第一章：凌晨三点的推送：一个普通人的数字生活切片

凌晨三点的北京，CBD的灯火还剩不到三成亮着，28岁的互联网运营张明被枕边的手机屏幕晃醒。屏幕上是一条来自京东的推送：“您关注的索尼WH-1000XM5无线降噪耳机今日直降200元，限时24小时。”

张明盯着屏幕愣了十秒。他上周确实在京东刷到过这款耳机，当时加了购物车，后来觉得价格有点高就搁置了，但他这周一次都没打开过京东，甚至昨天才刚清了一次浏览记录。他是跟同事随口提了一句“最近想换个降噪耳机，通勤太吵”，难道手机连他的聊天记录都监控？

他划掉推送，翻到上周的浏览记录，页面停留在耳机的商品详情页，停留时间只有12秒，他甚至没点进参数页，更没加购——直到刚才那条推送跳出来之前，他早就忘了自己看过这款耳机的事。

这已经不是他第一次遇到“被猜中”的事了。过去一个月，他刚说想喝冰美式，楼下的瑞幸就给他发了5元优惠券；他刚在淘宝搜了“露营垫”，抖音首页就刷到了三个露营装备的测评视频；他甚至没告诉任何人自己最近肠胃不好，外卖平台就给他推了好几家清淡饮食的店。他以前总觉得这是“算法懂我”，是数字时代给的便利，直到今天凌晨这条耳机推送跳出来的时候，他突然觉得后背有点发凉：这些算法，到底是怎么“懂”他的？

为了搞清楚这件事，他回忆了昨天一整天的生活，把那些他以前根本没注意到的、细碎到可以忽略的“数字痕迹”一一翻了出来，数到最后，他自己都吓了一跳：过去24小时里，他的手机、电脑、甚至线下消费的刷卡记录，一共被12个不同的系统采集了数据，而这些数据最终流向了至少6个他根本没听说过的第三方公司。

---

At 3 a.m. in Beijing, less than 30% of the lights in the Central Business District (CBD) are still on. Zhang Ming, a 28-year-old internet operations specialist, is woken by the glow of his phone on the pillow. The screen shows a push notification from JD.com: "The Sony WH-1000XM5 wireless noise-canceling headphones you followed are now 200 yuan off, limited to 24 hours."

Zhang Ming stares at the screen for ten seconds. He did come across these headphones on JD.com last week, added them to his cart, but put them aside when he thought the price was a bit high. He hasn't opened JD.com once this week, and he even cleared his browsing history yesterday. He only mentioned to a colleague the day before that he wanted to get noise-canceling headphones because his commute was too noisy — could his phone be monitoring his chat records too?

He swipes away the notification and flips to last week's browsing history, which stops at the product detail page of the headphones, with a stay time of only 12 seconds. He didn't even click into the parameter page, let alone add it to the cart

张明盯着那条索尼耳机的推送看了足足三分钟，直到手机屏幕自动暗下去，他才猛地坐起身，把床头灯按亮。他翻出自己昨天一天的使用记录，把每一个曾被他忽略的细碎操作一一列在备忘录里，每写一条，后背的凉意就重一分。

---

Zhang Ming stares at that Sony headphone push notification for a full three minutes, until the phone screen goes dark automatically. He then sits up abruptly and turns on the bedside lamp, pulling up his app usage records from yesterday and listing every trivial operation he had overlooked in his notes app. With each item he writes, the chill running down his spine grows heavier.

7:15 AM：被闹钟吵醒的早晨，从导航采集位置开始

The 7:15 a.m. alarm jolts Zhang Ming awake, just as it has every workday for the past three years. His first move, before he even gets out of bed, is to open Amap (the Chinese equivalent of Google Maps) to check the morning commute. He types in his office address as he brushes his teeth, and the app instantly pushes a notification: "The 10-minute walk from Jiandemen station to your office is currently 20% more crowded than usual; we recommend leaving 5 minutes earlier."

This single action triggers the first data collection of the day. Amap collects not just his current location, but his historical commute route, his usual departure time, and even his phone’s gyroscope data to confirm he is walking toward the subway, not driving. According to Amap’s privacy policy (which 99% of users never read), this data is not only used for navigation services, but also shared with its advertising alliance partners, including e-commerce platforms, coffee chains, and local restaurants.

