Go invest in games, like you've never been hurt before.

@Chen Zhiyan

"Don't leave this industry to people who don't love games."

Christmas Day, 2024. Shao Yun's last day at NetEase. An old friend said this to him.

Shao Yun had spent nearly two decades at NetEase. As former Senior Vice President and head of the Tianxia Business Unit, he oversaw the birth of megahits like Eggy Party, Knives Out, and Tianxia 3, while also incubating stylized breakouts like Identity V and Where Winds Meet.

A few months later, he and former colleague Fang Da founded Magic Find Ventures (MF Ventures). Fang Da holds a JD from Columbia University, previously served as VP of NetEase Games' investment division, and ran her own fund. She has backed numerous well-known companies including TapTap, Replika, and Sonantic, and helped shape NetEase's global gaming fund strategy.

They are building a $50 million early-stage gaming fund. They have just completed their second close, with LPs including HSG, IDG Capital, and several gaming industry giants.

The timing is intriguing.

Traditional financial VC firms have tried to dissect this "ninth art" through models and technical metrics. The result? Either arrogantly missing Genshin Impact and Black Myth: Wukong, or succumbing to FOMO and firing blindly when the wind shifts. Strategic investors with systematic resources, meanwhile, tend to prefer playing catch-up, rarely betting at the earliest stages.

Shao Yun, entering at this moment, wants MF Ventures to answer: Who are the next generation of super creators? And who has the ability to find and accompany them while they're still early?

At NetEase, Shao Yun interviewed countless young people and evaluated numerous creative pitches. He witnessed how China's brightest, most game-obsessed minds started from campus recruitment programs, survived all-night development crunches and the darkest moments of product stillbirth, and eventually emerged as fully fledged producers.

Shao Yun is someone who likes to extract abstractions from specifics. He says that if human brain training resembles large models, then over the past two decades he has been fortunate to possess one of the industry's highest-quality datasets, "with repeated validation and post-training optimization through data performance and player feedback."

Now, he wants to install that model into an institutionalized VC.

This is elsewhere's sixth new fund story, following Nebulon Ventures, Source Code Rhythm, Creek Stone, XiaoXiao Fund, and Impa Ventures.

The following is Shao Yun's account — including his journey toward game investing, and how he believes the discourse power of the gaming industry will return to super creators in an AI-driven future.

The Best Vintage

Christmas Day, 2024. My last day at NetEase. Before that, I had been at the company for nearly twenty years.

When leaving a major corporation, the paths ahead were actually quite clear: either fully retire, since the era's dividends had already given me decent financial returns; or start my own company and keep making games; or find a third way to continue outputting value to the industry — through investment.

Initially, I did plan to retire and travel the world. But that calm didn't last long. It was shattered by DeepSeek's explosion in early 2025.

The restructuring of technological paradigms meant that productivity and organizational relationships in gaming would face seismic shifts. This tsunami of change made it impossible for me to sit still. I immediately reached out to my classmates from Zhejiang University's Chu Kochen Honors College mixed class — they were partners at High-Flyer Quant. I told them: "If DeepSeek wants to explore consumer-facing games, I can help." But after discussion, their feedback was pure and simple: DeepSeek must stay fully focused on AGI.

I joked at the time that there might never again be a company in this world that I'd be willing to work for. But it was precisely at that moment that I felt a strong push — I wanted to empower more outstanding creators, to explore the uncharted territory where AI and games intersect. Selfishly, I wanted a front-row seat to see what the next generation of super innovation would look like.

So I began making personal angel investments, backing early-stage projects like "Mujian" and "Retrograde Plan," and co-launching the podcast "405 Gaming Bureau." During this period, top-tier funds also approached me about becoming a Venture Partner. These conversations helped me open entirely new perspectives beyond the gaming industry.

Then came the day-long serious conversation with my current partner, Fang Da.

Fang Da had invested in numerous funds and started her own. We thoroughly dissected the current industry cycle and reached a consensus: at the intersection of technological transformation, now is the best vintage for early-stage gaming investment. She also made me realize the limitations of personal angel investing — if you want to build a continuously rolling, mutually empowering ecosystem, you need to operate a more purely market-driven and disciplined institutional VC. Only then can you truly amplify industry know-how.

