Leaving Microsoft

Essay

I’ve handed in my resignation at Microsoft. There are still two weeks before my last day, but honestly, I don’t feel like waiting. So here are a few notes.

I once described my three and a half years at Microsoft as a painting made from acrylic that had dried solid: bright, stable, and completely unchanging.

Microsoft is, without question, an excellent employer: one of the best benefits packages in China, a comfortable office environment, and absurdly good WLB. The job itself was also relatively “easy” for me. I could finish everything efficiently in the way I knew best, barely break a sweat, and tick off each year’s KPIs. In terms of bandwidth, I could juggle my schedule at Tsinghua, my side projects, and everything else without much pressure.

Ever since the last desk move, I had been sitting in the corner of the sixth floor for two full years. That cubicle had long since molded itself around me: behind me was a huge 🐲 head I had drawn by hand; gym clothes hung beside it, and the wall was covered in sweat stains. The neighboring desk was piled with empty liquor bottles, protein powder, and assorted workout gear. My desk was lined with monitors, alongside the little fan that had followed me all the way from Alipay and a collection of inexplicable trinkets picked up from company events. Underneath was a workstation assembled from IT’s electronic junk and a pair of my wrecked slippers. My desk sign still had the drinks menu from our last “drunkard challenge” tucked into it.

I was far too comfortable in that cubicle. Every day I sipped coffee, munched on afternoon tea, listened to music, and drew design screens. I could have coasted there until the end of the world.

Sounds great, right? Of course it does. That’s also how a frog dies in warm water.

My product lead, level 67, was an “old-timer” who pecked at the keyboard with two fingers. Whenever I went out to eat with business colleagues, the conversation was about school-district apartments and middle-school admissions. The fitness group checked in enthusiastically every day, but one look at the actual business and we were doing exactly the same things as two years ago. No one could say people weren’t working hard. Yet somehow, no one seemed to care either.

“Isn’t that the best possible WLB? What are you complaining about?” Sorry, but to me, WLB in the context of salaried employment is itself a false proposition.

The logic is simple: judging a job by its WLB is still employee thinking. If you’re going to work for someone else, you should ask how that particular job helps you reach your long-term goals. Unfortunately, by most of the standards that matter to me, this job offered very little help. The goal of WLB is balance, and of course I want balance too. But when that day comes, I want it to be something I created for myself—not something granted by a benefits-heavy retirement home run by some Silicon Valley Indian guy, one that might pull out of China at any moment. (As luck would have it, Microsoft announced yet another round of layoffs in China today. 楽.)

As for the platform itself, Microsoft Research Asia may have been a cradle of talent for China’s internet industry at the beginning of the twenty-first century. Traces of Microsoft can be found in the early histories of many Chinese tech companies. But more than twenty years later, the product and engineering organization in China has become conservative, ambiguous, and peripheral. Our self-mocking description of ourselves was “the outsourced team of the outsourced team of the outsourced team of the outsourced team in America.” My direct manager also told me more than once that working at Microsoft now meant living with constant “ambiguity.” For me, that was a disaster.

At the industry level, Microsoft has delivered little that stands out in this wave of AI—not at the chip layer, not at the model layer, and not at the application layer. Mustafa, the EVP who just got the boot at MAI, seemed to do little besides sending one company-wide email every month. (Turns out being an EVP is pretty easy.)

As for me, losing enthusiasm was a very slow process. There was no decisive moment. Maybe it was one AMA email after another from LT; one feature after another withdrawn because of compliance requirements; one 1-on-1 and review after another. Anyway. Anyway. Time kills everything.

Still, I’m grateful to Microsoft. These three and a half years were a rare period of calm in my life, and I became much healthier and more mature, both physically and mentally. Its values also shaped me deeply. Empowerment is a word with real weight. I don’t know whether I have the ability to empower anyone else, but even without the ability, one should still have a dream or two.

So I’m going to try dreaming big. I’ll dream first, fall hard, and deal with the pain afterward.

I’ll be heading south soon. In 2022, the moment the lockdown ended, I scrambled back to Beijing. At the time, I truly never imagined I would leave again. But life is unpredictable. Who knows where I’ll be tomorrow—or whether what comes next will be a surprise or an accident?

June 2026
Desk 6C002, Building 2, Microsoft Beijing