neverland

Why smoking is unbannable on railway platforms in China

Sat, 27 Jun 2026 07:20:41 +0800

车站虽受国铁集团管理，但作为企业，无论国铁集团还是其下设的卫生监督所，都没有执法权。一名地方铁路系统人士介绍，车站和铁路卫生监督部门均无行政处罚权和罚没权，收缴罚款也无法上缴国家财政。

那么，谁才能开出罚单？华东交通大学铁路法治研究院院长朱新建称，铁路系统内真正有执法权的是铁路公安和铁路监管局。但铁路公安的执法主要集中在《治安管理处罚法》和《铁路安全管理条例》，让他们在站台上针对控烟执法不太现实；铁路监管局的职能是监督铁路运行安全和施工安全，不涉及旅客健康事务。

尽管 2019 年修订后的深圳控烟条例将高铁站台划入禁烟区，但时至今日，深圳北站尚未开出一张罚单。

“我们也提出来，如果你们不干，授权给我们来干。”深圳市控烟办相关负责人称，他们曾与铁路部门多次沟通，希望对方授权控烟办进入深圳北站候车大厅和站台执法，但车站认为涉及人身和铁路运营的安全问题，没有答复。

垂直管理又较为封闭的铁路系统，难向地方控烟人士敞开大门。朱新建分析，铁路系统有长期封闭独立的历史惯性，车站是保障运行安全的重地，有严格的区域管控。

在较早实现车站站台无烟立法的秦皇岛，也有相同困扰。秦皇岛市疾控中心相关负责人说，他们同样不被允许进入秦皇岛站的站台区域执法。

只能另辟蹊径。“我们抓住市爱卫会领导小组全面调整的契机，将秦皇岛站所有相关铁路段比如车务段、机务段等，悉数列入爱卫办成员单位，用’红头文件’明确了其控烟职责。”上述负责人说。

从纸面到落地，同样需要策略。该负责人称，他们曾拿着文件与铁路部门反复沟通，面对对方“承接旅游业务繁忙”的推脱，主动提出“你忙的话我们来帮你”，提供海报、控烟督导人员、营造宣传氛围的同时，借助河北省每年三四月份的控烟履约第三方暗访形成外部压力。目前站内环境已大幅提升，双方建立了良好的配合关系，“但和深圳一样，站台上尚无处罚案例，仍在持续推动中。”

华东交通大学铁路法治研究院副院长曾明生从现实层面分析，无论是铁路公安还是地方执法部门，都面临编制有限、任务繁重的现实约束，控烟与反恐、治安等事务相比，并非最紧急的执法事项。而且执法缺乏内在动力，铁路运营主体有“多一事不如少一事”的观望心态，担心激化一线矛盾、影响旅客满意度考核，缺乏主动作为的内生动力。

全国性无烟立法的缺位，也很难做到“全路一盘棋”。前述铁路系统人士分析，由于跨区标准不一，旅客从北京到广州，穿越多个省市，如果每个城市站台规定不同。这里禁烟处罚、那里不管，旅客将无所适从，容易引发公众舆论反弹。

Comment by 财新网友纯路人:

此事在《是，首相》中亦有记载。

关于规范使用人工智能工具：致《文艺研究》编辑部的一封信

Fri, 26 Jun 2026 23:13:19 +0800

徐贲：

学术共同体真正应当追究的是责任（responsibility），而不是来源（origin）。因为来源从来都是混合的，而责任必须是明确的。

传染性授权的得与失

Thu, 25 Jun 2026 06:39:42 +0800

feeshy:

[Paraphrasing an opinion video] A24 拍的《后室》避开了开源的后室 Wiki，选择了 YouTube 创作者 Kane Parsons 的个人宇宙，以换取独占的商业路径；而社群创作的「SCP 基金会」世界观则受困于其采用的 CC BY-SA 3.0 协议，导致好莱坞资本因合规风险不敢介入，其影视化只能在社群众筹的圈子里艰难推进。[End of paraphrase]

[双重] 授权协议的签订理论上需要全体版权人达成共识，这在实践中几乎不可能，光是联系历史上参与过编辑的每一位作者就是一项浩大的工程，更别提挑战性更高的后续如何分配「蛋糕」的问题了。

在 IP 孵化期（影响力从 0 到 10），传染性授权能依靠共享激发社区协同创作、扩大 IP 的传播范围与影响力；可到了破圈期（影响力从 10 到 100），传染性又反而会在某种程度上阻碍 IP 的进一步扩圈 […]

事实上，在许多为了传播的生态中，即使没有上游传染性授权的强制要求，创作者往往也会自觉共享。比如采用 OFL 开源协议的思源系列 CJK 字体 […] 即使没有协议强制，无数设计师和厂商为了促进排版生态的传播，依然自觉基于它开发衍生的开源字体回馈社区 […]

‘Welcome datacomp’

Tue, 23 Jun 2026 07:00:03 +0800

Jason Snell:

[According to John Norstad, the creator of the definitive free Mac antivirus app,] it’s a practical joke in the ROM code in some third-party keyboards [that if you didn’t type anything for a while they would themselves type “welcome datacomp”.]. The only solution is to get your bad keyboard replaced.

My guess is that this wasn’t a practical joke as much as some sort of firmware test gone awry, but regardless, “welcome datacomp” managed to appear in numerous publications, as Haller’s post details.

In memory of the man who put red and green squiggles under words

Tue, 23 Jun 2026 06:17:43 +0800

Raymond Chen:

In early versions of Word, the Spell Check feature was something that you explicitly invoked, and then you had to sit and wait while the program looked for all your potentially-misspelled words, and then showed them to you one at a time for a decision on what to do for each one. Word did introduce an Auto Spell Check feature to run spell check when the user was idle, so that when you hit the Spell Check button, the results were ready to go. However, the Auto Spell Check was still a blocking operation. As a result, a lot of users turned it off because it always seemed to decide “Now would be a good time to spell-check the document” just as you wanted to do something, forcing you to wait for the spell check pass to complete before you could, say, save and exit.

Tony [Krueger] made the spell checker much more unobtrusive so that it didn’t interfere with your foreground work. And when it found a problem, instead of waiting for you to trigger a spell check, it immediately drew red squiggles under potentially-misspelled words (and later green squiggles under potential grammatical errors).

