2025 Fall Rhythm

This quarter, I got into a better rhythm for work while still having enough time to relax and have fun. I hope to continue this rhythm into next quarter too.

Work

As the RL infra at Meta matured, I took on more tasks and I wanted to push myself to work more. I have been inspired by the work ethic of Chinese people at Meta. I can tell that everyone is devoted to the work, and that most people continue working after they go home. There is one infra engineer that is always in the office, and most nights when I went home, I would say goodnight to him on the way out he would still be working! That is a different level.

I got into a rhythm of working 40+ hour weeks for most of the quarter (for external readers, this 40 hours doesn’t include meetings, eating, random hallway discussions). This basically looked like being in the office until 11pm Monday through Thursday, and working for around half of the time on the other days (leaving some time to exercise and meet friends). I would push myself a bit even when I was tired to hit the hours, and I realized that at the end of the week it would get harder to focus and I probably was pushing the limit of sustainable, productive work. The main difference was to make working the default mode of operation and not spend much time on other things, like browsing the internet, exercise, or attending social gatherings. Going to the office and having a bunch of things I wanted to get done also helped obviously.

Overall, for me this was a pretty big improvement in productivity. The last time I tried to work this hard was during my AI residency, and I was working more this quarter than the best weeks during my AI residency! I think the main thing is that working in the office reduces a lot of distractions. I think I “reward hacked” this a bit though, because I was working a lot and being relatively productive, but I was probably not spending as much time thinking about whether I was being the most efficient. Perhaps a better way of working is to be purely “outcome-oriented”, where you just focus on whatever moves the needle, and not necessarily the number of hours worked or how much you concentrated.

I realized that I probably overindexed on Cal Newport’s concept of “deep work”. Deep work is most valid when you’re doing something that (1) is at the frontier of your intellectual productivity, (2) is non-super collaborative, and (3) doesn’t have much waiting involved. So it was appropriate the work that Cal Newport himself was doing: writing and theoretical computer science research. However, these assumptions are basically untrue in today’s framework of AI. First, most of the work in Ai research is execution and doesn’t require novel thinking. Obviously you still want your execution to be efficient and high quality, but the majority of it does not require periods of long, concentrated thinking. Second, I am working on many projects at once that depend on other people, and so it is unavoidable to incur a lot of cost in context switching. Finally, coding centered around AI doesn’t require the same cognitive load, because AI takes care of a lot of things that typically require human attention. And you have to wait for the AI to do work. The more important bit now is checking that the work is correct, and not having to produce all the code yourself anymore.

Health

For exercise, I mostly maintained my lifts without making much progress. It was good to lift with Tianfu once a week. The biggest revelation from lifting with him is that I stopped lifting hard. After a few weeks, I asked him why I wasn’t improving and he said we were actually training pretty casually. So we are trying to increase the intensity moving forward. For cardio, I played volleyball with Rishi, badminton with Kevin and Yi, and tennis with Suraj.

I got into the habit of eating Salad everyday, which is a pretty good diet and saves time and thinking. The one thing I’m concerned about is the amount of beef that I eat with the salad; my TMJ clicking started to get worse again, and I’m not sure if chewing beef is the issue or if inflammation from red meat is the issue.

Left, badminton with Yi; middle, tennis with Suraj; right, I ran up the lyon street steps once.

Life

AI race v1

I hosted an event that I called “race to AGI”. It was a larger scale follow-up to the lifting event I had last quarter with 7 people. Around 25 people came to the Panhandle and I set up a few events for four teams: max pull ups, tug of war, relay race, and then capture the flag. We had a Meta team, OpenAI team, Stanford team, and “everyone else” team. The Stanford and “everyone else” team ended up tying for first place, so they competed to see who could do 200 push ups first. For the tug of war, I got a super legit rope and gloves (I was worried about the rope breaking). This was probably the most unique event and I think people liked it the most.

Before hosting these events, I always get nervous and excited. But I am glad I did it, and plan to do more. The revelation is that I can host any event that you want, and doing something unique is probably more memorable than just a regular party or brunch. Keeping a child-like energy, and reliving the days of being a kid is so fun for me.

Race to AGI v1 at Panhandle.

Jae

My high school friend Jae came to visit, which was lovely. Jae is interested in starting a company and eventually moving back to SF, and it was great to be able to host him in that process. Jae has always been creative, introspective, and aesthetic and by spending time together I’ve learned a bit more about those things. One of my favorite qualities of Jae is that he has a disarming demeanor and can be emotionally open and vulnerable, which also brings out those qualities in myself, allowed me to also explore myself deeper. It was also a good exercise for me to live vicariously through his experience building a company from scratch, as one day I also plan to start a company and it’s good to experience that a little bit through a close friend.

Left, at Stanford EVGR bar; middle, Vasara hot tub; right, Halcyon night club.

Halloween

This is the first Halloween that I really made an effort to dress up. Jae had the great idea of being Saja boys, and I ordered a bunch of tattoos and our friend Megan loaned us her bright cardigans. I brought the tattoos into work and a few people joined in with us as Saja boys! Then we went to a Kpop demon hunters sing along movie and a few SF house parties. We ended the night eating Korean friend chicken in SF.

Left, with Jae at a SF house party; middle, Kpop demon hunters sing along; right, with coworkers at Meta.

San Francisco

The funny thing about SF is that I only really started to appreciate the city once I had moved out. In the past, I didn’t like the mission much and was too frugal (in time and money) to take ubers to other places, so I didn’t explore much. This time, without those constraints I spent time exploring Pac Heights, as I wanted to get to know the best neighborhood in SF. Yifei was also nice enough to let me stay at his place on Friday evenings, and I enjoyed using the hot tub and Sauna there with Yifei and Kevin. The conclusion that I came to is that San Francisco can be a good place to live, if you’re rich.

Left, sleeping at my old bedroom at Yifei’s; middle, hot tub at Canyon; right, Karoake with Yifei and Jae.

Left, with Dad and Jerry at Elephant Sushi; middle, celebrating Jerry’s promotion at Chīsai Sushi Club; right, hanging out with Jae and Kevin on thanksgiving.

Building around time

One thing that I have been trying to be better about is building around time as the limiting factor instead of money. I had some frugal habits growing up, which was the struggle of our parents’ generation. But I think saving time and investing in being able to work better is the most effective approach. I don’t think I spend too much by any means, but here are some examples of old and new habits I’ve taken on:

• Just ordering DoorDash on weekends instead of trying to bring extra food from the company lol (I used to do this at Google)
• Taking Uber because it’s less stressful than biking to work
• Ordering 30 socks so I don’t have to worry about running out and doing laundry at inconvenient times (strictly better)
• Buying a bunch of protein shake cups so I can use the dishwasher

Conclusions

Writing this from the comfort of my Mom’s home in Plano, I’m recharged and ready for a new year! Have a lot on my bucket list for the new year: