China’s 72-Hour AI Challenge: Survival, Society, and the Tech Soul

The room is a study in brutalist minimalism. It contains a single bed, a toilet, one bottle of drinking water, and a solitary roll of toilet paper. There are no windows to the outside world, no art on the walls, no books on a nightstand. The only sound is the low, persistent hum of a computer. This sterile, sealed-off space, located in a rented guesthouse on the sprawling outskirts of Shanghai, is not a prison. It’s a laboratory. And for 72 hours, it is home.

This was the setting for the “72-hour AI Survival Challenge,” an audacious experiment held from May 15-18, 2025. The rules were as stark as the room itself. Seven participants, chosen from over 300 applicants, were locked inside with a starting capital of just 100 RMB—about $14. Their only connection to the world was a computer pre-loaded with a suite of artificial intelligence tools and a non-smart mobile phone for receiving, but not initiating, contact.¹ The most crucial rule, the one that elevated this from a mere tech demo to a profound human trial, was the absolute prohibition on manually opening any traditional internet products. No browsers, no social media apps, no food delivery platforms. This wasn’t about using AI as a helpful assistant; it was about being utterly, terrifyingly dependent on it for one’s most basic needs.

As an American who has spent years chronicling the dizzying pace of China’s tech scene, I’ve seen my share of ambitious projects. But this felt different. This wasn’t just a hackathon; it was a human experiment, a live-action simulation of a future we’re all hurtling towards. Organized by Five Yuan Capital, a major player in China’s venture landscape, the challenge was designed to answer two deceptively simple questions posed by Meng Xing, the firm’s partner and the event’s initiator. First: “Can one survive solely with the help of AI?” And second, the more ambitious follow-up: “Can one achieve higher pursuits with the help of AI?”.¹ These questions framed the event not just as a quest for survival, but as a search for transcendence. In doing so, this 72-hour ordeal became a perfect microcosm of the nation’s frenzied AI ambitions, its punishing work culture, and the profound questions China is asking about its own technological soul.

Part I: The Challengers and The Code

The success or failure of this experiment hinged on two factors: the capabilities of the artificial intelligence and the ingenuity of the human subjects. The organizers curated both with the precision of a master strategist, assembling a cast of characters and a digital toolkit designed to test the absolute limits of an AI-native existence.

The Human Guinea Pigs

The seven participants were not chosen at random. A venture capital firm like Five Yuan Capital is in the business of betting on winners, and this challenge was, in essence, a high-stakes market research study. By selecting a diverse group from a pool of 307 applicants, they created a living portfolio of test cases, each representing a critical archetype within China’s burgeoning AI ecosystem.¹ They didn’t just want to see if a coder could survive; they wanted to see if an AI-only world could support strategy, creativity, entrepreneurship, and personal reinvention.

The cast included:

• The Academic, Li Yuchen: A doctoral student from the Mohamed bin Zayed University of Artificial Intelligence, the world’s first AI-focused university. He represented the pinnacle of theoretical knowledge, the pure research that underpins the entire industry. The question for him was whether deep academic understanding could translate into practical survival skills in a raw, unstructured environment.¹
• The Corporate Insider, Shiyi: A product manager from one of China’s internet behemoths. She embodied the established tech world, a world of structured processes and immense resources. Her challenge was to break free from that corporate scaffolding and wield AI with the lean agility of a startup founder.¹
• The Artist, Li Jianlei: A professionally trained young director. He was the crucial test case for AI’s role in the creative arts. In the isolation of his room, would AI prove to be a true co-pilot for storytelling, or just a glorified tool for generating assets to combat boredom?.¹
• The Builder, Ming: An algorithm engineer from a large model company. As the ultimate insider, someone who literally builds the systems he was now forced to rely on, his experience promised to reveal the vast, often comical, gap between how AI is designed and how it’s actually used under pressure.¹
• The Hustlers, Chen Ruixuan and Qu Hannan: A pair of university student entrepreneurial partners. They represented the next generation of hungry, ambitious founders who view technology not as a field of study but as a weapon in the marketplace. Their journey was about using AI not just to survive, but to find a market edge and “win” the game.¹
• The Convert, Chen Zhiyue: A liberal arts student who taught herself to code and became an independent developer. She represented the great promise of AI: the “democratization” of technology, empowering individuals from non-traditional backgrounds to build and create.¹

The Digital Toolkit

To navigate their sealed world, these seven challengers were given a specific set of digital tools. For a non-technical audience, it’s best to think of them not as software, but as extensions of the human mind and body.

