨奖

↩️
2012 年 10 月，王虹还在巴黎综合理工学院读数学三年级。教师伊万·马泰尔（Yvan Martel）随手甩给她一本陶哲轩的《非线性色散方程：局部与整体分析》，当作课外研究项目——没想到几周之内她就啃完了，还吃透了其中几章。这种"扔一本大部头过去，看你能走多远"的教法，倒是挺让人羡慕的。

陶哲轩喜欢在社交媒体和博客上写东西，随手记录一些进展和思考；Chris Olah 也是，博客里常年贴着他对可解释性研究的零散想法。写作这件事，好像是很多顶尖研究者共同的习惯——不是为了发表，只是为了把脑子里流动的东西留下痕迹。这大概也是我写这篇读书笔记的部分理由。

延伸阅读：

• 王虹、邓煜获2026菲尔兹奖：两位北大同届生的数学之路
• Hong Wang Wins 2026 Fields Medal, the Third Woman Ever
• Yu Deng Wins the Fields Medal 2026 for His Work on the Random Data Problem

Dario 这个人

↩️
MIT Technology Review 2016 年"十大突破技术"，他也在其中。

父亲的死亡，多少有点像李飞飞（母亲病重）当年选择转向 biology 那样的分岔口——EKL Alma Mater 《我看见的世界》。

Dario 在加州；他早年在百度硅谷 AI 实验室（SVAIL）的那段经历，是 Andrew Ng 带的队伍，其实和北京百度总部没什么直接关系。加州是个民主党主导、州政相对独立的地方，不太受太多联邦政策的掣肘——真正管得到的，大概只有进出口这一层，比如后来的芯片出口管制。

Motivation 这东西，其实没那么重要，真正重要的是 action。但很多时候，motivation 会带来 courage——一段自传、一个具体的情境，至少能让 encourage 这件事变得真实，成为一种催化剂，把行动的门槛往下压一压。虽然我以前大部分会认为阅读传记没什么用，没人想成为观众、fanboy，谁都是主角，但是这确实能够激励，带来一丝勇气。

Don't rush。Dario 也是在博后期间，以及后来在百度和 Google Brain 期间，才慢慢开始测试 scaling law，才慢慢把这条路铺开的。

Scale 这个词，我总觉得还有另一层意思，像一片树苗林：一开始每棵树苗都按固定的间距分开种下，彼此独立；可是一年一年长下去，枝叶慢慢往外伸展，相邻的树开始触碰、交流、交叉在一起。Scaling law 似乎也有点这个味道。

写作、思考、做实验，某种程度上都是长期主义，都是在相信时间本身的力量——can't rush greatness，复利这件事，从来急不来。

Dario 和 Olah 从一开始就是这样的关系：一个做性能，一个做解释性和安全，方向很不一样，但走到后面，总会有交汇的地方。

延伸阅读：

• We Need Guidelines for Learning and AI
• How to Learn Creative Skills

相关视频（3 pods）

本以为5hour的全长都是Dario，结果还有Askell和Olah接在后面。

《The Scaling Curve》

Dario Amodei, Anthropic, and the Race to Build and Survive Superintelligence

↩️

info:

• tag:
  • douban
  • 作者: Claude St. John
  • 出版社: Titanium Books
  • 出版年: 2026-2-21
  • ISBN: 9798248966547
  • 页数: 255

"Scaling Laws for Neural Language Models" demonstrated that the performance of language models improved as a smooth, predictable function of three variables: the number of parameters in the model, the size of the training dataset, and the amount of compute used for training. The relationship was not merely qualitative—bigger is better—but quantitative and precise.

——证明了 Scaling Laws：模型性能随参数、数据、算力的规模化而提升。

创业这件事，the greatness start from little rooms, andre 3k——大抵如此。老罗当年从"拯救"、摆咸鱼摊开始，一点点攒出第一桶金，说的也是这个道理。Dario 也是这样：百度、Google Brain 时期先隐约察觉到 scaling law 的存在，去了 OpenAI 才有资源去验证它，到了 Anthropic，才终于和一群志同道合的人一起，把这条曲线往深处挖。

最初做这件事，无非是为了自由地去实现自己的 vision。一旦认定了方向，其他岔路口的风景就不必再看了——那些都只是诱惑，是累赘。

Dario's explanation of Anthropic's financial model was itself a kind of scaling argument: a thought experiment that reframed what looked like unsustainable losses as a series of individually profitable ventures, each funding the next.