Photos

Sun, 21 Jun 2026 18:48:00 +0800

Note

Sun, 21 Jun 2026 15:33:00 +0800

看看报纸。

Photos

Sat, 20 Jun 2026 17:45:00 +0800

Photos

Thu, 18 Jun 2026 18:23:00 +0800

Photos

Sun, 14 Jun 2026 17:44:00 +0800

Photos

Sat, 13 Jun 2026 18:44:00 +0800

pikabubu

Sat, 13 Jun 2026 15:00:00 +0800

Photos

Wed, 10 Jun 2026 07:31:00 +0800

Photos

Sun, 07 Jun 2026 18:24:00 +0800

Photos

Sun, 07 Jun 2026 15:35:00 +0800

Oura 4 to 5

Sat, 06 Jun 2026 23:32:00 +0800

Photos

Sat, 06 Jun 2026 15:06:00 +0800

在香港骑共享单车：访客视角的经验

Fri, 05 Jun 2026 00:00:00 +0800

骑共享单车游香港，听起来似乎不是什么主流选择。的确，香港总体来说对单车不算很友好。特别是港岛、九龙一带，地形起伏大，开发密度高，基本没有条件开辟单车径，政府也明确不鼓励以单车代步。

不过，在临近深圳的新界区域，地势相对平坦，新市镇开发过程中也普遍规划了单车设施，可以骑行探索的部分其实颇多。再加上香港的共享单车市场经历一轮发展和洗牌，已经逐渐稳定，车型和使用流程都与内地无异；只要稍微学习适应当地的骑行规则，骑车游玩基本没有门槛。

去年下半年，我因为懒得走路，临时起意扫开一辆车，从荃湾骑到屯门，体验相当好，此后便开始主动选择骑共享单车。到现在，共享单车已经成为我在新界闲逛时的常用交通工具，既可以独自支撑半天的远途行程，也可以用来连接两段不方便转乘的公共交通，非常值得推荐。

位于天水围和康城的单车径

因此，这篇文章将以一个内地访客的视角，介绍在香港骑共享单车的方法和注意事项，并推荐一些我走过的骑行路线，希望能提供一种「自食其力」地探索香港的新思路。

>> 共享单车类型与收费

香港从 2017 年起出现私营共享单车服务，前后出现过六七家运营商，但目前只剩下两家，分别是香港本地品牌 LocoBike，以及内地平台哈啰旗下、2025 年进驻香港市场的 HelloRide。

骑行体验上，香港的共享单车已经与内地高度相似。从车型设计看——

HelloRide 近期投放的单车与内地的较新款哈啰单车几乎完全相同，即通体蓝色涂装，大部分还有手机支架；
而 LocoBike 亦从去年开始与滴滴青桔合作，计划在两年时间内更换全部旧款单车，更换后的车型与内地的青桔单车也几乎完全相同，少部分有加装的橡胶手机支架；
唯一的明显区别是，这些「港版」单车均加装了前灯。（香港法例规定，任何人不得在黑夜时间或能见度低的情况下，在道路上骑单车，除非该人在车头展示一盏白灯及在车尾展示一盏红灯。）

LocoBike 和 HelloRide 的新近车型（官方图片）

使用流程上，LocoBike 和 HelloRide 都支持 app 或微信小程序骑行，后者可以直接从微信支付换汇扣款；与内地的一个小差别是需要先打开相应的小程序再扫码，不能直接从「扫一扫」界面扫码骑行。两者均不需要押金。

LocoBike 和 HelloRide 的小程序界面

定价方面——

LocoBike 需要先购买「踩车券」才能开始骑行，其中最适合访客的方案是 20 港币的「一日通行证」，可于 24 小时内无限次使用；
HelloRide 的模式则更接近内地，不充值即可骑行并按时长计费，但价格偏高，为每 30 分钟 10 港币。因此，仍然建议提前购买「一日无限次卡」，价格为 28.8 港币，每次骑行的前 2 小时免费，之后则正常计费。

（券、卡的定价可能因账户新旧、时令优惠等因素而浮动。）

仅从成本而言，LocoBike 是明显更低的。但在我的观察中，可能由于投放时间更近，HelloRide 的车况往往好于 LocoBike（尤其是服役时间长的旧款车型），而且手机支架也更稳固。因此在有得选的情况下，我一般宁愿多花点钱骑 HelloRide。

>> 可骑行区域与导航方式

如开头所述，香港适宜骑行的区域主要集中在新界。2020 年 9 月，新界已建成总长约 60 公里的单车径网络，其主干线由屯门经元朗、上水、粉岭、大埔、沙田，东至马鞍山。此外，荃湾海滨、将军澳及东涌一带也铺设了单车径。以上大致也是 LocoBike 和 HelloRide 的运营范围。因此，只要计划行程的起止位置位于上述范围之内，一般都是可以找到单车路线的。（至于维港海滨的休闲单车径以及离岛越野骑行，并不在共享单车的可达范围，本文暂不讨论。）

屯门至马鞍山段的单车径干线（香港土木工程拓展署）

说到导航，你或许会抢答说，我知道我知道，在香港要用谷歌地图。但遗憾的是，谷歌似乎完全放弃了香港骑行的使用场景，经常规划不出任何骑行路线。如果一定要用的话，可以用驾车导航凑合凑合。这是因为在香港步行需要频繁走上天桥并借用商场等室内空间，步行导航的路线对于骑行的参考意义有限；而单车是可以上行车道的（详见后文），所以跟着汽车走一般问题不大。

相比之下，高德地图反倒能在香港规划出比较合理的单车路线，脏是脏了点，还是推荐捏着鼻子用一下。如果是苹果用户，直接用内置地图即可。

此外，如果需要更详尽、权威的路线描述，可以使用香港运输署推出的「香港出行易」应用。不过，出行易只能规划出全程单车径的路线，而且它可能是你见过最难看、最卡的应用之一，所以只适合用来确认路线局部的细节。

谷歌、高德和出行易的路线规划

需要注意的是，根据香港法例，单车不得驶入若干场所，包括游乐场地、泳滩、郊野公园、隧道以及快速公路。在新界北部的边境地带骑行时，还须特别留意不要误闯边界禁区。因此，规划骑行时，应当注意检查导航路径，不要误入上述范围；骑行中，也应当关注各类约束单车的交通标志和道路标记，避免违例。

部分约束单车的标志

例如，下图分别为轻轨蓝地站对出的交汇处，以及青山公路新田段与新深路交界处。在前者位置，中间岔路将要上坡驶向元朗公路，而元朗公路属于快速公路。注意路牌上黄色的 9 号干线标志，以及「禁止行人、单车进入」，即可知不能骑行进入。在后者位置，新深路通往落马洲管制站，即深方皇岗口岸，属于禁区。注意路牌上红色的「道路封闭」标志和「禁止行人、单车进入」标志，亦可知不能进入。

>> 骑行规则与注意事项

提到香港与内地骑行的区别，一般都能想到香港的道路通行方向与内地相反，采用靠左驶交通：前行使用左侧道路；除超车外，使用左侧车道。这些规则同样适用于单车骑行。

但除此之外，在香港骑行时，关于车道选择、车辆操作和停车等事宜，还存在一些与内地形成的直觉不同之处，值得在初次骑行前知悉。下面就是我根据自己理解整理的主要规则，来源均为香港法例及香港运输署编写的《道路使用者守则》，后者特别值得在香港骑行前完整阅读了解。

