It found an award from more than a decade ago, then entered me in a competition I had never attended.
Xian Weijia · Original draft 12 Mar 2025 · Adapted 8 Nov 2026 · 7 min read
“Trust me. Nobody in the world understands Xian Weijia better than I do.”
Xian Weijia
The main problem with this statement is that Xian Weijia is also the person making it. A conflict of interest. Enjoy it accordingly.
In March 2025, I asked AI to look me up. It dug out old records I had not managed to find on Baidu, including my first prize in the Hope Cup mathematics competition in 2012. An award more than a decade old suddenly reappeared on my screen.
Impressive. My personal publicity department had finally hired someone who could search the archives.
Then I kept reading. The new employee had ideas of its own: it left out things that had happened and supplied things that had not. The real Xian Weijia sat there, checking another Xian Weijia’s life, line by line.
1. First, a few additions to the résumé
AI found the Hope Cup result but missed how I entered Chengdu Experimental Foreign Languages School through the “Campus Times” examination. I came first in mathematics. Hold the applause: this was not the school’s own independent admission exam. Fewer people took it, and there was less competition.
There was also the Huasai competition: 75 points and a second prize in 2011, then another second prize in 2012. As I remember it, a second prize or above in 2011 brought an invitation to take the selection exam for the youth palace’s “Bullfrog Class,” a class for strong students.
I went.
Ridiculously hard. Who was taking these exams? I finished near the bottom.
The publicity department could turn this into “multiple awards and an invitation to a selective examination.” Smooth reading. The near-bottom finish disappears. Every sentence might be true, yet the person they assemble is looking suspiciously impressive.
The next episode should restore a normal body temperature to this résumé.
2. The fifth postulate. A lasting pain.
In 2018, I took part in the independent admissions assessment for mathematics at Sichuan University. I did not earn the bonus points or end up in the department. This is how I remember the exchange:
Professor: What do you usually read?
Me: Euclid’s Elements.
Professor: What is the fifth postulate?
Me: …
An explanation: I had a feeling this professor taught geometry, so I offered that title. Whether my guess was right, I do not know. The answer was already out.
I responded with silence.
And so Sichuan University’s mathematics department suffered the loss of Xian Weijia.
Please never ask me about the fifth postulate again. A lasting pain.
AI did not know this conversation. It could retrieve the award but not the few seconds in which I had no answer. The internet preserved “good at mathematics,” but not “consider the follow-up question before naming a book.” The searchable version of me was, conveniently, a little more accomplished than the actual one.
Well, now I have written the story down myself and supplied the missing training material. Future publicity work may become slightly harder.
3. My pretraining was fortunate
I was fairly good at mathematics as a child. My own estimate put me somewhere in the top few per thousand. Please note: my estimate. Xian Weijia grading Xian Weijia. The conflict of interest continues.
A large part of it was luck.
I had books such as Ju Yi Fan San, a mathematics practice series, and teachers who took their work seriously. The teachers who taught me mathematics as a child, and those who looked after my learning in high school, deserve thanks. Many problems I could solve were ones they had patiently taught me to approach.
My middle-school teacher deserves thanks too. Full marks, though, I am reluctant to give: I liked fooling around with girls, kept getting caught, and lost my lunchtime nap to standing as a punishment.
Whatever happened to boys help boys???
Later, I noticed a similarity between training a person and training AI, at least as a joke: good textbooks and teachers first, then a beating from the real world. Pretraining, followed by reinforcement learning.
It is only an analogy. People remember who taught them, and who deprived them of a nap. Gratitude and a grudge can coexist. Models do not yet have to negotiate quite this kind of teacher–student relationship.
Back to the résumé. Keep only the awards and the teachers, books and opportunities disappear. The achievements begin to look as though they sprang from my brain unaided. This version benefits Xian Weijia. Xian Weijia is highly suspicious of it, and a little reluctant to delete it.
4. It found me, then wrote a sequel
Omissions are manageable. Add what is missing. The trouble was that AI did not seem satisfied with the life already available.
It mentioned a paper of mine from 2022 but got the timing wrong. I wrote it after graduating, for a master’s application, not during my undergraduate degree. In my original draft, my own assessment of the paper was brief and unflattering: filler.
It also said I had entered the Mathematical Contest in Modeling, MCM, and written a paper about tennis.
I had done neither.
So, the tennis-playing Xian Weijia: where are you? Could you come and claim your paper?
I genuinely wondered whether there was someone else with my name. I do not know where the system went wrong, so I cannot turn mistaken identity into another certain answer. What I do know is that these were not my experiences.
That was the difficult part. The Hope Cup record was real, but the text did not change fonts when it entered fiction. The correct material earned my trust; the mistakes carried on in the same voice.
If it had been nonsense throughout, I could have closed it. Instead, it first got a few obscure facts right, then calmly sent me to a competition I had never entered.
I can personally correct a story about Xian Weijia. What about someone I do not know? I would probably find that story just as easy to read.
5. Perhaps some blanks can remain
At the time, I thought people still held things AI could not access: private conversations, stories passed by word of mouth, publications that had never been digitized. Those seconds of silence over the fifth postulate belonged in that category.
My original draft called them “data that cannot be quantified.” More precisely, some had not been recorded; some were recorded but not public; and some would lose the relationship and tone around them even in a recording.
Recording devices and camera-equipped glasses, however, make it harder for an encounter to leave no record. Being recorded, being searchable and being accessible to a particular model are different things. Still, once a record exists, who keeps it, who sees it and how much later it can be retrieved become concrete questions.
Something said over dinner may originally have been intended to last only until dinner ended. Turn it into searchable text and its lifespan grows. Its audience may change too.
I do want AI to understand me better. I am less keen to turn every casual conversation into an oral-history interview for a future archivist.
Enough. AI is dancing on the boundary. 🕺
So what do we do?
6. Criticism finished. Check the delivery.
At this point, I should disclose the postscript to the original draft.
In May 2024, I bought a Limitless AI pendant. When I wrote the piece in March 2025, I was waiting for it to ship and asking whether a friend in the United States could receive it and bring it back to China.
Yes. One paragraph criticized AI for recording our lives. The next sought a delivery address for an AI device that records life.
Better leave that part in. Delete it and I become a thoroughly principled technology critic, as suspiciously tidy as the excellent résumé without the fifth postulate.
Having complaints about AI and wanting to try what it can do have always been able to sit at the same table with me. Even waiting for delivery need not interrupt the criticism.
So I am keeping the opening line:
“Nobody in the world understands Xian Weijia better than I do.”
For now, at least, I still remember which competitions I never entered, which paper I never wrote, and which admissions interview is best left unmentioned.
As for the fifth postulate, look it up yourself. Do not ask me.