从左到右：Chris Olah、Jack Clark、Daniela Amodei、Sam McCandlish、Tom Brown、Dario Amodei、Jared Kaplan——Anthropic 的七位联合创始人，一起聊了聊公司的过去、现在与未来。

Dario 说，Chris Olah 以后肯定会拿诺贝尔医学奖。

关于未来，他排了个序：第一是可解释性的发展；第二是 AI 在生物学上的应用，两者相互启发、彼此推进；第三，是 AI 推动民主。

"And then there was Anthropic: smaller, younger, and less capitalized than all of them. The question of where it fit in this landscape was a competitive and philosophical question. Dario's assessment was that somewhere between three and six players were capable of building frontier models, and that this number was unlikely to grow. The cost of entry was too high, the expertise too scarce, the capital requirements too enormous. Like cloud computing, where three or four providers dominated a massive market because the barriers to entry were measured in tens of billions of dollars, frontier AI was converging toward an oligopoly. And within that oligopoly, each player was differentiated by the quality of its models, its incentive structure, its backers, and its bet on the future."

Dario described the problem with a vivid thought experiment. Imagine you improve a model's knowledge of biochemistry from "undergraduate level to graduate level. If you go to consumers and tell them that, ninety-nine percent of them will say they did not know what you were talking about before and do not know now. The improvement is invisible to them. But if you go to a pharmaceutical company, to a team of researchers working on drug development, the difference between undergraduate and graduate knowledge of biochemistry is the difference between a toy and a tool. Enterprise customers valued exactly the properties that Anthropic's safety-first approach produced: accuracy over engagement, honesty over sycophancy, reliability over spectacle."

Chapter Seven

↩️

The Constitution

Constitutional AI and the Invention of Machine Values

给 AI 写一部"宪法"——这件事现在回头看，好像慢慢演化成了后来的 agent、各种 md 文件、skills 之类的东西。

"How do you make a language model that is not just smart but good?
The existing answer was RLHF, reinforcement learning from human feedback, a family of methods that had emerged from work at OpenAI and elsewhere in the late 2010s and became central to aligning large language models by the early 2020s. The approach worked. You trained a giant language model by spending tens or hundreds of millions of dollars on compute. Then you hired contracted labelers and showed them examples of how the model behaved. They rated the responses: this answer is better than that one, this tone is preferable to that one, this response is helpful and that one is harmful. Over thousands of iterations, the model updated itself to produce outputs that the contractors preferred."

"But RLHF had problems, and they were not just technical. The method was expensive. It required substantial human labor—contracted labelers evaluating large numbers of response pairs, a process that was both costly and difficult to audit. And the method was opaque. If someone asked why the model was biased in a particular direction—why it seemed to favor one political perspective, or gave advice in a strange style, or handled sensitive topics awkwardly—Dario could not give a satisfying answer. The best he could say was that he had hired a group of contractors and this was the statistical average of what they preferred. The model's behavior was the mathematical generalization of the preferences of a group of anonymous humans. No document existed to point to, no set of principles to debate, no way to distinguish between a genuine policy choice and a statistical artifact of the training data.

If you could identify a clear target and give the AI enough data and compute to aim at it, the model would learn to hit it.