>>> 车道选择

初次在香港骑行的人最容易犯错的大概就是「在哪个道骑」。在有单车径的地方走单车径，自然是天经地义。（根据香港法例，凡道路有一部分划给单车使用，任何人不得在该道路的其他部分骑单车。）但如上文介绍，单车径的覆盖范围毕竟是有限的。如果途径路段没有单车径，应该选哪个道呢？

如果你来自深圳这样人与非机动车抢路的文明城市，可能习惯性地想，毕竟我是个人，那就上人行道吧。如果你这样想，那么将会和我上车的前五分钟一样，在路人的白眼中不断怀疑自己是不是个人。实际上，无明显需要而在行人路上策骑，即属妨扰罪，可处第 1 级罚款（2000 港币）或 3 个月监禁。

（香港法例并不区分刑事犯罪与行政违法，条文中的「罪」「罪行」「犯罪」均系 offence 的直译，可指任何违反订有罚则的法律条文的行为，因此其性质并不都如内地语境中一样严重，当然仍需注意避免。）

根据排除法，你已经知道正确答案是车行道。香港法例将车辆（vehicle）定义为「任何经构造或改装为用于道路上的车辆，而不论是否由机械驱动」（轻轨和电车除外）。换言之，单车在香港被视为一般车辆，除非有另行规定，可以（并且应当）在机动车道上行驶。这也意味着在车道上时，单车须单排前进，除非超车，不能与其他车辆并排行驶。

值得一提的是，与汽车不同，单车可以进入巴士线（即公交车道）。除非巴士线上竖有禁止单车驶入的标志，否则单车可沿巴士线骑行，且不受标志所示可通行时间的限制。不过在巴士线上骑行时，应在情况许可下让后方的巴士超车，并留意有无行人突然走出马路。

>>> 过路口、转向和超车

如上所述，既然单车被视为车辆，骑行者也须遵守适用于机动车的交通标志、灯号（即信号灯）、道路标记和交通规则，犹如是在驾驶汽车一样。基于这一原则，以下这些场景的处理方式可能与你在内地骑行的习惯不同，值得提前留意：

（一）过路口。如果原本在车行道上骑行，过路口的方式与其他汽车相同：应遵守车行道的红绿灯，遇红灯须在停车线后等候，或排在其他候灯车辆之后或之间，而不能与其并排。出于安全考虑，最好在通过路口前先转入最左侧的直行车道，以避开车速较快的右侧车道。

特殊地，如果遇到的是不设红绿灯的回旋处（即环岛），则应靠左驶入，并始终贴着左侧边缘顺时针骑行，直至从所需的出口离开。

此外，根据运输署的建议，若路况复杂、车流较大，以行车方式过路口令你感到不安全，也可靠边下车，推行经斑马线过路口。

如果原本在单车径上骑行，遇到路口时通常会看到「终止」的地面标记或交通标志，以及「骑单车者到此下车，由附近行人线过路」的指示牌。此时应按指示下车推行，待人行道的绿灯亮起后推行通过斑马线，再在对面单车径的起点继续骑行。

位于元朗朗日路的单车径及路口

还需注意的是，香港单车径是迁就地形和路况而铺设的，经常在中途终止，换到道路对侧，在此过程中需要借用行人天桥、隧道。（这些路段在运输署官方单车径地图及「出行易」应用中以红色标注。）因此，如果你在骑行过程中发现前方突然找不到单车径，多半是因为已经错过了分岔位置，应当折返并寻找相关路标。

「出行易」中的东涌单车径地图，其中红色部分为经桥梁、隧道过路位置，而单车径在过路后将切换到道路对侧

（二）转向。香港靠左行驶，左转沿道路边缘前进即可，相对方便。但须注意，香港没有「红灯转向」的道路设计，只有在绿灯时才可左转，而不像内地那样，只要没有右转箭头灯时便可径直右转。

至于右转，则应提前小心变道至右转车道，并提醒自己右转之后须汇入左手边的车道，与内地相反。

左转及右转的处理方式

此外，根据运输署的建议，在转向时应当以手势发出信号：伸出左手表示左转，伸出右手表示右转。不过，这并非强制要求，况且单手骑行有一定风险，路上也很少看到有骑行者使用手号。

（三）超车。共享单车的速度上限很低，一般没有超车的需要。如确需超车，应从前车右侧超越，并尽可能保持至少一道车门的距离。只有在仅设一条行车线的单程路（即单行道）上，或在即将右转的车辆之后，才可从左侧超车。

>>> 停泊单车

香港法例规定，在设有街道照明的道路上，单车只能停放在有单车泊车处标志的指定泊车处。共享单车小程序的地图上也都会标出能够泊车的位置，超出范围就不能还车。不过，没有必要在规划行程之初就选定停车位置，只要将导航终点选在港铁站、公共交通交汇处、单车汇合中心等设施附近，一般都容易在附近找到泊车处。

至于在骑行中途，如需作短暂停留休息，可以把单车停放于行人路上或路旁，但要确定单车不会阻碍行人、构成危险或影响视线。

泊车处标志

有共享单车的地方就有乱停乱放。实际上，虽然香港共享单车表面上会非常严格地执行还车范围，但和内地一样，只要静置一段时间就会自动还车，并且一般也不会触发罚金。可能正是因为这样的漏洞，如今新界许多街道上也已经是共享单车乱停成灾。当然，本文还是提倡文明骑行，不要在指定位置外停车。

>> 推荐骑行路线

文章最后，推荐几条我自己走过的骑行路线，供有兴趣尝试的读者参考。值得指出，单车骑行的乐趣之一就在于偏航；我所列举的路线也大多是在按网上攻略行走时「东张西望」而探索出来的。因此，我也建议读者也无需拘泥于本文或网帖所给出的路线，看好大致方向后随兴所至即可，相信会有更多收获。

>>> 荃湾西—屯门

（约 17 千米起步，需 1.4—1.9 小时起步，具体视乎所选终点。）

这就是我第一次在香港骑行的路线。荃湾西—屯门本身是一条热门的轻徒步路线，沿途左手边一直是海景，并且全程都是大路，坡度少而平缓；唯一的缺点就是有点长，走到最后容易乏味。改为骑行后，风景不减而时间更好控制，亦方便衔接后续交通。

