Days slip by without you noticing, and just like that, Chinese New Year 2021 is around the corner. Looking back at the past year, it added yet another vivid stroke to an already colorful life.

A while ago I came across a multi-year diary online – available in 3, 5, and 10-year editions – where you can flip to the same date and see what you wrote in previous years. On a whim I bought the 5-year version. After writing in it for a while, I found it somewhat impractical: to fit five years on a single page, each day’s entry is limited to three or four brief sentences. It ended up gathering dust. Blogging is far more practical – no such constraints, room to write freely, and everything indexed by date.

GitHub Contribution

I glanced at the Contribution graph on my GitHub homepage and was surprised to find that 2020 was my most active year since I registered – 982 commits total, roughly 2.7 per day. The heatmap does show a sparse stretch from July through October, though. Work was simply too busy during those months to maintain my open-source projects, and I recognized at the time that this was a problem.

GitHub 2020 Contribution

Leaving Kuaishou

My time at Kuaishou was short, but it was relentlessly busy. From the moment I joined in late 2019, there was never a breather. The Lite edition had to develop new features while staying in sync with the main app. At first, product was cautious – Lite only carried a subset of the main app’s features. Then, as growth campaigns started showing results, the floodgates opened: everything the main app had, Lite had to have; things the main app didn’t have, Lite needed those too. Just like that, Lite lost its positioning entirely. After a few months of iteration, it was arguably more bloated than the main app. We used to joke that “Lite” didn’t mean “smaller and faster” – it meant “lite-speed releases” (the main app shipped biweekly; Lite shipped every week). By July I was building similar-content recommendations for Lite, and the periodic big sync with the main app had to continue. Each sync was excruciating – I had already been through two rounds, with the third about to begin.

The “big sync” was essentially a branch merge. Lite had originally been forked from the main app’s codebase, with both products living in the same repo and iterating independently. Since the main app’s engineering team was far larger, code divergence grew exponentially, and the app’s foundational architecture simply couldn’t support this kind of product structure. Syncing became a monumental challenge. From an engineer’s perspective, this was the kind of problem that should have been solved systematically as early as possible – the longer you wait, the higher the cost of each sync.

Maintaining multiple product lines from a single codebase was nothing new at Didi. It dates back to the The One project after the Didi-Kuaidi merger. To support rapid iteration across business lines, the entire company paused feature development for a month and restructured the app’s foundational architecture. That investment paid off enormously: in later projects like the Uber merger and internationalization, incubation time for new business lines went from months to weeks. That’s the productivity gain a solid architecture rebuild delivers.

As for why I left: overtime was never the issue. I regularly contributed to open-source projects until 2-3 AM without feeling drained. The real reason was that I could already see exactly what I’d look like three years down the road. Before I resigned, a friend urged me to hang on a few more months and vest some options first. I left anyway. When Kuaishou went public recently, many friends asked if I had regrets – I had missed out on a small fortune in just a few months. I said: “Money matters, but every day of your life should be spent in joy and hope.”

Turning Down Didi

After word got out that I had left, former colleagues at Didi reached out, hoping I’d come back to level up the mobile engineering org. After years of grinding through business demands, the foundational architecture needed an upgrade. By that point I had already accepted an offer from Coupang and wasn’t looking at other opportunities. But after several conversations, I started to waver – after all those years at Didi, there were real attachments. It was a tough call, until I faced the soul-searching question: “What do you actually want?”

What do I actually want? Title? Options and stock? None of those seemed like what I wanted most. After more than a decade in the industry, for every decision I make, I always ask myself: “Is this the right thing to do? If not, what would be?”

When you stay in one place too long, hearing the same voice for too long, you develop an illusion that the whole world works the same way. You grow accustomed to it, even internalize it as values. I realized what I wanted most was different voices, different worlds, different values – to learn and grow through the collisions. As for titles and stock, if you genuinely create value for the world, those things come eventually. Chase them directly, and they tend to slip away.

Joining Coupang

In China’s tech industry, foreign companies used to be the hot ticket. Then domestic giants rose, and everyone pivoted to homegrown companies for career prestige. Foreign companies barely registered – maybe still somewhat in Shanghai, but definitely not in Beijing. I had spent time at SAMSUNG years earlier, and there were things about working at a foreign company that left a deep impression.

Culture and Values

I remember being sent to training in the first month at SAMSUNG. A group of new hires traveled to a hotel on the outskirts of Nanjing (where headquarters was located) for a week of classes on the company’s history and culture, team activities, and talent shows. Working with those colleagues was genuinely fun. Every week there was a Happy Hour – Korean colleagues would make sushi for everyone. I even showed off once by making pumpkin cakes. Even years after leaving Shanghai, every time I visit I can still get the old crew together for a night out. Coupang has Happy Hour too – celebrating birthdays, chatting about light topics. At Coupang, the most frequently mentioned values are:

  • Wow the Customer
  • Deep Dive
  • Think Systematically
  • Hire & Develop the Best
  • Aim High and Find a Way
  • ……

From Coupang Leadership Principles

Security

The most unbearable part of SAMSUNG was the security checks. Every time you entered or left the office, you passed through security gates – practically airport-level screening. Guards would scan you with handheld devices. Since the restrooms were outside the secure zone, going to the bathroom or stepping out for a smoke meant another round of screening. Some colleagues joked about whether getting scanned that many times a day might affect fertility.