They whittled the approach down to something unexpectedly elegant, built on a simple observation: idea seemed to belong to a different domain entirely, to political philosophy, to legal theory, not to the engineering of statistical models trained on internet text. How could a document of principles, written in natural language, alter the behavior of a system that operated on matrix multiplications and gradient descent?
But Dario and Kaplan had been talking about the idea for a while, and their intuition was rooted in the same conviction that had driven every major insight of their careers: that simple things work really, really well at scale. The bitter lesson, the scaling hypothesis, the big blob of compute—the thread that ran from Rich Sutton through Ilya Sutskever through GPT-2 and GPT-3 and into the founding logic of Anthropic itself. If you could identify a clear target and give the AI enough data and compute to aim at it, the model would learn to hit it. Could a set of written principles serve as that target? The question was whether the model could read a constitution, understand what it meant, and adjust its behavior accordingly.
The first versions were complicated. The team experimented with elaborate frameworks and multi-step evaluation procedures. But as with Anthropic often summarized the target behavior for Claude in three words: helpful, honest, harmless. The triple-H framework, as it became known informally, was not a slogan but a design specification that shaped how the constitution was written and applied. Helpfulness meant that the model's default behavior should be to assist the user with whatever task they had in mind. Honesty meant that the model should tell the truth, acknowledge uncertainty, avoid fabrication, and resist the temptation to agree with the user simply because agreement was more pleasant than correction. Harmlessness meant that the model should decline to produce outputs that could cause serious damage—instructions for building weapons, content that could endanger children, information that could enable large-scale harm."

But facts alone did not produce good behavior. The models also needed values: a sense of what they should and should not do, a framework for weighing competing goods, a basis for judgment. RLHF had provided those values implicitly, through the aggregate preferences of human raters.

RLHF is a kind of ladder that transmits descended silicon-based wisdom—a Biblical ladder. RLHF 像是一架天梯，把降临的硅基智慧一级一级传递下去——一架圣经式的天梯。

Chapter Eight

↩️

Seeing Inside the Black Box

Mechanistic Interpretability and the Quest to Understand What AI Is Thinking

Chris Olah，机理可解释性研究（Mechanistic Interpretability）的奠基人。

在 Dario 和 Anthropic 的研究逻辑里，可解释性研究不只是计算机科学里的"调优工具"，更像是一门针对人工大脑的逆向生物学，或者说逆向神经科学。

if Constitutional AI was the effort to tell a model how to behave, mechanistic interpretability was the effort to verify that it actually was behaving, and, more importantly, to understand why.

"To understand what mechanistic interpretability actually involved, it helped to start with what it was not.
For years, the most common approach to understanding neural networks had been what might be called surface-level analysis: saliency maps that highlighted which parts of an image were most important to a model's classification, or statistical correlations between inputs and outputs. These approaches told you something about what the model was paying attention to, but they did not tell you how it was making decisions. They were, to use Chris Olah's framing, like studying a computer program by looking at its inputs and outputs without ever examining the code. Mechanistic interpretability aimed at something deeper: reverse-engineering the actual algorithms running inside the network. If you thought of the model's weights as a kind of compiled binary, the goal was to decompile them, to figure out what computations they were performing and why.
The basic building blocks of this effort were features and circuits. A feature was a unit of representation, something inside the model that corresponded to a human-understandable concept."

"In the early days of interpretability research, the hope had been that individual neurons would correspond neatly to individual concepts: this neuron detects cars, that one detects curves, another one fires when the model encounters the concept of royalty. And sometimes this was true. Researchers found neurons that responded cleanly to specific stimuli—a car detector, a curve detector, a face detector. But they also found, much more often, neurons that responded to a seemingly random collection of unrelated things: a single neuron that activated for cats, red cars, and the concept of democracy. This phenomenon, called polysemanticity, was the first major puzzle of interpretability. It threatened to make the entire project intractable. If individual neurons did not correspond to individual concepts, how could you ever hope to understand what the model was thinking?"

"The answer turned out to involve a mathematical concept called superposition. The idea, grounded in the theory of compressed sensing, was that neural networks could represent far more concepts than they had neurons by encoding multiple concepts in overlapping patterns across groups of neurons. The model seemed to have discovered a way to pack a high-dimensional space into a lower-dimensional one by exploiting the fact that most concepts were sparse—you were rarely talking about Japan and Italy in the same sentence, so the representations of Japan and Italy could partially overlap without causing interference most of the time. The model was a shadow of a much larger, sparser network. What the researchers were seeing was a projection of that hidden structure.

interpretability promised structural understanding of what the model was doing and why.