这条路线的起点是港铁荃湾西站。从 B 或 D 出口前往荃湾公园，即可看到沿海铺设的单车径。沿单车径向西行到尽头，可从海安路隔音屏障这个网红打卡地走上青山公路。

继续向西，一路沿青山公路各段骑行，途中可以看到从青衣岛延伸出的多座大桥，也可按兴趣在深井、黄金海岸等处下车游览，最后到达屯门。

后续可以选择继续向西南骑行，沿海荣路前往屯门码头、蝴蝶湾一带看海；或者继续沿青山公路屯门段往北，前往屯门市中心一带的商圈。最后通过屯门市广场附近的 B3 或 B3X 巴士即可返回深圳湾。

>>> 上水—河上乡—塱园湿地

（去、回程各约 5 千米，各需 0.4–0.6 小时；停车游览时间另计。）

这条路线可看到边境村庄景观，特别推荐 11 月左右前往，更有香港少见的秋收稻田。起点是港铁上水站，出站后即可看到大量共享单车停泊，沿单车径前进，不久可看到石上河。

继续向西北骑到河流分叉处，转向西南沿双鱼河下行，即进入河上乡范围。村口的「豆美味豆花团」是一处有名豆品店，主打 15 港币不限量续碗的豆花，另有其他豆制品小吃，值得一试。

继续沿河往西南，即可看到塱园湿地的入口（亦可从河上乡村口对出小路进入），需要下车步行进入。其中面积颇大，划分为农业区和生态区，分别种植多种农作物，及为水鸟和水生生物提供栖息地，建议预留 1—2 小时的参观时间。参观完后，仍选择靠近双鱼河一侧（西侧）的出口离开，骑车原路返回即可。

>>> 大围—大美督

（去程约 23 千米，需 1.9—2.6 小时。）

大围—大美督可谓是小红书上最网红的一条骑行路线；如果你搜索「香港骑行」，几乎一定会在前几条笔记中看到这条路线。客观地说，这条路线能出名是有理由的：起迄点交通便利；全程都有完善单车径，沿途可观看到城门河景和吐露港海景；后段地形有一定起伏，有骑行乐趣而又不会造成太大体力压力；终点还适合观看日落和烧烤。

但同时，这也是最令我心有戚戚的一次骑行。就在我原本规划骑行这条路线的当周，宏福苑发生了大火——正好位于线路途径的位置。结果，我在周六（11 月 29 日）从近距离看到了余烬刚灭的火场，以及满是物资募捐点和捧花悼念者的大埔街头。远处另一个「网红机位」——慈山寺的巨型观音像俯瞰着下方鳞次栉比的比华利山别墅群——此时也似乎也显出几分凭吊之意，又或是天地不仁的讽刺。

这条路线的起点是港铁大围站，从 E 口出站即可看到大量共享单车停泊，出发后一路沿单车径行走即可。其中，到达大埔海滨公园时，与通常推荐的路线不同，可选择转向东行，前往公园尽头的码头观看吐露港，然后向北穿过大埔工业邨即可返回汀角路上的单车径。

回程时，如尚有体力，可骑行 9 千米左右返回大埔。否则，也可从大美督搭 75K 巴士到尾站，即可从港铁大埔墟站乘东铁线返回深圳。

作为本线路的支线，汀角路中途向北可达的慈山寺、以及中途向南可达的马屎洲特别地区，都是非常值得一游的目的地。不过这两处各自需预留半天时间游玩，并不适合挤进上述长途骑行中，可以在「二刷」时直接从大埔出发前往。

>>> 朗屏—丰乐围—天水围

（约 16 千米，需 1.3—1.8 小时。）

这也是一条很有趣的边境路线，可以看到新界西北的鱼塘景观，并从香港方向遥望深圳湾对面南山区的闹市景象。从港铁屏山站出发，沿福喜街向北进入盛屋村，再往北便是丰乐围的成片鱼塘。（如想有更丰富的体验，可以先行一段 3 公里、爬升 200 余米的横洲—髻山山路，俯瞰下方景色，然后从盛屋村开始骑行。）

丰乐围并无正式单车径，全程需在鱼塘间的田垄上穿行，要注意颠簸，以及不要惊扰民宅门口的犬只。（你还会看到很多问候祖宗的禁止倒垃圾告示。）此处地处乡野、颇显荒凉，抬头却尽是深圳湾对面的高楼大厦，颇有几分对比反差的趣味。

向东行到鱼塘尽头，即到达边境路，道路右手一侧是以铁丝网限制进入的米埔自然护理区。沿着边境路大致向西北方向上行，即到达尖鼻咀，此处设有洗手间和观景台，可以登高眺望并作休整。

从尖鼻咀往西进入深湾路，在道路分叉处转向东南，就来到天水围渠附近的单车径。沿渠一路骑行回到港铁天水围站附近，乘 B2P 巴士即可回到深圳湾口岸。

>>> 康城—坑口—将军澳

（约 7.5 千米，需 0.6—0.8 小时；停车游览时间另计。）

将军澳一带的规划与传统香港风格差异颇大，有一种「消灭街道」的倾向，主要商业都藏在以天桥连接的大型室内商场中，小店和街市几无发展空间，步行探索容易乏味。但就骑行而言，这反倒留下了宽敞的街道空间；将军澳的单车径也是相比之下特别宽的。

这条路线的起点在港铁康城站。如果从深圳前往，可以先乘巴士或旅游大巴到观塘，然后乘港铁将军澳线到尾站。从 B 口出站，沿将军澳跨湾大桥的指示牌行走，穿过地面的公交总站，就能看到单车停泊处。沿将军澳南海滨长廊向北骑行，经过第二座桥（北桥）时左转过桥进入将军澳。

继续沿单车径向北往坑口方向，在环保大道高架下骑行，依次会路过香港单车公园和常宁游乐场，都是值得下车步行探索一番的公共空间。随后转由绿地另一侧的宝康路折返将军澳市中心，向南回到海边，沿将军澳南公园骑行。到尽头后右转向北，可以到达香港知专设计学院、调景岭体育馆两处有特色的建筑群。最后沿唐明街抵达唐明街公园的北侧入口下车，穿过这座「高架公园」即可从港铁将军澳站离开。

>> 结语

以上就是我作为访港旅客，对于骑行共享单车的一些心得总结。经验所限，难免错漏，还请有更丰富本地见闻的朋友指正。

最后想说的是，每个城市都有许多面向。游览的动机、心境、方式不同，看到的样子也就不同。骑单车所能至之处，大都不是那些能适合发定位的网红打卡地。但在耳边市井街道的喧嚣中，轮下陡缓簸平的触感中，远处山海城乡的变幻中，相信你能形成对香港更立体、真实和个人的印象，而不只是在西九龙站牌下、中环天桥上和坚尼地城路标旁，接受和复制他人的印象。

有用资料

Photos

Thu, 04 Jun 2026 13:34:00 +0800

Photos

Wed, 03 Jun 2026 19:38:00 +0800

Photos

Sun, 31 May 2026 18:38:00 +0800

Photos

Sat, 30 May 2026 21:25:00 +0800

Day for Night

Fri, 29 May 2026 17:57:14 +0800

Marcin Wichary:

But there’s a trick [to invert black and white while preserving colors]: do a regular invert and then rotate the hue halfway through. Through the magic of math, this is the same as inverting just L in the HSL space, which means the colors are preserved.