Their primary tool was a suite of General Large Language Models (LLMs). This was their lifeline—their Google, their email, their therapist, and their only real window to the outside. It was the primary interface for asking questions, formulating plans, and attempting to communicate with the world by generating text that, they hoped, could be acted upon by someone on the other side.¹

Next were the Programming and Development Assistance Tools, including programs like Cursor and Trae, alongside a local Python environment. These were their hands. When the pre-built world failed them—when they couldn’t simply ask the AI to order a pizza—these tools allowed them to try and build their own solutions. They could attempt to code a rudimentary web browser from scratch, write a script to connect to a food delivery API, or create a simple webpage to showcase their work. These were monumental tasks without direct internet access, akin to building a car engine with only a box of parts and a set of blueprints.¹

Finally, they had Multimodal Generation Tools. In the sterile, profound isolation of the room, these tools became the heart. They allowed for emotional expression and a defense against madness. Participants could generate images to combat the crushing boredom, create audio soundscapes to break the oppressive silence, or write and visualize video scripts to document their surreal experience.¹

The only permitted human-to-human contact was a non-real-time internal message board. This detail is crucial. It prevented the kind of easy, real-time collaboration that could solve problems too quickly, forcing each participant back into a solitary dialogue with their AI, making the experiment a true test of human-machine symbiosis.¹

Part II: The ‘Hundred Models War’ Raging Outside

To fully understand why a venture capital firm would stage such an elaborate piece of technological theater, one must zoom out from the sealed room and look at the brutal, hyper-competitive national landscape that created it. The 72-hour challenge was not a whimsical experiment conducted in a vacuum. It was a calculated strategic probe, and the pressure cooker environment inside the guesthouse was a direct reflection of the ferocious market pressures raging just outside its walls.

This conflict has a name in China’s tech circles: the “Hundred Models War” (百模大战, bǎimó dàzhàn). The term aptly describes the period of “barbaric growth” that saw hundreds of AI large models flood the market, followed by a “great wave sifting sand” as companies fight for survival and dominance.² The scale is staggering. By early 2025, over 300 large models had been released in China, with nearly 200 having completed the official government filing (

bèi’àn) required to operate.³

But the war has entered a new, more difficult phase. After the initial gold rush to simply build a model, the critical question has become, as one industry report put it, “how should large models be used?”.⁵ Finding a viable path to commercialization is proving to be, for many, “harder than training the model itself”.⁵ It is precisely this billion-dollar question that Five Yuan Capital’s experiment was designed to answer. By creating a controlled environment, they could generate proprietary data on which AI applications have real-world utility, a decisive advantage in a war where everyone is desperately searching for a winning strategy.

This war is being fought on multiple fronts, with the government itself fanning the flames. Cities like Shanghai and Chengdu have launched massive support initiatives, establishing “Model Power Communities” to foster local ecosystems and designating AI as a “No. 1 Innovation Project”.³ The competitive landscape has fractured into several key factions, each with its own strategy for victory.

Faction	Key Players	Core Strategy	Market Position
The Incumbent Giants (“Elephants Dancing”)	Baidu (Wenxin Yiyan), Tencent (Yuanbao), ByteDance (Doubao)	Leverage massive existing user bases, cloud infrastructure, and traffic advantages to rapidly scale AI applications and acquire users.2	Dominant in user numbers but facing pressure to prove deep technological innovation beyond their established “walled gardens.”
The VC-Backed “Little Tigers”	Zhipu AI, Moonshot AI, Baichuan Intelligent, MiniMax, 01.AI	Achieve “unicorn” status through rapid, massive fundraising. Pursue specialized or “differentiated” paths (B2B, specific consumer apps, vertical industries like medicine) to avoid direct competition with the giants.2	Agile and highly valued, seen as being on the cutting edge, but facing immense pressure to commercialize and justify their sky-high valuations.
The State & Academia	Government initiatives (e.g., Shenzhen’s “voucher” system), national supercomputing centers, university labs	Provide foundational support through favorable policies, massive computing power (like the “East Data West Computing” project), and a steady pipeline of talent.2	The strategic backbone of the entire national effort, focused on ensuring China’s long-term competitiveness rather than short-term profit.