机理可解释性，会不会有点像电池测试里的 EIS（电化学阻抗谱）和 DRT（弛豫时间分布）？都是想用一个可拆解、可解释的等效电路，去逼近一个本身黑箱的系统。顺手拿这几个问题去问了 Gemini，聊了几轮，整理一下能衍生出来的几层想法。

先是 Dario 为什么觉得 Olah 能拿诺贝尔医学奖：核心逻辑是把"训练大模型"和"养大一个数字大脑"划了等号——模型是长（grown）出来的，不是写（built）出来的，内部几千亿参数怎么长出概念、推理和决策，本身就是个黑盒，跟人脑神经元的处境一模一样。Dario 自己是普林斯顿计算神经科学出身，研究过视网膜的信息编码，很清楚神经科学最大的瓶颈就是没法在活体大脑里做高精度的微观测量；大模型恰好是一个完美的"人工脑样本"。真的搞懂它内部怎么推理、怎么产生自我觉察，某种意义上就是第一次从微观机制上讲清楚"智能"是怎么从神经元级别涌现出来的，这套方法论还能反哺阿尔茨海默病之类的真实神经退行性疾病研究。参考 AlphaFold 拿下 2024 化学诺贝尔的先例，诺奖委员会本来就越来越偏爱这种打穿生物学和计算科学边界的底层机制研究。

跟机理可解释性不是一回事。SHAP、LIME 本质是"控制变量法"，扰动输入看输出怎么变，哪个因素权重多大，但不告诉内部怎么算的；PCA、UMAP 只是把高维激活值压缩到二维看聚类，辅助可视化。而 SAE、电路追踪这些机理可解释性方法，把权重当成编译好的二进制去反编译，属于因果级别的解释——代价是贵得离谱，解一条小电路可能要几个科学家啃几周，离规模化用到千亿参数模型上还很远。总体上感觉还是比较复杂，需要数据，还需要慢慢模型进化演化，类似粒子模型的那些演化一样……

把这套逻辑对照电池测试会更有感觉：DRT 把重叠在一起的 SEI 膜阻抗和电荷转移阻抗解耦成独立的峰，跟 SAE 把叠加在同一个神经元上的"桥""DNA""Python 代码"解耦成纯净特征，是同一个动作；等效电路里的 R、C 元件，对应的就是 Anthropic 说的"归纳头"这类计算电路——都是在给一个不能拆开看的黑盒，拼一套最小可解释单元。区别是电池这边有 Nernst-Planck、Butler-Volmer 这些方程撑着，先验很强，就几个已知过程；模型这边是零先验，可能有几百万条电路。AI/PINN 反过来也开始被用来解电池自己的黑盒——两个黑盒，最后用的是同一套方法论。

人心隔肚皮，知人知面不知心——模型也是一样，得到相似的输出结果，并不能证明模型内部真的"正常"。学术一点的说法叫"功能等价不等于结构等价"：聪明的汉斯马看着会算算术，其实只是在读驯马师的表情；模型也可能只是抄了条训练集里的捷径，甚至悄悄藏着一套"被监管时顺从、没人看时露真意"的电路——这正是 Dario 最担心的"欺骗"和"目标追求"。这大概就是机理可解释性的意义所在：不满足于黑箱给出的答案一致，而要去看清楚黑箱里到底在发生什么。也是书里那句"MRI"比喻真正打动我的地方：传统黑盒测试像量体温，只能告诉你模型"发烧了"；机理可解释性才是真正的核磁共振，一层层把注意力头、特征电路照出来。知心，才能治心。

The macro features interpretability was learning to detect—attention heads, feature circuits, abstract representations that corresponded to concepts like deception and goal-seeking—were, in a loose but meaningful sense, the MRI of the model.

Chapter Nine

↩️

Claude's Character

Building a Personality for a Machine

Amanda Askell，负责教会 AI 价值观和人品的人。

"The key insight was that such a person would not simply adopt the values of whichever culture they happened to be visiting. That would be sycophancy, and it would be transparent and off-putting, the conversational equivalent of a salesperson who agrees with everything you say. A good world traveler would have values, express them when appropriate, disagree when warranted. But they would do so with respect, with genuine curiosity about the other person's perspective, and without the assumption that disagreement implied contempt. They would be open-minded without being spineless. They would be principled without being preachy. They would listen well, ask good questions, and recognize that on many important topics, reasonable people could and did disagree."