…and, more importantly, [this is] available in CSS as a filter: instead of invert(1), use hue-rotate(180deg) invert(1).

一个冷门的速查日历方法

Wed, 27 May 2026 16:22:00 +0800

就是终端里的 cal 命令了。之所以喜欢用这个首先是因为快：我平时固定是开着终端的（作为 vscodium 的底部面板），码字过程中直接打个 cal（pun intended）就能看当月日历，比打开日历应用或者 menubar item 都要快。其次因为是终端输出，就是纯文本，可以直接复制到笔记里参考或者进一步标注。当然缺点就是比较简陋，不过我本来就喜欢这种野兽派纯文本风格，所以反而是加分项了。

以下是一些常用选项和参数：

# 显示 2025 全年
cal 2025 # shorthand for `cal -y 2025`

# 显示今年 3 月
cal -m 3

# 显示本月和紧邻的前后月份
cal -3

# 旋转行和列（每周为一列）并显示周数
ncal -w # shorthand for `cal -N -w`

更多用法可以看手册页 man cal。如其中的 history 一节所说这也是个上古命令了，从 Version 1 AT&T UNIX (1971) 就有，几十年就是这样下来了（

Note

Wed, 27 May 2026 13:00:00 +0800

don’t feel even a bit prepared for homeownership

Love Is to Be Invested in Someone’s Continual Expansion

Tue, 26 May 2026 20:20:00 +0800

Henrik Karlsson:

[E]ach time she’d grown and changed, I had felt something open up in myself. Her changes had become prompts that pushed me toward a higher, truer version of myself, just as my changes had been for her. Change answering change, it was a virtuous loop.

With your children, it becomes very clear that the love you feel has little to do with who they are or what they do.

In more superficial relationships […] you assign value to the other person because of what they do. […] But with your children, the value, to a large extent, flows in the other direction.

Adult love isn’t fully unconditional like this. There are things you would never value, even if your lover does, and there are things they could value that would make you stop loving them. But still, there is a sense that to love someone means to extend your care to the things your partner cares about to some degree.

[T]here are […] two levels at which you can extend your values like this. On the first level, you do what I did when I cut down the pines; you do it out of duty and without connecting to the values.

[T]his kind of care that doesn’t come from a deeply felt space can be grating and unpleasant both for you and the person you give it to.

[T]here is a second level at which to approach this, a more transformative level. It is when you actually learn to feel what the other person is feeling. It is when you stretch out with your imagination into the unknown and learn to merge their values and perspective into your worldview. This is one of the most profound experiences that love gifts us.

To aspire is to want to want something before you actually want it.

The aspirant sees that she does not have the values that she would like to have, and therefore seeks to move herself toward a better valuational condition. She senses that there is more out there to value than she currently values, and she strives to come to see what she cannot yet get fully into view. (Agnes Callard, Aspiration)

If I am learning, for instance, Russian, I am confronted by an authoritative structure which commands my respect. The task is difficult and the goal is distant and perhaps never entirely attainable. My work is a progressive revelation of something which exists independently of me. Attention is rewarded by a knowledge of reality. Love of Russian leads me away from myself towards something alien to me, something which my consciousness cannot take over, swallow up, deny or make unreal. (Iris Murdoch, “The Sovereignty of Good over Other Concepts”)

[T]his, to me, is one of the most profound rewards of love: that it can expand you so that more of the world becomes real to you. You can provide others with the love and care they need to transform and discover new aspects of themselves and of reality, and then, by meeting them where they are now, you can learn to see what they do, and so see further.

Which will change you.

And so change them again.

Photos

Tue, 26 May 2026 19:30:00 +0800

Photos

Mon, 25 May 2026 18:54:00 +0800

Note

Sun, 24 May 2026 22:26:00 +0800

new findings from m&s

Photos

Sun, 24 May 2026 18:55:00 +0800

Photos

Sat, 23 May 2026 11:54:00 +0800

Photos

Fri, 22 May 2026 16:12:00 +0800

Repetitive Cycles

Tue, 19 May 2026 22:07:00 +0800

Hairuo Jin:

[Chinese retail investors’ (Sanhu 散户)] attachment [to derive meaning from their past failures, unsatisfactory present, and expected future] can be situated in what scholars have described as a form of political depression, particularly what Lauren Berlant (2011) calls cruel optimism: an attachment to a promise widely recognised as fragile, even illusory, yet difficult to abandon, despite hostile circumstances.

This resolve stems from a larger developmental temporality that has deeply structured life in reform-era China and under which malfunction is often treated as evidence of progress rather than failure. From the 1990s, periods of market turbulence were seen less as structural failure than as moments of adjustment within an evolving, maturing system. Sanhus of older and younger generations continue to describe the stock market as ‘primitive’ and ‘developing’, echoing the vocabulary used by the state to narrate the course of national development.

The Slow Death of the Power User

Mon, 18 May 2026 15:54:11 +0800

fireborn:

The YouTube tutorial is the perfect emblem of this rot. Tutorials are not documentation. A tutorial teaches you to perform a specific sequence of steps to achieve a specific outcome. The steps are usually correct for the specific scenario the tutorial covers. If your scenario differs — if something’s changed, if you get an error the tutorial didn’t anticipate, if you’re using a different version — the tutorial has given you no tools to respond. Documentation teaches you to understand a system: what its components are, how they interact, what the configuration options mean and why they exist, what the error messages indicate. One produces people who can follow instructions. The other produces people who understand what they’re doing. The industry has enthusiastically replaced the latter with the former and called it democratization.

Photos

Sun, 17 May 2026 19:22:00 +0800

Photos

Sat, 16 May 2026 19:36:00 +0800

Photos

Sat, 16 May 2026 18:52:36 +0800

Photos

Sat, 16 May 2026 17:57:00 +0800

The Life and Times of an American Tween

Fri, 15 May 2026 21:13:00 +0800

Anna Wiener:

The Structure of Adolescence

[Middle school marks] a time when kids begin to pull away from their parents, discard childish pursuits, and pursue, full thrust, the exhausting project of individuation. It is a period of intense, hormonally driven emotional flux. Self-consciousness sets in. The adult world is studied and emulated in a manner that suggests praxis but no theory. There is an aspect of camp to it all: a kind of LARP or drag, as young people transition from play-acting adulthood to inhabiting it. Actual adults are ancillary. Tweens and teens look to each other for clarity and guidance on how to behave and how to feel, all the while gambling with each other’s social confidence and self-esteem. It is natural, and it is psychotic.

