Vibe Coding Should Start From Childhood

The air carried the familiar mix of free pizza, coffee, and youth.

It could have been any hackathon. Except for the people at the keyboards — among the 200 competitors, 13 were under 18 and still in middle school. The youngest was 13, barely a month into seventh grade. These faces, impossibly young, had even the Gen Z contestants joking about their own age.

There was an 18-year-old humanities student who brought an IKEA plush shark to pair programming. There was a 13-year-old girl with three years of coding experience who declared that "kids make great things because they play." There was an eighth-grader who found writing code "meaningless," so he built an entire automated development pipeline using OpenClaw instead.

One team, called "Page One," had four members averaging 13.5 years old. Their captain carried around a project banner that read: "Not future creators. We are creators now."

This was the scene at the Xiaohongshu Hackathon Finals.

The business world has always had its generational power dynamics. The venture capital world's current obsession with post-95s and Gen Z likely traces back to a cohort of 40-somethings who proved themselves in the last cycle. Beyond the natural turnover of talent, there's a drive to pass down accumulated experience, resources, and capital — to catch every wave, and to achieve something like spiritual fulfillment through mentorship: you only become a real "big brother" by lifting up your little brothers. This relationship requires a manageable age gap, a rough alignment on the rules of the game, and at least some dependence on the older generation's help.

But in an era where everything can be Vibe Coded, facing a group of untrained kids who haven't even finished compulsory education — the game changes completely.

Once, they accessed and understood the world through adults, schools, and companies. Now these intermediaries are no longer necessary. They don't need your help. At an age when you thought they could only build sandcastles, they proudly show you their products.

The next generation is already entering elsewhere.

The Self as Method

Page One's 48-hour product was called "Shu Doctor." Built on a self-trained quantitative prediction model and a five-AI-agent multi-round debate mechanism, it offers data-driven note diagnosis, score comparison, and one-click optimization for Xiaohongshu creators. The idea came from team members Yang Xizhe and Chen Yuxia, both Xiaohongshu creators themselves. In their daily posting, they kept running into what they called "traffic mysticism."

"Sometimes a carefully produced video falls flat, while a random rant about my phone's power button gets millions of views," Chen Yuxia said. "Instead of blind testing, why not fight magic with magic?"

A simple frustration. A product born from it.

This pattern is common among the new generation of builders. Their initial inspiration often comes from small frictions in daily life.

Thirteen-year-old Chen Yuxia built an aggregated DNS resolution desktop app for the most basic of reasons: he hated the hassle. He was tired of pulling out his phone to scan QR codes and log into cloud service backends every time he managed a domain. So he called APIs from major providers and built a desktop tool that managed everything in one place, smoothing out the tedious workflow entirely.

RPONE, the 18-year-old humanities student, simply missed Android's "screen recognition" feature after switching to iPhone. She learned as she went and built an iOS app called Screenpick.

Thirteen-year-old Yang Xizhe extended this problem-solving impulse to broader groups. After finding himself overwhelmed by WeChat group messages, he conceived a message summarization and priority assistant. At a previous hackathon, his middle school team developed a makeup assistant for visually impaired people using 128 facial point mapping technology — simply because they'd noticed that disabled people also wanted to "enjoy beauty."

Thirteen-year-old Lü Sitong, a seventh-grade girl, found code incomprehensible — and knew her mother couldn't understand it either — so she built an app called Little Fox Explains Code.

She designed three perspectives in the software: "Mom language," where obscure code becomes "a refrigerator that holds vegetables"; "Little brother language," where it becomes "a box that holds candy." She also built a workflow converting long texts to mind maps, addressing the "no time to read long books" problem, and monetized it over 200 times on the platform.

She first encountered AI at age 10 through Coze. Because she liked playing international server games and often needed to voice-chat, she desperately wanted to practice spoken English. But human tutors were expensive and couldn't be on call 24/7. So she built an agent called "Frog Tutor" to address this deeply personal pain point.

She used natural language to set strict grammar correction instructions and gave it personality — if she mispronounced something, the virtual frog would patiently correct her in standard English, then playfully ribbit.

This sensitivity to real but perhaps small pain points makes them naturally resistant to polished narratives.

Ma Yuhan, a 14-year-old ninth-grader from Beijing, once built a so-called ROS-guided vehicle for a traditional science competition. He didn't hide his disdain: "It was basically useless, just storytelling." The chassis was too low to navigate stairs. This time, he built an aerial photography robot and won a special category award.

Talk with these middle school geeks for a while, and you'll notice they almost never say "ceiling" or "moat." Their version of hacker spirit is simple: no compromises. Find a problem, fix it.

Build in Public

Seventeen-year-old Xian Xinglang couldn't afford the 688 yuan annual Apple Developer fee during early development, so he asked for help on Xiaohongshu. The post spread through influencer "Hua Shu" and unexpectedly reached Manus founder Xiao Hong. Xiao sponsored the fee and encouraged him to build "positive feedback loops." This propelled Xian into the global top 50 of Apple's Swift Student Challenge.

RPONE similarly benefited from social distribution. She priced her app at 1 yuan, documented her stumbling development process on Xiaohongshu, and unexpectedly achieved high conversion with over 3,000 downloads. In the comments, users didn't just demand updates — they made precise feature requests like "add zoom" or "recognize any image." As more users followed and downloaded, the app broke into Apple's top ten paid tools list, peaking at number six.