I began tracing this logic to analyze the global gaming VC market. I found a fascinating gap: worldwide, people who have actually shipped top-tier products and then transitioned to early-stage investing are extremely scarce. And in the chaos of a major cycle shift, this is precisely an extraordinarily rare window.

But what truly made up my mind, beyond the commercial logic, was what an industry friend said to me:

"Shao Yun, don't hand this industry over to people who only talk business but don't love games."

Big Corporations Are Failing, Individuals Are Rising

When I was in college, in Zhejiang University's Chu Kochen mixed class of 200-plus students, the vast majority chose to go abroad or join foreign enterprises after graduation. Entering the gaming industry was an extremely非主流 choice. But I probably spent the most time on games during my studies.

Back then, school computers required swiping cards in the computer lab. Chu Kochen students had unlimited time on their cards. We'd take that unlimited card, bring a large bottle of Coke, and play CS all day in the lab.

I joined NetEase through campus recruitment in 2006, employee number 2041. There was no concept of "big tech" back then. The company was on Keyun Road in Guangzhou. Everyone lived nearby, rented places together, ate all three meals together. We'd go to the office on weekends mainly because the company had better internet — we could game together. It was a state of pure play-and-build.

You could say I was fortunate to experience every cycle of Chinese gaming: from pirated single-player games as a student, to the rise of PC online games after starting work, to competing for mobile-era traffic dividends from a big corporation's perspective. In the major transition from client-based games to mobile, NetEase successfully crossed over, but several former client-game giants slipped amid technological shifts. The mobile era ultimately ushered in the "Four Little Dragons" represented by miHoYo.

In the previous cycle, big corporations' operating logic was often "playing catch-up": because product innovation cannot be precisely predicted, corporations preferred to amplify and harvest innovation's dividends through production capacity, quality, mature operations, and marketing resources.

But today, a new major cycle has arrived. Both the production and demand sides of gaming are undergoing simultaneous upheaval. The traditional investment methodology is being thoroughly deconstructed by AI.

The most intuitive change is that content production costs have been flattened. 3D assets, art, voice acting, copywriting, and prototyping are all accelerating. Small teams now have, for the first time, the opportunity to leverage AI-native workflows to achieve the visual complexity and content volume that previously required massive bureaucratic organizations.

Meanwhile, players' blind deference to "high-cost industrialized blockbusters" is also loosening. On decentralized platforms like Xiaohongshu and Bilibili, big corporations can no longer monopolize player attention through budgets. More and more small products, even one-person creations (OPCs), are emerging. As long as the content's strength is sufficient, algorithms will push them directly to players.

We recently invested in an otome game. I asked the founder: "Players of the current top female-oriented games have extremely strong guard and resistance toward AI-generated illustrations, AI voice acting, and copywriting — they even fight about it on social platforms. How do you view this burden?"

She gave me a penetrating answer: "These people aren't my core users right now. I'm serving players who accepted AI from day one. I believe the next generation of players will only grow."

This is where new-generation AI games begin digging their moats.

AI Reshapes Everything

Traditional game stickiness has two layers: the first is financial sunk cost based on money and time spent — any spending or time investment counts; the second is physical sunk cost based on muscle memory and operational proficiency (like your positioning and aiming skills in shooter games).

AI-native games are constructing a third, and highest, layer of moat — emotional and data sunk cost. Because the player has engaged in hundreds of hours of unique, irreversible dynamic interaction with this world or character, its understanding and sedimentation of you is something no one can copy or transfer.

In fact, developing evergreen games was never "building a car," but "building an assembly line that can produce different cars." This requires producers to construct a complex system where talent, tools, and organizational structure are efficiently coupled.

AI has already entered game development from the most fundamental tool layer. As AI efficiency improves, talent requirements are trending toward the "T-shaped talent" defined by Valve.

Organizational structure is also beginning to change. Early AI content was difficult to present directly to players, requiring extensive human filtering — particularly the producer's own taste — with AI production serving cost reduction and efficiency gains. But some AI content pipeline outputs can already face players directly. This portion is what we call AI-Native. Game producers are building more and more AI-Native pipelines, becoming "super individuals" within organizational structures, amplifying their own capabilities by hundreds of times.