The psychoanalyst Erik Erikson described the period between twelve and eighteen as one of essential identity formation: a time of trying on personas and roles. Writing in the mid-twentieth century, Erikson was sensitive to the ways that society could shape personality, but no one could have anticipated the explosion of potential identities, interests, aesthetics, subcultures, and life styles that children would be exposed to by globalization and, later, the internet.

Adolescent anxiety and depression have been on the rise for years, and there is abundant public debate about why: economic inequality, strained family ties, sleep deprivation, smartphones. Childhood has never been easy, but these days the on-ramp to adulthood seems somehow shorter and more perilous.

The Private Room as Self

For any adolescent, a private bedroom is both sanctuary and mood board. It’s a safe haven for experimental selfies, a shelter for incognito-mode Google searches, an internal switchboard for the exchange of secrets and dreams.

Ritual and the Projection of Future Selves

It would be easy, having made it safely past the shores of puberty, to make light of these rituals and ablutions. But the rituals are very important — in fact, they’re everything. They’re experiments with externalizing private self-perceptions, and dalliances with potential selves. They’re a way of projecting into the future: to imagine being noticed, maybe even seen.

The Clothing of Aspiration

It felt like an A.I. hallucination of nineties fashion. I was immediately transported to my own preteen cabinet of horrors [...] In a season of knockoff Kate Spade bags, bar-mitzvah pencil skirts, and gratuitous training bras, it had produced in me a sick, horny covetousness: not for the clothes themselves, but for the idea of womanhood — flirty, filled-out, smoochable — that they implied.

In the early eighties, Neil Postman, the media theorist and critic, argued that the boundary between childhood and adulthood had long been one of information asymmetry — adults had secrets, whereas children had only their natural instincts and limited experience — and it was eroding [...] Children were exposed early to death, sex, violence, and money; one consequence, Postman claimed, was an undifferentiated culture in which children acted like adults and adults acted like children.

[I]n their eager imitation of maturity, they conveyed an incomplete but inevitably uncomplimentary reflection of adult life.

The Cognitive Dissonance of Being Twelve

Early adolescence was a sustained state of cognitive dissonance. It was a time of learning to take yourself seriously while wearing Unicorn Snot; of being clothed and transported by your parents while seeing them as critically as anyone ever would. It was a period when you could be aware of violent, sexual, criminal abuses of power; immediately identify the sexism of a “vintage” workplace dramedy; and mostly find the boys in your own life annoying. It was exhilarating, exhausting, and fleeting.

Young people changed quickly and not on any predetermined schedule. From the relatively static vantage of adulthood, it could feel like any given day with a kid was your last day with that version of them.

One of the most salient features of twelve, after all, is that it comes right before thirteen.

The Brevity of the Final Season

Watching Mira and her chatty, cheerful gaggle, it was strange to know, with no necessary qualifications other than having once been a preteen girl myself, that within the next year, or two, or three, they would be pulled into all manner of minor dramas and major insecurities, make transformative cultural discoveries, experience lifelong personal revelations. They’d have crushes, keep secrets, be teased, or be cruel. They’d be tempted to betray themselves or others. They’d outgrow their own taste, pull away from their siblings, stop laughing so hard and so freely at their parents’ jokes. They’d toss the heavily flavored balms and bottles of candy-scented spritz in the subsequent room refresh, time capsules from the final season of childhood.

Photos

Wed, 13 May 2026 19:10:00 +0800

AI 裁员潮卑鄙生存指南 The Unethical Guide to Surviving AI Layoffs

Wed, 13 May 2026 13:05:00 +0800

atmoio:

So Coinbase laid off a bunch of people last week, and the CEO sent out this letter, and by letter, I mean tweet, a really beautifully written tweet, explaining that the layoffs were not because of mismanagement. No, no, no, no, no. Brian Armstrong was a victim. We’re all victims. The AI was just too damn productive.
上周 Coinbase 裁掉了一大批员工，CEO 发出了一封信——确切地说是一条推文，写得极其「漂亮」。他解释说，裁员并不是因为管理不善。不，不，不。Coinbase 老板也是受害者，我们都是受害者。怪只怪 AI 的生产力实在太 TM 高了。

And I mean, he tried. He got on his knees and he begged the GPUs to slow down. It’s like, please think about the families, but it was no use. The productivity gains were already realized. Now, this isn’t an economics video. This is a survival video. The question is, what do you do to survive? You become the problem.
我是说，他尽力了。他甚至跪下来哀求 GPU 慢一点。他就差没说，「请为那些家庭考虑一下吧！」但没用，生产力的飞跃已经实现了。听着，这不是一段经济学视频，这是一段生存视频。问题是：为了活下去，你该怎么办？答案是：让自己成为那个「问题」。

You are going to become the most insufferable, the most LinkedIn sounding motherfucker, the most aggressively pro-AI person in your entire company, starting tomorrow. I’m serious. I want you to go to ChatGPT, and I want you to ask it for a list of 100 AI buzzwords. And I want you to print them out and tape them to your bathroom mirror. And every morning, while you brush your teeth, I want you to practice them.
从明天开始，你要变成全公司最令人难以忍受、说话最有「领英范儿」、最激进拥护 AI 的混蛋。我是认真的。我要你去问 ChatGPT 要 100 个 AI 热词。把它们打印出来，贴在你浴室的镜子上。每天早上刷牙的时候，给我对着练。

Ralph Loop. Ralph loop, stochastic, agentic, orchestration, multi-agent, neocloud. By the time you walk out of the door, you should have a low-grade fever. You should be slightly nauseous. This is good. This is a positive symptom.
拉夫循环（Ralph Loop）。拉夫循环、随机性、代理性、编排、多智能体、新云服务。当你走出家门时，你应该已经烧到了低烧。你应该感到轻微的恶心。这很好，这是积极的症状。

Then at 9 a.m. sharp, you’re going to send an email to your entire organization. Include members of the board as well. The email subject is going to be “Some thoughts on where we’re headed.” The body of the email will be: “Hey team, I’ve been giving this a lot of thought, and I think it’s time we became an AI-first company. I have some experience in this space, and I’m happy to be the Sherpa on this journey. Reach out if you’d like to set up some time.”

然后，在上午 9 点整，给全公司发一封邮件。记得把董事会成员也抄送上。邮件主题写成：「关于公司未来走向的一些思考」。正文这么写：「嘿，团队，我一直在深思这件事，我认为现在是我们转型为『AI 优先』公司的时候了。我在这个领域有一些经验，很乐意成为这段旅程的向导（Sherpa）。如果你想找时间聊聊，随时联系我。」

Okay? Send it, just send it. And at this point, if you can make this email leak, this will put you in the major leagues. Well, that’s a strategy for another video. Now, you’re going to host weekly workshops. Okay? You’ll call them “AI Office Hours.” Every Friday at 3 p.m. in Conference Room B. You’re gonna print flyers that say, “Got AI questions? I’ve got AI answers.”

