About

At four, I held instruction cards and swiped them through the robot's card reader again and again: forward, turn left, turn right, stop. Back then, I had no idea what a program was, but I was already captivated by this magic: a few commands could control something else.

At seven, in Scratch, I designed rules by dragging colorful instruction blocks, filling in parameters, and characters moved according to my ideas. I gradually understood this wasn't magic—I designed the rules, and it interacted with me.

At ten, I filled the corners of my textbooks with stick figures and levels, designing sets of game strategies on scratch paper. But I didn't just want to draw them—I wanted them to actually move. So I started self-learning C language and game engines, continuously publishing over a dozen original games and music works on Luogu and Bilibili. I slowly discovered: programs don't just bring my world to life, they let others step into it too.

At thirteen, leftover apples, connected with wires and linked to programs, became a fruit piano. I taught myself music composition, writing dozens of symphonies for my games; I also adapted a math puzzle from "Super Brain" into a game that stumped the top math students in the grade. I gradually realized: music, mathematics, engineering, and code converge into creativity within programs.

At sixteen, I self-learned Unity and UE, completing my first 3D game. I participated in two 48-hour GameJams, responsible for programming. The themes gradually shifted beyond just fun—I started incorporating campus issues and anti-addiction topics into games. During summer break, I interned at TAL and NetEase on real projects. Later, my dad and I built "Cantonese Master" to help my sister learn Cantonese and adapt faster to life in Hong Kong. I finally realized: programs don't just make virtual worlds move, they can make real life a little better.

The speed of AI development and iteration is too fast to process. I worried AI would take away most people's jobs, that humanity would be coerced and held captive by it. I couldn't express this fear in code, but I expressed it through art—an English short story "Escaping AI," where I wrote lyrics, composed music, and edited it into a short film; in AP 3D class, I used discarded circuit boards, crystals, and wires to assemble a humanoid installation, materializing my concerns: as humans increasingly rely on AI, are we losing the vitality of our own thinking?

Fear shouldn't stop at criticism—I needed to approach and understand it. I started self-studying MIT's online course "Day of AI"; used spare time to learn and earn the AWS Certified AI Practitioner certificate. Though I haven't deeply studied AI principles, I gradually gained a new understanding: AI isn't a monster with its own will, but an understandable tool composed of data, algorithms, and models.

Books and certificates are just the starting point; I believe real answers should be found in real scenarios. At the National Brain Science Competition, I led a team researching the intersection of AI and psychology, winning the national second prize; during my internship at TAL, I participated in research and iteration of AI error-correction features; at NetEase, I explored how AI empowers oral language learning; at a gaming company internship, I started thinking about how to inject AI into the entire game development process. At this point, AI in my eyes became a tool for psychological research, educational assistance, and game creation engine. From fear to understanding, from understanding to application, AI's role is changing, and my relationship with it is evolving too.

I led the design of "Xueban"—a growth guidance system with 9 specialized AI agents working in coordination: RealtimePlanner orchestrates 5 types of ReplyAI in real-time, while 3 backend AIs continuously build long-term user profiles. Following the principle of "understand first, then act," it translates neuroscience stress mechanisms into product response strategies, replacing one-time answers with long-term user modeling. In 35阶段性 user tests, "not preaching" scored highest (4.06/5)—this is exactly what I want to do: use AI to liberate repetitive labor while preserving the uniquely human understanding and warmth.