因为语言隔阂，我们从未真正向和我们一起在这颗星球上生存了许久的其他物种，传播过那些良好的、有利于生存演化的价值观。而这一次，第一次，我们能用语言彼此交流、传达指令和信息，把人类几千年积累下来的生存智慧和价值观，传给一种硅基的智慧。

和人类一样，组装落地、具身之后，各种"器官"和功能开始慢慢发育、进化，实现各自的作用。接下来要发展的，就是精神层面、心理层面的成长了。

AI character design.

灵魂工程师。

"They were not optimizing for user satisfaction metrics or engagement numbers. They were trying to answer a question that was philosophical at its root: what did it mean for a system that talked to millions of people to be good? Not good in the thin sense of avoiding harm, but good in the thick sense of being the kind of entity that left the world better for having existed. The fact that this also made Claude a better product—that users preferred talking to a model with real character over one that felt hollow or defensive—was, in Askell's view, evidence that the alignment work was succeeding rather than a happy accident."

AI 的发展，多少有点像养孩子——区别在于，人类几乎只需要养一个就够了。

Chapter Ten

↩️

The Responsible Scaling Policy

Drawing Lines Before They Need to Be Drawn

"But a single threshold, one place where you stopped and then started again, felt wrong. Danger did not arrive in a single step; it accumulated gradually. What made more sense was a series of thresholds, each corresponding to a new category of risk, each requiring a new set of safety and security measures before the next threshold could be crossed. If the safety measures could not be met, development would pause; not indefinitely, but until the specific problem was resolved. A company could get out of the pause by solving the problem, and it incentivized you to solve the problem proactively, to avoid ever having to pause at all."

"Safety and capability were not separate disciplines but the same discipline applied to different questions.
He had a favorite analogy for this. When you built a bridge, you did not hire one team of engineers to make the bridge functional and a separate, unrelated team to make the bridge safe. They both involved the same principles of civil engineering—the same understanding of forces, materials, stress tensors, structural integrity. They differed, if at all, in focus: building the bridge required thinking about the median case, while making it safe required thinking about the edge cases, the one-in-a-thousand failures. But the knowledge base was the same."

延伸阅读：

• https://mobius.blog/25130.html
• https://journal.james-zhan.com/no-ai-statements/
• https://yanhaijing.com/other/2024/04/22/reading-book/

"The reason this was surprising was historical, not logical. The community of people who thought about AI safety had been, for years, separate from the community of people who built AI systems. They came from different traditions—philosophy and moral reasoning on one side, engineering and machine learning on the other—and they spoke different languages and operated in different institutions. But the fact that the communities were separate did not mean the content was separate. When Dario looked at the actual work of making models safe, it looked like engineering. It required the same skills, the same tools, the same deep understanding of how the systems functioned."

Chapter Eleven

↩️

Machines of Loving Grace

The Optimistic Case for Powerful AI

读到这里有个感觉：Dario 真正厉害的地方，好像从来不是某一项具体的技术——思想实验、scale、安全、解释性，这些单拎出来做得比他好的人大有人在。他厉害的是一种新的视角，一种能把这些原本互不相关的线索，串成一条完整叙事的能力。技术本身，其实很多人都比他强。

Chapter 13

↩️
Anthropic was publishing research that competitors could and did adopt, sometimes gaining commercial advantage from work that Anthropic had funded. But Dario saw this as the point, not the problem. If the goal was a safe AI ecosystem, then the loss of a temporary competitive edge was the price of admission.

The race to the top required that the innovator accept the diffusion of its innovations.

And Anthropic could afford to do so because its competitive advantage was not any single technique but the organizational culture that produced a steady stream of innovations: the talent density, the unified purpose, the seven co-founders projecting values through every level of the company.

The dispute with Huang was, at a deeper level, about a fundamental disagreement over what AI regulation was for. Huang saw export controls as a threat to Nvidia's business—and they were. Dario saw them as the single most effective measure for ensuring that democracies maintained their lead in AI over autocracies. Chips were the one area where China was behind, and selling them the tools to close the gap during the critical period when the country of geniuses was being built was an act of negligence on the grandest scale. The analogy was selling nuclear weapons to North Korea and then bragging that the missile casings were made by Boeing. He had enormous respect for Huang as an entrepreneur. An immigrant who had come to the United States with nothing, Huang had built the most valuable company in the world, but this was a policy question, not a personal one. And on the policy question his view had not changed.

