This happens to coincide with the reading club's anniversary event, so I will write down the past year — what happened, and what I took from it.
This time last year, I was at home, cramming for the postgraduate entrance exam.
The university library was closed for renovation, so I simply moved my desktop machine from school back home. Days were for studying; in the evenings I opened VRChat and chatted with friends from all over the world. Oddly enough, my spoken English improved faster in that period than during exam prep — language, it seems, only really moves when you actually want to talk to someone.
In November the library reopened and I went back to prepare. That was the most efficient stretch I have ever had. On exam day I felt calm; on the evening after the major paper, I walked across the snow-covered campus one last time and felt nothing but quiet.
One failed attempt, and one overnight train
On the day results came out, I had missed the interview cutoff for my first choice. Honestly, it was not a surprise — I am genuinely not good at exams.
Then came the round of emails to supervisors and waiting for replies. A school I had prepared well for fell through because of a policy change, and I missed one round of transfers. Later I received an interview invitation from another city, bought an overnight high-speed rail ticket and went — written test in the morning and interview in the afternoon the next day, with two online interviews squeezed in before and after.
Just before the noon deadline the next day, I clicked confirm on the admission offer. The major is Mechanical Engineering — a properly orthodox name, and convenient when introducing myself to foreigners. Back home I emailed the supervisor, we got along well, and that is how I joined the group.
Beijing: the first time what I made actually worked
At the end of April I interviewed at a robotics startup in Beijing, got through, and started right after the May Day holiday.
Everyone in my intern cohort had a much prettier academic record than mine, and for a while I felt quite inferior. But once I actually started working, I found that whether you can bring a board to life and chase a problem to the end has far less to do with your degree than I had assumed. In those months I built up a good deal of hardware and algorithm experience in high-power brushless motor drives, and — more importantly — perspective: at the end of August, before the internship ended, I went to the World Robot Conference; and because of work I got into the joint-motor production and testing area of a humanoid robotics company, and saw the real production lines and test benches behind the phrase "embodied intelligence".
The biggest gain from that internship was not an extra line on my CV, but learning for the first time the difference between "finished" and "working": finished means you handed the task over; working means it actually turns.
Now: days at the workbench
In September I enrolled and joined the research group. My supervisor is very decent — no clock-in, no mandatory group meetings, just move your work forward at your own pace. That was a great relief. It is genuinely busy, but when I sink into the work I get a bit of solid satisfaction.
The anxiety this year has clearly decreased. Not because there is less to do, but because I finally have something I can push forward a little every day.
Deeper dive: how I handle anxiety
Most of my anxiety used to come from thinking — and what I thought about was usually unverified inference: will the job market be terrible, will degrees keep devaluing, is this the right path. Later I gave myself a dumb method:
- Write the vague worry down and split it into a few questions that can actually be checked;
- Check what can be checked; mark the rest "unknown" and stop chewing on it;
- Keep only one or two smallest actions per day — finishing them counts as a win.
There is nothing profound in this, but it has pulled me out of idle spinning a few times.
As I write this, I am still putting AI to work. It burns something like three hundred million tokens a day, all to polish one demo, so I can go after funding for this project. With funding I can let go and build robots at real scale — that thought is genuinely exciting.
Three things I learned this year
The second is that the body is more honest than the mood. I pulled one all-nighter to hit a deadline, and the next two days were basically written off; since then I have drawn a shutdown line — if I really do stay up, the next day is resignedly a no-task day.
The third: you only really understand something when you can explain it so others understand it. During the internship I put together a set of drive-related tutorials and documents, and I got more out of writing them than out of reading them.
This time last year I thought the fork in life was an exam. Now I know the fork is there every day; most of the time nobody tells you that is what it is.
From waiting for results to making things took me one year. It was not a smooth year, but every part of it counts.
One more thing: if I win a prize in the reading club's year-in-review event, I will convert it into DeepSeek tokens for my own use — running agents every day does cost real money, and that is probably the most practical way to spend it.
Comments
Loading comments...