Beyond physical security, there was information security. The company network could access Google and GitHub, but almost everything else – messaging apps like QQ and WeChat – was blocked. Going to work felt like entering a sealed-off world. Besides your phone, no personal electronics were allowed in the office. I once tried to bring an iPad and was flatly stopped by security. Even company-issued notepads had to be collected and returned. And that was just the office – the factories were even stricter. You had to let them put a tamper seal over your phone’s camera before entering. If the seal showed any sign of loosening when you left, you were in serious trouble.

Foreign companies seem to take information security far more seriously than domestic ones, which reflects how much intellectual property matters. In China, chatting on QQ or WeChat on your work computer is completely normal. At foreign companies, work is work and personal is personal – don’t mix private property with company property.

At Coupang, while not as strict as SAMSUNG, information security is equally valued. The office network and personal device network are separate. Fortunately, my time at SAMSUNG had prepared me, otherwise the adjustment might have been rough. I also made some changes to my work habits. I used to nap for an hour or more after lunch, then start working – showing up late in the morning and leaving late at night. Now, to stay focused, I skipped the nap entirely, powered through my daily tasks, and left on time instead of dragging until 9-10 PM.

Hiring

My daily work boils down to four things: hiring, meetings, coding, and writing docs. For a fast-growing team, hiring is paramount, and Coupang is extremely rigorous about it – not just for candidates, but for interviewers too. Each interviewer focuses on a different dimension for every candidate: Round 1 is algorithms, Round 2 is engineering-oriented coding, Round 3 is advanced coding, Round 4 is system design. Each interviewer stays in their lane and designs questions appropriate to the role level. After each interview, a detailed evaluation is required:

Question: (Description)

  • ……

Expectation: (What we expect from the candidate)

  1. …
  2. …
  3. …

Solution: (Candidate’s approach)

  1. …
  2. …
  3. …

Pros (Strengths of the candidate’s approach)

  1. …
  2. …
  3. …

Cons (Weaknesses of the candidate’s approach)

  1. …
  2. …
  3. …

Gap to Next Level (Distance to the next role level)

  1. …
  2. …
  3. …

Each round lasts an hour, and writing the evaluation takes another hour – all in English, like composing a short essay. The evaluation has to be as objective and factual as possible. You can’t slap on vague labels like “weak Java fundamentals” or “poor Android knowledge.” You have to state exactly which question they couldn’t answer. The whole hiring process differs from many domestic companies. In most Chinese tech firms, if one interviewer says no, the candidate doesn’t advance. At Coupang, each round is independent – the next interviewer doesn’t see the previous one’s evaluation before the interview begins. This ensures that a single negative review doesn’t prematurely kill the pipeline. The process only terminates early if more than half the interviewers give a NO.

Meetings

Meetings at Coupang are not casual affairs. If you regularly need to discuss with people across countries and languages, you have to do a lot of prep. Documents are a must. If people from other countries are involved, you book simultaneous interpreters (Chinese-Korean, Chinese-English), and naturally, the documents need to be in English.

I also picked up a new skill here: taking meeting notes in English in real time while speaking, ensuring everyone stays on the same page. You can’t rely entirely on interpreters – after all, the discussions are highly technical, and the interpreters may not convey everything perfectly. For complex topics, I’d brief the interpreter beforehand and go over key jargon and abbreviations. Once, an interpreter asked me to explain what a “reference” was. After a long explanation, she quietly asked: “So it’s like a pointer?” I was floored that she knew what a pointer was – would have saved a lot of breath.

Documentation

Beyond technical discussions, there are architecture reviews attended by members of the architecture committee. Before your review, you need to prepare thoroughly – the most important thing being the document, written in a standardized format. For developers, the time spent writing docs nearly rivals coding time. This stands in stark contrast to many domestic companies where no documentation exists at all – everything is passed down by word of mouth. Beyond technical docs, there’s another critical document: the Onboarding guide. Every team maintains its own, ensuring knowledge is preserved and doesn’t vanish when someone leaves.

Coding

Domestic companies tend to be relatively lax about code quality management – usually just Check Style and a quick Code Review. At Coupang, code quality is taken very seriously. Before a Code Review even starts, the change must pass the CI pipeline – SonarQube, Android/Swift Lint, and more. Only after the pipeline passes will other engineers review the business logic. After review, automated tests run, and only if everything passes does the code get merged.

Performance Reviews

Coupang‘s performance evaluations are grounded in Leadership Principles. Only items covered by the Leadership Principles are evaluated – nothing outside that scope. Whether it’s self-assessment or peer review, everything must be expressed in SBI (Situation-Behavior-Impact) format, aiming for objectivity and factual grounding. Peer reviews work this way, and so do manager-to-report reviews.

Overall, well-defined processes are what distinguish Coupang from many domestic tech giants. Then there’s the engineering culture: refusing to sacrifice product quality to hit deadlines. After all, when something breaks in production, it’s the engineers who take the blame and spend time running incident reviews – and incident reviews are no walk in the park either. Better to nip that behavior in the bud.