A small pop-up appears in the bottom corner of the screen (marked as a *sidebar note for digital edition readers*: Sidebar: What is an ad alliance? Think of it as a digital "logistics hub" where data from different internet companies is matched, tagged, and sent back to relevant advertisers. You never see this hub, but your data passes through it every day.)

Zhang Ming swipes the pop-up away without a second thought, just as he has done a thousand times before. He doesn’t know that the "internet operations specialist, 25-30 years old, lives in Chaoyang District, commutes via Line 10 subway" tag has already been added to his profile in the ad alliance’s database.

---

7:45 AM：地铁上的20分钟，12次无意识的点击

By 7:45 a.m., Zhang Ming is squeezed into the Line 10 subway, holding his phone with one hand and the overhead rail with the other. He opens Douyin (TikTok’s Chinese counterpart) to kill time, scrolling through the feed for 20 minutes until he gets to his stop.

He watches three videos about workplace skills, two reviews of new smartphones, one video of a chef making braised chicken rice, and one random clip of a guy fishing by a river. He likes the braised chicken video, leaves a comment saying "I ate this three times last week," and swipes past the fishing video in 0.8 seconds.

Each of these actions is recorded. Douyin’s algorithm notes that he stayed on the braised chicken video for 8 seconds, 2 seconds longer than the average user, and that he left a comment, so it tags him as "high interest in braised chicken rice, recent consumption history of the dish." The fishing video, which he swiped past in less than a second, is still recorded as a "negative feedback" signal, to avoid showing him similar content later.

What he doesn’t know is that Douyin’s ad system has already matched his profile with Meituan’s (the leading Chinese food delivery platform) ad target list. Meituan pays Douyin to push food delivery coupons to users who have shown interest in braised chicken rice in the past week, so Zhang Ming will see a Meituan coupon for his regular braised chicken restaurant when he opens the app at noon.

---

12:15 PM：午休的外卖，算法帮你做了选择

At noon, Zhang Ming sits down at his desk to order lunch. He was originally going to order the mala tang (spicy hot pot) he ate two days ago, but the first store that pops up on Meituan’s homepage is the braised chicken restaurant he visited three times last week, with a 5-yuan discount coupon attached.

He hesitates for two seconds, then clicks the braised chicken store. "It’s cheaper anyway," he tells himself, not realizing that the mala tang store he wanted was pushed to the second page of the search results, because the algorithm calculated that he had a 78% chance of ordering braised chicken today, based on his Douyin viewing history, his past order records, and even the fact that he searched for "braised chicken recipe" on Baidu last night.

This is the first time he notices the algorithm’s hand in his choice, but he doesn’t think much of it. After all, the food is good, the discount is real, and it saves him the trouble of scrolling through dozens of stores. He pays with his Ant Group’s Alipay, and the transaction record is instantly uploaded to Ant Group’s server, adding another tag to his profile: "spends 25-35 yuan on lunch on workdays, prefers meat dishes."

---

2:30 PM：摸鱼的两小时，数据采集的密集期

After lunch, Zhang Ming has two hours of free time before the afternoon meeting. He first opens BOSS Zhipin (a leading Chinese job recruitment platform) to check for new job opportunities, as he has been considering switching jobs for the past month. He updates his resume, adding a new project he completed last quarter, and browses three job postings for operations specialist roles with a salary of 18-22k per month.

Then he opens Maimai (a Chinese professional social networking platform similar to LinkedIn) and searches for "internet industry layoffs 2024," staying on the search results page for 5 minutes to read a few posts about layoffs at major tech companies. He closes Maimai and opens Baidu to search for "how to relieve lumbar strain from sitting too long," because his lower back has been hurting for the past two days.

Each of these actions is collected by multiple platforms. BOSS Zhipin records his resume update and job browsing history, tagging him as "operations specialist with 3 years of experience, actively looking for new jobs, expected salary 15-20k." Maimai’s search record is shared with BOSS Zhipin via a data cooperation agreement, so BOSS Zhipin knows he is concerned about layoffs, and will push him more "stable, well-established company" job postings later. The Baidu search for lumbar strain relief is s