明白了吗？发出去，直接发。这个环节，如果你能让这封邮件「不小心」泄露出去，你就能进入顶级玩家的行列。不过那是另一个视频的话题了。接下来，你要举办每周工作坊。明白吗？把它们叫做「AI 办公时间」。每周五下午 3 点，B 会议室。去印传单，上面写着：「有 AI 问题吗？我有 AI 答案。」

Nobody else in the company will be doing this, and certainly no one else will have flyers. But here’s the trick. The workshops are not for them. They’re for you. They are reconnaissance. You are going to find out in real time exactly how little anybody at this company actually understands AI, and you’re going to exploit this gap to figure out exactly what content to serve your audience—your audience being the AI simpletons on your team.
公司里没别人会这么干，更没人会去印传单。但这里的诀窍是：这些工作坊不是为他们准备的，而是为你准备的。这是侦察行动。你要实时摸清楚公司里的人对 AI 的了解到底有多匮乏，然后利用这个信息差，确定该给你的观众提供什么内容——你的观众就是你团队里那些对 AI 一窍不通的傻瓜。

Now, if your CEO has never heard the phrase “Ralph Loop”—oh, man. You are less than 30 days away from your next promotion. I’m not even exaggerating. Walk into his office, close the door, and say, “Hey chief, been experimenting with something. It’s called Ralph Loops, and I think it could change literally everything.” And he’s going to say, “What’s a Ralph loop?” And you will say, “Give me $18,000 worth of API credits and I’ll show you.”

如果你的 CEO 从没听过「拉夫循环」这个词——我靠，那你离下一次晋升不到 30 天了。我一点都没夸张。走进他的办公室，关上门，说：「嘿，头儿，我一直在实验一个叫『拉夫循环』的东西，我觉得它能改变一切。」他会问：「什么是拉夫循环？」然后你会说：「给我价值一万八千美元的 API 额度，我展示给你看。」

Now, you won’t actually do anything, because you can’t do anything, because nobody knows what they’re doing. But by the time he figures that out, you’ll have a new title, an equity bump. What you want to be doing is automating. Talk about automation constantly. Nothing arouses the slumbering capitalist than the mention of automation.
其实你什么都不会做，因为你也做不了什么，反正大家也都不懂。但等到他发现这一点时，你已经拿到了新头衔和更多的股权。你要做的是不断谈论「自动化」。没有什么比提到「自动化」更能让沉睡的资本家感到兴奋了。

Drop names too, bro. Talk about specific team members you can automate out of existence. Be like, ”Yo, I automated Gary, bro.” Tag Gary in the message, tag him in Slack in a very public channel. Be like, ”Yo, I just automated @Gary. His function has been Ralph-looped,” and tag your CEO in the same message. You think you’re getting laid off after that, bro? Like, are you out of your mind?

还要指名道姓。谈论那些你可以炼化的具体团队成员。比如：「哟，我把加里给炼化了，兄弟。」在 Slack 的公开频道里艾特加里，说：「哟，我刚把 @Gary 炼化了。他的职能已经被『拉夫环』替代了。」顺便在同一条消息里艾特你的 CEO。你觉得这样之后你还会被裁员吗？兄弟，你疯了吗？

This is how you survive the storm. It does not matter who is right and it does not matter who is wrong. A storm is not either right nor wrong. Like, dude, if you’re an AI contrarian at your company right now, what are you doing? Resign, dude. Resign voluntarily, man. This is highly disgraceful.
这就是你度过风暴的方式。谁对谁错并不重要，风暴本身没有对错。老兄，如果你现在还在公司里唱衰 AI，你在干什么？辞职吧。自己主动辞职吧，这太丢人了。

The only place you should be talking about AI realism is here with me or with your dog. Do not let anyone, not even your own wife, hear you be negative or balanced about AI. Are you kidding me, dude? There’s only one way to make money off being an AI realist, and I’ve already cornered the market. And I’m barely getting by, man.
你唯一可以讨论「AI 现实主义」的地方就是我这里，或者对着你的狗说。不要让任何人，甚至包括你妻子，听到你对 AI 持负面或中立的态度。开什么玩笑？当一个「AI 现实主义者」只有一种赚钱的方法，而那个市场已经被我垄断了，即便如此我也只是勉强糊口。

The most important thing, the absolute most important thing is that you are no longer going to do any work. Okay? You’re not going to write any code. You’re not even going to type. You’re going to dictate. You’re gonna use a voice tool. You’re going to speak to Claude. You’re going to speak to your team on Slack. You’re going to speak in meetings and workshops, and at no point are your fingers ever going to touch a keyboard.
最重要的事情，绝对最重要的一点是：你不再做任何实际工作了。明白吗？你不再写代码，甚至不再打字。你要用口述。用语音工具，对着 Claude 说话，在 Slack 上对着团队说话。在会议和工作坊里说话。你的手指绝不能再碰到键盘。

Because we have transcended labor, my friend. We have ascended. Engineering is no longer a craft. Engineering is a metaphysical practice now. We do not write code. We commune with code. We cleanse the repo of bad energy. We are philosophers of the codebase.
因为我们已经超越了劳动，朋友。我们升华了。工程不再是一门手艺，工程现在是一场形而上的修行。我们不写代码，我们与代码对话。我们清除仓库里的负能量。我们是代码库的哲学家。

And at this point, if you have a black turtleneck, put it on, okay? If you don’t, get one. Get two. Have a backup turtleneck. Steve Jobs did not have a backup. That was his mistake.
这时候，如果你有一件黑色圆领衫，穿上它。如果没有，去买一件。买两件，备一件。乔布斯当年就是没有备用的，那是他的失误。

Now, I want to address the people in the comments who are about to type, “Hey man, this is super messed up. This is cynical and it’s bad advice. I would never do this.”