For these developers raised in social media, "build in public" isn't performative. They freely share buggy prototypes and livestream their clumsy hand-built processes.

Yang Xizhe has already taught millions on social platforms "how to use AI to memorize vocabulary." He considers this another form of open source. In his comments, fans send encouragement, and one even played his video during their school's "first day of class" assembly.

Though Yang is a polished cross-platform creator, he's noticed that AI content gets more feedback on Xiaohongshu. "It's like a giant library of living people." He's grown accustomed to searching for the latest on OpenClaw and foundation models there, teaching others in comments while learning himself.

This open-source ethos carried into the hackathon.

In Yang's previous experience, teams at traditional hackathons treated each other as rivals to guard against, terrified of idea theft, rarely interacting. But at this competition, Yang and the Page One members were surprised to find that the "adults" exchanged ideas freely. Teams that were supposed to be competitors sincerely encouraged them to break into the top ten, hoping they'd keep building their project.

Thirteen-year-old Chen Yuxia has gone further down the "public expression" path.

His initial online presence came from a family interaction. To explain "flashing a ROM" to his non-technical parents, he earnestly recorded an explainer video. What was just family documentation, along with a later video complaining about a broken phone power button, went unexpectedly viral — 9 million views combined.

As a tech influencer with hundreds of thousands of followers, commercial agencies came calling.

Chen once signed with an MCN but found the process cumbersome and the cuts excessive. The 13-year-old middle schooler terminated the contract and began negotiating brand deals solo. He has clear commercial principles: "If I don't like the product, I absolutely won't take it."

In the past month, he's taken three product experience reviews — from Kimi, MiniMax, and Feishu.

Among the middle school builders elsewhere spoke with, nearly everyone has a Xiaohongshu account. They don't just post products and document development there — they learn coding, follow "developer gods," accumulate seed users, and even connect with VC investors.

What's evident is that this AI-native, social-media-raised generation is making the entire build chain public — from ideation, validation, and iteration to user acquisition, team building, fundraising, and partnerships. Everything happens in communities like Xiaohongshu.

This habit of building in public from the very first spark of inspiration has become the natural innovation mode for a new generation of creators.

Starting Young

Today's AI-native post-2010s kids carry much of the spiritual DNA of their 1980s-born parents.

Yang Xizhe's coding journey began in second grade when his father handed him The Legend of Zelda. Faced with heavy academic pressure, his parents didn't enforce "grade levels" or "perfect scores." They did the math: 110 out of 120 was excellent enough. The remaining time was his to explore programming and AI.

Ma Yuhan's mother plays the demanding "project manager" at home: grades can't slip, homework must be done on time, and she radar-scans for science competitions and hackathons. But on the core question of "what to build," she steps back and stays silent. From age 10, when he started tinkering with drone payload systems at home, to the current house full of microcontrollers, soldering irons, and robot parts — she's let him be.

At this hackathon, her attitude and "famous quote" made it onto an event poster: No need to book Ma Yuhan a hotel, he'll definitely choose to sleep at the venue with his teammates.

These parents of "AI natives" occupy a delicate moment. The AI wave rolls on, but educational institutions turn slowly. How should children face the future? Test scores are clearly no longer the only answer. Whether it was the STEAM courses popular in recent years, or the more direct support for development and creation seen at this competition, the paths to a future ticket have diversified.

Turn the clock back ten years, and China's tech innovation narrative was still top-down: elite university students and executives exiting internet giants were the main entrepreneurial force, while VCs sat in mentor chairs dispensing wisdom. Today, creators are emerging from every corner of society.

The barrier to innovation is falling.

Near the end of this 48-hour hackathon's demo night, judge Cao Xi of Monolith asked the contestants: "Do you still need VC money?"

At previous tech competitions, Jiang Muran won 200,000 yuan with a one-person-plus-AI pipeline project. Xian Xinglang received an invitation to Apple headquarters in the US. Middle school teams have begun defeating university and graduate student teams in competitions, pitch battles, and roadshows.

On the technical foundation, large models, the spread of Vibe Coding, and continuous improvements in hardware infrastructure have lowered the "how" barrier. Meanwhile, these young builders have formed a subtle symbiosis with their communities: young builders propose ideas and vibe-code crude products, breaking the elite narrative of tech circles; in turn, the "build in public" community ecosystem provides them with an extraordinarily tolerant product testing ground.

In RPONE's story, Xiaohongshu's comment section functioned as a product manager's dashboard: someone happily paid 1 yuan and left a long review; someone sharply pointed out system compatibility bugs; more people, while demanding updates, began feeding precise requirements — "Can you add a local zoom feature?" "I want it to recognize images directly from my local album."

RPONE responded just as quickly. Classes during the day, then following the comment section's guidance to have AI modify code at night, rapidly releasing iterative versions in the community.

More young people with nothing to their names have thus achieved not just the democratization of expression, but of creation itself.

Around the time of the competition, a speech video by Lü Sitong went viral online. The 13-year-old girl's talk was titled "A Story of Someone Born in the AI Native Era."

You could say: vibe coding, start them young.

Cover image: Pieter Bruegel the Elder, Children's Games, 1560, Kunsthistorisches Museum