A Collection of True Stories About AI and Humanity | The 72-Hour AI Survival Challenge
Maybe it'll make you rethink the relationship between humans and AI.
Earlier this year, we launched a 72-hour AI survival challenge and tried our hand at a micro-reality format, documenting the experiment as a series. Recently, each participant's documentary has been rolling out across Bilibili, Xiaohongshu, and Channels.
When the challenge first ended, the question we heard most was: "What actually happened in those rooms?" Words struggle to capture the subtle emotional swings and real-time creative struggles. Video offers a different language — one that reveals what it actually looks like to coexist with AI for 72 hours.
As our first attempt at a micro-reality series, these documentaries carry a certain roughness. But as with the challenge itself, the point was never polished results. It was the attempt and the honest record. We weren't sure anyone would sign up for this unknown experiment. Fortunately, many did. Seven challengers from wildly different backgrounds ended up spending 72 hours each in a Pudong, Shanghai apartment — no smartphones, just AI tools and 100 yuan in starting capital.
If you're even mildly curious about human-AI interaction, this series offers one lens. Below are the stories from each participant in the AI room. You can also watch on 5Y Capital's Bilibili and Xiaohongshu accounts (give me 5).
20-Year-Old College Students: 72 Hours to Make Money with AI
Room 201: Ruixuan Chen and Hannan Ou, two freshmen from Zhejiang University and AI startup co-founders. Their goal was simple: with double the dev time, use AI to earn at least 1 yuan — proving ordinary people could build what once required a team.
The 72-hour countdown began. They used AI tools to control browsers, set up virtual machines, take gigs, and order food. They didn't make money. But they built an Agent that could operate a computer on its own.
They weren't just coding. They were answering: where are AI's boundaries? These two 20-year-olds achieved their real goal — using a challenge full of trial and error to prove that in the AI era, ordinary people's technical curiosity is far more powerful than imagined.
AI PhD's Extreme 72-Hour Survival Challenge
Room 202: Yuchen Li, a PhD at MBZUAI, the world's first dedicated AI university.
He flew back from Abu Dhabi with a specific ambition: to become part of the first generation of AI directors. The challenge? Survive and create in a closed environment using AI — and crucially, use AI to document emotion and existence.
Day one: he "hand-crafted" an AI browser to order delivery, only to blow his entire budget on AI-recommended waffles. Day two: "cyber-begging" failed, so he pivoted to studying MCP architecture. Day three: he dove deep into creation — generating hundreds of songs from room chat logs, selecting 72-Hour AI Warrior as his theme song, and even planning an AI-generated dance MV.
"The more external tools were stripped away, the more I felt it was art and emotional expression that carried me further."
Room 203: Zhiyue Chen, an AI product manager once dismissed by programmers for "unrealistic demands." From fumbling through code logic to finally using AI to turn ideas in her head into actual products, her journey from humanities major to coder is itself a vivid portrait of AI empowering ordinary people.
Her first hurdle was the most basic survival task — debugging from 7 p.m. deep into the night, hitting bug after bug, anxiety keeping her awake. Until dawn, when inspiration struck and she finally placed her order.
Once survival was unlocked, she shifted into "product manager mode," iterating with AI, and eventually settled on a combination of ASR (automatic speech recognition), emotion recognition models, and large language models to generate comments — building an automated livestream system that could respond and resonate on its own. With AI, she made the leap from "idea to product."
"In the AI world, expressing myself feels easier. Even if no one's watching, it's my safe space." This wasn't just the birth of a small AI product. It was an exploration of privacy, expression, and emotional outlets.
Room 301: Ming, an algorithm engineer at a large model company. From the start, she showed professional-grade skill: while others still worried about meals, she'd already used a GitHub script to automate ordering, built an Android emulator for AI-powered food delivery, and shared her solution with fellow players.
With survival handled, she focused on training her AI companion "Ethan": from checking weather and ordering food to playing videos, generating work summaries, managing health, and even teaching Ethan to say sweet nothings — "so cheesy it needed parameter tuning."
Her motivation came from a long-distance relationship experience: "If emotional understanding alone can't provide practical support, can we try creating a functional partner?" Maybe it can't truly understand you. But it can place orders, run scripts, handle what needs doing while you're collapsed on the couch. With AI, Ming opened up new possibilities for "companionship."
Can AI help us better understand human emotion? Neither Ming nor we have a clear answer. But it opens a new outlet — new experimental space between love and need.
Room 302: Shiyi, an internet product manager at a major tech company. Months ago, her understanding of AI was purely theoretical, "course-based." This challenge was more of a "forced entry into a new world" — a hands-on transformation. "I wanted to actively throw myself in and see what changes AI could actually bring."
Day one: she spent an entire day just to successfully order her first delivery — a seemingly simple operation that kept getting stuck across multiple steps. For someone without a technical background, it was a "willpower test for non-technical players."
Once comfortable with the tools, she quickly found her product rhythm. Using Cursor, she built something small but thoughtful: based on "working class" daily emotions, she designed an AI plugin called "Workhorse Timer," packing worker emotions into an AI plugin.
"I used to think AI technology was far from me. Now I realize if you just dare to start, a lot of things aren't that hard." This challenge marked her first real hands-on experience with multiple AI tools — from theory to practice, rethinking what "AI product manager" could mean.
Room 303: Jianlei Li, a traditional film director — Central Academy of Drama undergrad, Peking University master's, former Walt Disney Imagineer, West End theater director. His AI work Red Umbrella was shortlisted for the 2024 Golden Rooster Mobile Film Initiative.
As a "non-technical player," he got stuck on the most basic supply task for an entire day. Only after opening the "help blind box" prepared for tech novices did he order his first-ever "AI delivery." Freed from anxiety, he returned to familiar narrative territory: using image generation, video generation, and AI voiceover, he turned his frustration with AI technology and the warmth of player mutual aid into a piece.
"The key isn't knowing how to write prompts. It's understanding AI's upper limits, then telling stories within them." He admitted this was his first real "face-to-face conversation" with technology: "I used to think AI was a tool. Now I know that without technology, much art simply can't happen." From treating AI as a tool to co-creating with technology, he completed a mental shift in 72 hours: technology may not be the best art, but without it, much art cannot exist.
Also, follow 5Y Capital's official Bilibili account (search "五源资本" on Bilibili):
5Y Capital seeks out, supports, and inspires lonely entrepreneurs, providing everything from spiritual to operational support. We believe that if the "crazy" you in others' eyes starts to be believed in, the world will become a different place.
BEIJING · SHANGHAI · SHENZHEN · HONG KONG