现在，我要对那些准备在评论区留言说「嘿，哥们，这太变态了，太愤世嫉俗了，这是坏建议，我绝不会这么做」的人说几句。

And I want you to listen to me, okay? This little Gandhi stance you’re taking will not pay off. The CEO of your company is currently taking the Coinbase memo, and he’s asking ChatGPT 5.2 to draft one for his own company. He needs the views, man. He needs an invitation to the All-In Summit, okay? He needs Chamath to mention him on the pod.
听我说。你这种苦行僧的立场是不会有回报的。你公司的 CEO 现在正拿着 Coinbase 的裁员备忘录，让 ChatGPT 5.2 给自家公司也起草一份。他需要流量，老兄。他需要「All-In 峰会」的邀请函。他需要大 V 在播客里提到他。

He’s in his office studying Brian’s tweet. He is whispering, “We have made the difficult decision. We have made the difficult decision,” out loud just to feel it in his mouth. And you taking a stance against AI will not change his mind. It will not change the trajectory of AI. It will not make a fart’s worth of difference.
他正躲在办公室里钻研 Coinbase 老板的推文。他正低声念叨着：「我们做出了一个艰难的决定……我们做出了一个艰难的决定。」反复出声地念，只为了寻找那种感觉。而你反对 AI 的立场不会改变他的想法，不会改变 AI 的轨迹，连个屁的影响都不会有。

You want to take a stance? Go be vegan, man. Go open an account on Threads. But at work, in this climate, being a realist will get you canned, bro. You have one obligation, and that is to make sure there is a roof on top of your family’s heads and food on the table. Stop being such a dick, dude, and provide for your family.
你想表达立场？去当素食主义者吧，去 Threads 开个账号吧。但在职场，在这种气候下，当现实主义者会让你被开除。你只有一个义务，那就是确保你家人头上有瓦遮雨，桌上有粮果腹。别再犯轴了，去养家糊口吧。

And whenever it gets to be too much, come back here, okay? Because between you and me, you and I know what’s actually true. And it’s that AI is a calculator. It’s not the singularity. It’s a damn tool. Reasonable people know this.
什么时候觉得受不了了，就回我这儿来。因为你我心知肚明，什么是真正的真相：AI 就是个计算器，不是什么奇点，它只是个工具。理智的人都知道这一点。

And now you know why everyone around you is pretending that AI is the second coming of consciousness. It’s because they’re getting promoted. They’re keeping their jobs, and you’re not. Put on your turtleneck. And I’ll see you on the other side.
现在你也明白为什么你周围的所有人都在假装 AI 是意识的第二次降临了。因为他们在升职加薪，他们在保住工作，而你没有。穿上你的圆领衫，我们彼岸再见。

True resilience is not about bouncing back

Tue, 12 May 2026 23:10:00 +0800

Philip T Starks:

Endurance is getting through; recovery is getting better. Despite Holling’s distinction [that real ecological resilience wasn’t about snapping back, but about persistence], we have continued to confuse the two, and this confusion has consequences. Whereas endurance can be demanded of someone, recovery requires support. Glorifying endurance shifts attention away from systems and places expectations on to individuals.

You break down, repair, and rebuild something new. The challenge itself doesn’t make you stronger, your response does. No recovery, no gain. In the gym, we understand this. In life, we pretend the damage didn’t happen, and call it strength.

Note

Tue, 12 May 2026 20:39:00 +0800

电鸡（nvidia inside）

Note

Mon, 11 May 2026 20:32:30 +0800

米家用了两年跪了，换了个不能联网的 dumb 微波炉，神清气爽

Note

Sun, 10 May 2026 19:11:00 +0800

美妙

Photos

Sun, 10 May 2026 18:47:00 +0800

Most vibe-coded tools are not for you

Thu, 07 May 2026 05:29:06 +0800

Fabrizio Ferri Benedetti:

Tools created through AIs hurt my brain, I decided, because they’re often devoid of three fundamental qualities of good tools: reach, sociality, and finish.

In a way, thanks to AI, we’re making tools of ourselves (pun intended): much like we do when we create agentic skills out of interesting sessions, we condense our thoughts and intent into code. The resulting tool will carry our cognitive fingerprint: to use it effectively, you must be attuned to the way its creator thought, which is neither ideal nor fair. A tool with reach loosens that grip. It works for people who don’t think the way its maker does, or it grows until it can.

What makes tools social is a subtle combination of affordances, community participation, and the dialectic exchange between users and developers.

The lack of finish also means that tools are not thoughtfully designed, which makes their evolution improbable. A good tool can be refined, expanded, modularized, perhaps even merged with another tool. A thoughtful tool invites its own evolution; a careless one resists it. When you can’t tell where a tool’s boundaries are, it can’t grow.

Some of the best tools in history started as scratches-for-an-itch projects. They became universal because someone cared enough to finish and socialize them. For the stuff we create using LLMs to become tools, a README file and a catchy GitHub project name are not enough: We ought to ask ourselves whether what we’ve made can survive contact with someone else’s problem, and whether we’d trust it enough to maintain it.

BEWARE SOFTWARE BRAIN

Wed, 06 May 2026 15:32:25 +0800

Nilay Patel:

[E]veryone in tech understands how much regular people dislike AI. What I think they’re missing is why. They think this is a marketing problem.

[S]oftware brain [is] see[ing] the whole world as a series of databases that can be controlled with the structured language of software code.

Once you start seeing the world as a bunch of databases, it’s a small jump to feeling like you can control everything if you can just control the data.

It turns out software brain has a limit — the government isn’t software. People aren’t computers, and they don’t live in automatable loops that can be neatly captured in databases.

At some point, the database stops matching reality. At that point, we usually end up tweaking the database, not the world. But the AI industry has fully lost sight of this, because AI thrives on data. It’s just software, after all. And so the ask is for more and more of us to conform our lives to the database, not the other way around.

This intoxicating similarity between law and code trips people up all the time. People are constantly trying to issue commands to society at large like it’s a computer that will obey instructions.

But law isn’t actually code, and society and courts aren’t computers. […] You simply cannot take the facts of a case, the law as written, and predict the outcome of that case with any real certainty, even though the formality of the legal system makes people think it works like a computer — that it’s predictable.

But at the end of the day, it’s actually ambiguity that’s at the very heart of our legal system. It’s ambiguity that makes lawyers lawyers. Honestly, it’s ambiguity that makes people hate lawyers because it’s always possible to argue the other side, and it’s always possible to find the gray area in the law.

There’s real value in introducing AI to business, because so much of modern business is already software: collecting data, analyzing it, and taking action on it over and over again in a loop. Businesses also control their data, and they can demand that all their databases work together.

Most people are not collecting data about every single thing that they do. And if they’re collecting any at all, it’s stored across lots of different systems […] Those systems don’t talk to each other and maybe they never will, because there’s no reason for them to. Asking people to connect them all freaks them out.

[I]t is a failure when you ask people to adapt to computers. Computers should adapt to people. Asking people to make themselves more legible to software — to turn themselves into a database — is a doomed idea.

Not everything about our lives can be measured and automated and optimized, and it shouldn’t be.

And so the tech industry is rushing forward to put AI everywhere at enormous cost — energy, emissions, manufacturing capacity, the ability to buy RAM — and locked into the narrow framework of software brain without realizing they are also asking people to be fundamentally less human. They then sit around wondering why everyone hates them.

Photos

Tue, 05 May 2026 16:57:00 +0800

Note

Mon, 04 May 2026 19:51:00 +0800

黑心厂还没倒闭呢