Part III: The Great AI Airdrop

While the stated goals of the experiment were to test the limits of AI for survival and higher pursuits, there is a deeper, more strategic way to understand the event. To decode the true purpose of the 72-hour challenge, we can borrow a concept from the wild world of cryptocurrency: the “airdrop.” The AI Survival Challenge can be best understood not as a scientific experiment, but as a sophisticated “Corporate Airdrop”—a novel strategy designed to acquire elite talent, generate proprietary market intelligence, and build a powerful ecosystem.

Deconstructing the Airdrop

In the crypto space, an airdrop is a marketing strategy where a new project distributes its digital tokens for free to the wallets of active community members.¹¹ The goal is to generate buzz, reward early adopters, and build a loyal user base.¹³ Often, users must perform simple tasks—like sharing a social media post or joining a community group—to qualify. This is known as a “Bounty Airdrop”.¹⁵

Now, let’s map this framework onto the AI challenge. Five Yuan Capital was conducting an airdrop, but the asset being distributed wasn’t a digital coin. It was something far more valuable in China’s hyper-competitive tech scene: opportunity.

• The “Token” as Opportunity: The prize for enduring the 72-hour ordeal was a chance at life-changing career opportunities—the potential for millions in venture funding, elite job offers, glowing media exposure, and priceless access to Five Yuan Capital’s powerful network.
• The “Task” as Survival: The challenge itself was the task. This was a “Bounty Airdrop” on an epic scale. The participants were performing an incredibly demanding set of tasks to prove their worth and qualify for the reward. In crypto parlance, they were “airdrop farming,” but the crop they hoped to harvest was a career-altering breakthrough, not a handful of tokens.¹⁴
• The “Exclusivity”: By selecting only seven individuals from a field of over 300, the organizers turned this into an “Exclusive Airdrop”.¹² This targeted a pre-vetted group of high-value individuals, ensuring that the “tokens” of opportunity were distributed only to the most promising candidates.

The Venture Capitalist’s Payoff

From the perspective of Five Yuan Capital, this “airdrop” strategy yields a return on investment far greater than any traditional PR campaign.

First, it is high-fidelity talent scouting. Forget resumes and 30-minute interviews. The firm got to observe seven top-tier candidates under extreme psychological and technical pressure for 72 straight hours. It is arguably the most intense and revealing job interview ever devised.

Second, it generates proprietary market intelligence. While their competitors read analyst reports, Five Yuan Capital received 72 hours of raw, unfiltered data on the true strengths, weaknesses, and breaking points of the current generation of AI tools. They discovered real-world use cases and frustrating dead ends long before anyone else.

Third, it provides narrative control. By orchestrating the event and framing the central questions, the firm positioned itself as a thought leader at the absolute center of the AI conversation, shaping the industry narrative in a way that benefits its own investments and strategic goals.

Finally, it is an act of ecosystem building. The challenge attracted a constellation of talent, media attention, and other startups, creating a powerful ecosystem with Five Yuan Capital at its core—a private-sector version of the government-led “Model Power Communities” sprouting up across the country.³

Conclusion: Waking Up from the Simulation

As the 72-hour clock finally ran out, the doors to the sealed rooms were unsealed. Seven challengers emerged, blinking in the bright, humid Shanghai air, returning to a world of unfiltered sunlight and direct human contact. The immediate results were a mixed bag of small victories and frustrating failures. Some had managed to use AI to code rudimentary tools; others had struggled to complete even the simplest tasks, like ordering a meal. The director had created some compelling animatics, and the entrepreneurs had the bones of a minimum viable product.

