On the stage at Harvard University, as the giant screen lit up with the presentation of the Xiaomeng AI project, Seven took a brief, steady breath. Before him sat young innovators from around the world, leading university professors and expert judges. Behind him was the work of nearly two years—condensed into this single moment, shared with his team.

At that instant, Seven knew clearly that what he had brought to the stage was not merely an AI system, but a response from a young generation in China: when technology becomes powerful enough, are we willing to take responsibility for others?

That response ultimately set Seven and his team apart in the CTB (China Thinks Big – Global Innovation Research Challenge), where they emerged from more than 4,000 teams nationwide to win the highest national award. At the subsequent global finals hosted at Harvard University, they were awarded third prize. Yet more important than the trophies was what the journey revealed—a different possibility: pupil innovation is not a distant future, but a reality already unfolding.

At Wellington College Education (China) – Hangzhou, we encourage pupils to pursue their interests, explore across disciplines, and embrace challenges that have no standard answers. We believe that every pupil is a multidimensional individual—able to be, at once, a programmer, a photographer, a poet, or a scientist. It is within such an open and diverse educational environment that pupils discover their passions, develop critical thinking and creativity, and learn to connect personal interests with social responsibility.

Seven's story is a vivid embodiment of this philosophy. From a spark of inspiration to collaborative practice to presentation on the global stage, his experience shows us that true innovation is not only about technological brilliance but also about bringing warmth to the world.

Seven is currently a lower-sixth pupil in the Senior School. Unlike many high-achieving peers, his growth has never followed a single academic path. Instead, it resembles a constantly branching exploration curve. As a child, he loved dismantling household devices and following online tutorials to build smart home systems. Later, he became fascinated with photography and video production, using images to tell stories. And when the wave of artificial intelligence arrived, he returned to his 'geek' identity—building his own computer servers and studying model training, deployment and optimisation.

These interests may appear scattered, yet they have always revolved around one core drive: turning abstract ideas into something tangible.

Seven often says that the moments he enjoys most are those when a thought like "what if this could work" appears in his mind—and he immediately begins turning it into reality, step by step. For this reason, he is rarely satisfied with "standard answers". Instead, he asks more profound questions: Why is it so? Under what conditions does this conclusion hold? As the world is continuously dismantled and rebuilt in his mind, innovation becomes a natural outcome.

In 2023, large language models were entering the public eye. That same year, Seven encountered countless stories—around him and online—of peers struggling with emotional challenges: some silenced by being misunderstood, some afraid to seek help because of stigma, others trapped by the scarcity of professional mental health resources. These stories stayed with him like a thorn, prompting a persistent question: If I have the technology, can I do something?

This question gave birth to the earliest form of the Xiaomeng project.

Seven did not immediately aim to 'build a complete system.' At first, he wrote his first line of code on his own computer, trying to teach AI how to listen. He was unsure of the direction, but kept testing, revising and conversing—observing how the AI responded in different situations. Late at night, the blue glow of the screen reflected on his face as he rewrote prompts again and again, to make the responses sound less like a machine and more like someone who genuinely wanted to stay and talk.

Looking back, he recalls that period as one of learning by doing. Days were spent in class; nights in documentation; weekends buried in open-source projects on GitHub. For the first time, learning was no longer about exams—it was about solving a real problem.

Using OpenAI documentation, open-source resources from GitHub, and his own prompt engineering, Seven built an early prototype of a WeChat-based mental health support AI. It was simple and limited at first, but it could listen, respond and accompany. What truly changed his understanding of the project was an unexpected and profound piece of feedback: a friend who had long struggled with depression told him that through conversations with Xiaomeng, she had found the courage to seek help from an adult and eventually receive professional treatment.

In that moment, Seven felt no excitement—only a long silence. For the first time, he realised that his code was no longer just a technical creation, but a path that could shape someone else's life. What followed was not pride, but a heavy sense of responsibility: if it could help people, it had to be treated with absolute seriousness.

It was then that he decided to transform Xiaomeng from a technical experiment into a long-term public welfare project.

Seven with his teammates Jayden and Kevin

As the CTB competition progressed, the project entered a new phase. Seven formed a team with pupils from different schools to rebuild Xiaomeng systematically. He led the overall concept, system architecture and functional design, and wrote the core of the research paper and presentation. Jayden translated ideas into a stable, functioning system, handling server deployment, model evaluation and backend logic. Another teammate, Kevin, conducted on-campus research, gathering feedback and refining the direction. Through countless discussions, revisions and restarts, the team gradually found a balance between technology and reality.

After extensive competitor analysis, the team established a clear principle: Xiaomeng must understand teenagers better than general AI, be warmer than traditional tools, and be more professional than ordinary chatbots.

This was not a slogan, but a guideline reflected in every technical decision.

They repeatedly discussed what the AI should never say, when it must encourage users to seek real-world help and where its boundaries lie. Under the guidance of a doctoral researcher in psychology, the team continuously refined the model—teaching it empathy without replacing judgment, companionship without creating dependency.

Through fine-tuning and prompt engineering, Xiaomeng learned to prioritise listening over giving blunt advice. By integrating professional psychological databases via RAG (Retrieval-Augmented Generation), its responses adhered to medical and ethical boundaries. The team also innovatively added phone access so that children without smart devices could still be heard. A backend risk-detection system was built to trigger alerts in extreme cases, providing a vital safety net.

Yet clarity of vision did not guarantee smooth implementation.

Deploying Linux servers, refactoring complex platform source code, enabling multi-platform access, and ensuring stability and security went far beyond curriculum requirements. Countless late nights were spent staring at error-filled terminals, starting over again and again. It was during this period that Seven truly understood what 'engineering thinking' meant: programming is not just writing code—it is the continuous ability to solve problems.

When the system finally compiled and ran stably, what he felt was not just accomplishment, but the weight of responsibility because he knew that on the other side of the screen was someone waiting to be heard.

Xiaomeng Main Page

Carrying the honour of the national highest achievement award, Seven and his team stepped onto the global stage at Harvard. Surrounded by outstanding pupils and leading professors from around the world, he felt an unprecedented sense of awe. International judges raised sharp questions about AI ethics, privacy and accountability—questions that forced him to rethink the boundaries of innovation.

"That experience taught me that true innovation lies at the intersection of technology and humanity," he reflected.

From that moment on, he became more determined that the future was not just about building more powerful models, but about using technology to respond to the overlooked corners of the world.

Looking back, Seven believes that the greatest support Wellington gave him was not a single resource, but an environment—one that encourages interdisciplinary thinking, allows trial and error, and values the process as much as the outcome. Here, literature can meet film, computer science classes discuss ethics, and pupils are guided to tackle questions with no single correct answer. It was in this environment that he learned to place technology, psychology, ethics and social responsibility on the same map of thought.

"If not for this environment, I might have only seen it as a technical exercise," Seven says.

When speaking of the future, Seven hopes that Xiaomeng can become a light in young people's loneliest moments. Yet he also says something that sounds paradoxical, but profoundly moving:

"I hope one day someone tells Xiaomeng, 'Thank you, but I don't want to talk to you today—because I've made a friend in real life.'"

Because in his heart, true innovation is not about replacing people, but about helping them return to one another.

From a campus in Hangzhou to a stage at Harvard; from a single "what if" thought to a functioning system—Seven has shown that pupil innovation can be cutting-edge and gentle, technological and humane.

This is not just the story of one pupil, but the outcome of an education already taking shape—and the future Wellington is committed to nurturing.