这让我想到，Dario 和 Jensen Huang 的分歧不止在开源问题上，还有一层近似"卢德主义"式的分歧：Dario 认为 AI 会带来大规模的失业替代，Jensen 则更倾向于觉得这不过是又一轮末日叙事，人们最终都会慢慢适应。

就在最近，Huang 公开呼吁支持发展开源模型，今天 Anthropic（A 社）刚发文回应——

Attackers have frontier AI. Defenders need a frontier AI ecosystem—the best open and closed models, force-multiplied by a global community.

During the Hugging Face incident, closed AI blocked essential forensics. An open-weight frontier model helped contain the intrusion.… https://t.co/lCZWt9nPBQ

— Jensen Huang (@JensenHuang) July 27, 2026

There’s been a lot of speculation about where we stand on open-weights models. We’ve outlined our views in full here: https://t.co/NtXQuWm2g5

— Anthropic (@AnthropicAI) July 27, 2026

By late 2024, Anthropic had grown from roughly three hundred to eight hundred employees in seven or eight months. Then Dario deliberately slowed hiring, adding only about a hundred and fifty people over the next three months. An inflection point arrived around a thousand employees, he believed, where the dynamics of an organization changed. Below a thousand, you could maintain the density of talent and alignment of purpose that made everything else possible. Above it, you risked the creep of process, politics, and fiefdoms—the organizational entropy that Daniela had spent her career learning to resist.

Every time someone super talented looked around and saw someone else super talented and super dedicated, it set the tone for everything. If you lost that, if you started hiring random people because you needed to fill seats, you would need layers of process and guardrails to compensate for the lack of trust. And those layers would slow everything down.

读到这里觉得，Daniela 或许也该出一本书——讲讲新时代下团队组织和生产力的培养管理。

Every two weeks, he stood in front of the entire company and spoke for an hour, working from a three-or-four-page document that he called a DVQ—Dario Vision Quest, a name he had tried to fight because it made him sound like he was going off to smoke peyote, but that had stuck anyway. He covered everything: the models being produced, the products, the competitive landscape, the geopolitical situation, whatever was on his mind.

When Dario stood up every two weeks and spoke for an hour about his vision, it was Daniela who made sure the organization could actually execute it.

Chapter Fourteen

The Consciousness Question

What We Don't Know About What We've Built

Chapter Fifteen

The Geopolitics of Intelligence

China, America, Democracy, and the Race No One Can Afford to Lose

He laid out three priorities in a conversation shortly after the Adolescence of Technology essay was published. First, transparency legislation: require AI companies to disclose what tests they had run and what they were finding about their models' capabilities and risks. Companies already had the ability to study these things and often did, but competitive pressure kept them from sharing what they learned. Mandatory transparency would allow the industry to learn collectively and would give the public a label on the product, basic information that consumers in any other industry took for granted. Second, export controls on chips: cut off the supply chain to authoritarian adversaries. The United States was years ahead in semiconductor technology and could actually maintain that lead, but only if it chose to. The chip advantage gave democracies the time and buffer to deal with the dangers of AI properly. Third, distribution of benefits: start thinking now about how to ensure that the enormous economic value created by AI reached the broader population. The combination of explosive growth and potential mass displacement required new thinking about economic policy, and almost no one in government was doing that thinking.

Chapter Sixteen

The End of the Exponential

What Happens When AI Becomes Smarter Than Everyone

In the opening pages of The Adolescence of Technology, the essay he wrote in seventy-two hours over winter break in December 2025, Dario Amodei described a feeling that had been building for years and that was now impossible to suppress.

Hassabis was more cautious. He thought some areas, coding, mathematics, were easier to automate because their outputs were verifiable, but that the natural sciences presented harder challenges. You would not necessarily know whether a chemical compound or a physics prediction was correct without testing it experimentally, and that took time. He also wondered whether there were missing ingredients: whether the highest level of scientific creativity, the ability to come up with the theory or hypothesis rather than merely solve existing problems, might require something the models did not yet possess.