All of this is accelerating into reality in 2026. In early 2025, a game producer attempting to finetune open-source LLMs — so-called "distilling oneself" — was still a fairly high-barrier activity. This year, the generalization of AI tools in public consciousness has even brought panic.

How to understand the rapidly expanding boundaries of AI capability and eliminate the panic brought by uncertainty? I think the best way is to do it yourself.

So within MF, I also built an investment research tool containing my own taste and most work contexts, called "Zhengzheng." Every night I've been vibe coding, even occupying a large portion of my gaming time. At least on this small investment research tool, through continuous use and iteration, the AI-Native pipeline has already changed our own working patterns.

Changing Two Variables Births Giants — What About Three?

Value creation in the gaming industry is essentially woven from four underlying drivers: the creative origin (what to make), production efficiency (how to make it), interaction paradigm (how to play), and distribution model (how to reach users).

In past commercial history, changing one variable is enough to birth a phenomenon-level breakout. Changing two variables simultaneously will nurture a new generation of industry giants. Previous cases include: Angry Birds emerging from mobile touchscreen interaction; 4G networks enabling the real-time interaction of Honor of Kings; short-video platforms like Douyin and Kuaishou allowing UGC game content clips to bring in new users at low cost, and so on.

Among these four variables, except for the "creative origin" which still needs to be driven by the human soul and cannot be automated, the other three are being comprehensively reconstructed — production efficiency is rapidly leaping, interaction paradigms are becoming "thousand people, thousand faces," and distribution models will in the future be redefined by countless Agents.

Three underlying variables simultaneously undergoing disruption means a major reshuffling of industry order, and also means gaming investment must thoroughly smash the empiricist "rearview mirror" of the past.

When traditional financial VCs invest in games, their script is always the stereotypical three-question combo: "Do you have a successfully shipped representative work? What's your product benchmark? How does your monetization model work?" Many investors don't play games themselves, let alone perform pattern recognition at the extremely early demo stage. They cannot penetrate product surfaces to assess creators' underlying value; or they swing to another extreme — only caring about grand technical concepts and visions on PowerPoint.

But in MF Ventures' coordinate system, we have a completely different standard for identifying people. We are looking for those who can engineer their underlying aesthetics, judgment, and philosophy as Fantasy Creators into AI pipelines.

They must possess near-obsessive devotion to creation, while making AI their most efficient execution partner; they must be sufficiently resolute to endure long-cycle loneliness, rather than being refined egoists.

Therefore, we resolutely avoid people who do generic competitor benchmarking, throw out big concepts to stack technology, or chase trends and frequently pivot. We want people who can make AI serve human experience — whose fantasy is to give more people better fantasy.

Games Are the Endgame of Everything

This utopian-sounding provocation originated from a discussion between Fang Da and me. We were talking about Elon Musk's prophecy: "Everything will change in the next 5 to 7 years."

If you follow first principles, what is the endpoint of this change?

With AGI, world models, and embodied intelligence running at full speed, technology's productivity flywheel is iterating at a pace of 10x cost reduction annually. Humanity will inevitably be comprehensively liberated from heavy, survival-oriented material labor. When AI continuously surpasses us in data, computing power, and knowledge retrieval, this is by no means a depressing thing — because these were never the core of what makes us human.

Handing efficiency over to AI, which is better at it, is precisely liberating humanity and returning to us our most precious and least algorithmically averageable part: subjective experience itself.

Unique spiritual projection, complex and authentic emotions, the pursuit of meaning and belonging — these will become the protagonists of life again. People will be able to more calmly define what they truly want, and materialize it in the digital world, experiencing myriad lives within unidirectionally flowing time.

But I don't believe the future leads to the stereotypical metaverse narrative of Ready Player One — everyone living in a highly centralized unified world defined by super-giants. That prophecy is too mechanical, too arrogant. The future world I believe in is uneven, pluralistic, and seamlessly blended with reality.

Finally, I chose Magic Find as our VC's name. It comes from an attribute that veteran Diablo players all understand — Magic Find. In the game, hardcore players will spare no cost to stack their MF value, solely to increase the probability of dropping set items and high-tier treasures amid endless randomness.

Much like VC, isn't it?

Cover image: Joseph Wright of Derby, The Alchymist, in Search of the Philosopher's Stone, Discovers Phosphorus, 1771, Derby Museum and Art Gallery