But stepping out of the guesthouse, the real question wasn’t just whether the seven challengers had “survived.” It was what kind of future they had survived into. The experiment, in its brilliant and brutal design, had laid bare the central tensions of our technological moment. The sealed room was a stage for the great bǎimó dàzhàn, a hyper-competitive war for the future of intelligence. The participants’ grueling performance was a poignant reflection of the neijuan culture that defines a generation of tech workers. And the entire event was a masterclass in strategic maneuvering, a “corporate airdrop” designed to secure an advantage in that war.

Is this AI-mediated life, which the experiment sought to simulate, one of greater freedom and creativity—an escape from the relentless grind? Or is it a more efficient, more totalizing version of the digital panopticon we already inhabit, a tool that perfects the very pressures it promises to alleviate? The 72-hour clock has stopped, but for China, and for the rest of us, the real experiment has just begun.

Works cited

1. 72小时全程纪实：人类首度与AI亲密接触 – 36氪, accessed June 13, 2025, https://eu.36kr.com/zh/p/3333024150333961
2. 中国大模型“狂飙”又一年：“大浪淘沙”后“由有到专”-新华网, accessed June 13, 2025, http://www.news.cn/finance/20241227/cf83dc526d124218a56b3557a6342286/c.html
3. 打响“百模大战”，成都大模型赛道迎来加速跑_中国网四川, accessed June 13, 2025, http://sc.china.com.cn/2025-02/28/content_43038286.html
4. 智瞰AI｜戴文渊：“百模大战”不是太多，而是远远不够-新华网, accessed June 13, 2025, http://www1.xinhuanet.com/tech/20250213/6547dab3be0c4e6eb744881e78041b30/c.html
5. 上证深一度| “百模大战”进入新阶段商业化之路有待探索 – 时政, accessed June 13, 2025, https://news.cnstock.com/industry,rdjj-202404-5224897.htm
6. 996工作制- 维基百科，自由的百科全书, accessed June 13, 2025, https://zh.wikipedia.org/zh-cn/996%E5%B7%A5%E4%BD%9C%E5%88%B6
7. 中国打击“996”能否解决年轻人的“躺平”焦虑- BBC News 中文, accessed June 13, 2025, https://www.bbc.com/zhongwen/simp/chinese-news-58419492
8. “996”是互联网行业用工史上的一段弯路–经济·科技–人民网, accessed June 13, 2025, http://finance.people.com.cn/n1/2021/0909/c1004-32221937.html
9. “996”是互联网行业用工史上的一段弯路 – 光明网, accessed June 13, 2025, https://news.gmw.cn/2021-09/09/content_35150180.htm
10. 中国码农对“996”不满持续升级当局严控舆论维稳, accessed June 13, 2025, https://www.rfa.org/mandarin/yataibaodao/meiti/bx-11202020110149.html
11. coinswitch.co, accessed June 13, 2025, https://coinswitch.co/switch/crypto/crypto-airdrops/#:~:text=A%20crypto%20airdrop%20is%20a,by%20completing%20certain%20simple%20activities.
12. Crypto Airdrops Explained – A Complete Guide – Token Metrics, accessed June 13, 2025, https://www.tokenmetrics.com/blog/crypto-airdrops
13. What are crypto airdrops? — Bitpanda Academy, accessed June 13, 2025, https://www.bitpanda.com/academy/en/lessons/what-are-crypto-airdrops
14. What are Airdrops? – Crypto Council for Innovation, accessed June 13, 2025, https://cryptoforinnovation.org/what-are-airdrops/
15. What is a Crypto Airdrop?: The Risks and Rewards | Walbi | AI Help Center – Intercom, accessed June 13, 2025, https://intercom.help/walbi/en/articles/8569026-what-is-a-crypto-airdrop-the-risks-and-rewards
16. Best Free Crypto Airdrops of 2025: Your Go-To Guide – BitDegree, accessed June 13, 2025, https://www.bitdegree.org/crypto/tutorials/best-free-crypto-airdrops