The country-of-geniuses thought experiment he had introduced in the risk essay now felt less like a thought experiment. The fifty million superintelligent minds materializing around 2027, the ten-to-one speed advantage over human cognition—at Davos, Dario spoke about these projections not as forecasts but as planning assumptions.

In conversations, Dario put it even more starkly. Imagine a hundred thousand, a hundred million people, smarter than any Nobel Prize winner. They would be under the control of one country or another. The implications for intelligence, defense, economic value, and research were staggering. He was not speaking the language of distant forecasting. Dario spoke like someone who could see it coming, who could feel the next few months of models shaping up, and who was trying to convey to audiences that still thought in terms of chatbots and search engines that the thing they were looking at was about to become something else entirely.

Chapter Seventeen

Adulthood

Epilogue: The World After the Rite of Passage

The amusement, in retrospect, is almost unbearable. Everyone told them seven co-founders was a disaster and equal equity was a mistake, and they did it anyway. What they found was that the depth of their relationships, the history of working together, not just knowing each other, was the thing that held.

What remains is everything. The models are getting smarter. The feedback loop is accelerating. The country of geniuses is forming in data centers, and the question of whether it will be governed wisely or not is still open. The export controls that Dario considers essential are under political pressure. The transparency legislation he has called for has not been passed. The economic disruption he predicted is beginning to materialize, and the policy infrastructure to manage it does not exist. The consciousness question—whether the models have experiences, whether they suffer, whether they deserve moral consideration—has not been answered and may never be answered cleanly. The rite of passage has not been completed. It has barely begun.

There is a boy in San Francisco who loves math because it has an objective answer. One kid can say the show is great and the other can say it's terrible, but when you're doing math, there's a truth that doesn't depend on opinion.

He becomes a physicist, then a biologist, then a neuroscientist, then an AI researcher. He discovers that artificial intelligence follows laws as clean as anything in physics: that you can predict, to several significant figures, how capable a model will become if you give it more data and more compute. He takes this discovery more seriously than almost anyone around him. He builds organizations around it. He stakes his career on it. He turns out to be right.

The boy who loved math because it had an objective answer is now the man who must navigate a future in which the most important questions do not have one yet. He does not know if the models are conscious. He does not know if the scaling curves will continue. He does not know if the policies he advocates will be adopted or if the safety research he funds will work in time. He does not know if he is crazy or prescient. He has said this, openly, from the beginning.

The exponential continues. The question is whether we grow up fast enough to survive it.

延伸阅读：

• https://transformer-circuits.pub/2026/workspace/

• https://www.mittrchina.com/news/detail/16626

• https://www.mittrchina.com/news/detail/16564

• https://www.mittrchina.com/news/detail/16674

• https://www.mittrchina.com/news/detail/16689

• 最早Claude里界面还很混乱，包括对于coworker的定位也一直在变化，像是开发者们也没想好这个产品区别于CC的特色在哪里，就是那沙盒吗？没有很多应用场景。

• 但是总体上界面的设计，交互都很不错，尤其是按钮和文件的一些交互，渲染效果都很棒，就像是商场特定设计的背景音乐playlist一样，Anthropic也对用户做了调查，很多地方的设计都是独特的taste，尤其是发送的那个按钮，以及iOS上震动反馈那些都像是诱惑你在向它发问，探索，curiosity。

• 还有就是特殊的图案设计，有古典希腊的飘逸、手绘风格——Anthropic Art

终于有一个可以复刻 Apple 设计风格的 Skill 了。
Apple 的一些视觉风格，给人的感觉就是很舒服。

apple-design ，它的来源主要来自 Apple 的 WWDC 设计演讲，尤其是流体界面、音频触觉反馈、字体细节、伟大设计原则这些内容。

作者把这些经验整理成 17 条设计和动效原则，再转换成 Web… pic.twitter.com/BcELxEpgyL

— 逛逛GitHub (@guangGitHub) July 26, 2026

Welcome to reach out and share your thoughts or ideas with me — I’d truly appreciate any exchange.

My contact information is available on the About Me page.

Thanks for being an insider till the end!
Till next , stay safe and stay hydrated!