This article was completed at the end of 2025 and sat in my drafts for a long time. Today, I finally have the opportunity to share it. It's quite lengthy, so I've created a reading guide to help you quickly navigate to each section:
Introduction: What happened in three months, why I had to write down this experience, and what reusable experiences and conclusions you'll gain from this long article. New Project Introduction: What exactly is MiroFish, moving from "analyzing the past" to "predicting the future," and why prediction is the natural next step for BettaFish. Experience Sharing: Both BettaFish and MiroFish are products of ten days of Vibe Coding. The concept of super-individuals can really succeed and will only get better. I'll share a replicable workflow for collaborating with AI to turn ideas into runnable projects. Story After Going Viral: The collapse and rebuilding after gaining 20k stars in one week, with offers and investment opportunities flooding my inbox, and the real needs I observed in China's AI market. 30 Million Investment and Incubation: From choosing Shanda to Shanda choosing me again—an AI Native organization providing an unreserved stage for super-individuals. Final Words: The MiroFish team officially releases a recruitment call for super-individuals. We're looking for similarly "restless" young minds.
Introduction
Three months ago, I was still anxious about internships.
My thinking was simple back then: since my academic path was settled, I could finish my graduation thesis ahead of schedule and free up time during senior year to find an internship. So, in the last ten days of summer break, I used Vibe Coding to complete the open source project BettaFish. Honestly, in my mind this was just a basic operation—a daily coding workflow, integrating years of college learning into an engineering project. I didn't put my whole heart into it, nor did I think it would make any waves. Even my PhD senior criticized it directly after seeing it: "This isn't a qualified graduation thesis, you need to put more thought into revising it."
I once thought that would be the end of it.
The first traffic turning point for BettaFish came when it reached 1k stars. At that moment, I truly felt this matter couldn't be more complete. As many friends in the comments said, as a college student's work, its technical capability might have already matched the highest stars it could achieve. I thought since it's ending, let's end gracefully, so I wrote an article and posted it on social media platforms titled "What Can a 1K+ Star Open Source Project Bring to a College Student," preparing to draw a perfect conclusion to this project.
I have two names I must deeply thank here. One is Linux Do, a forum so pure it's moving—every feedback, every like was the original motivation for me to continue managing open source. The other is the WeChat account "why technology," also a member of the L-site community, who was the first to discover my article and provide an in-depth interpretation. I remember after that WeChat article was published, I secretly read it countless times, feeling incredibly happy inside. It was precisely that article that brought the initial traffic inflection point for BettaFish.
But this article today isn't to review BettaFish again. If the previous article was about the resume boost from 1k stars, this article is where I can finally sit down and seriously talk about the almost magical story behind 36k stars.
After the open source project went viral, within one week the stars increased by 20k, quickly topping the GitHub trending list. My inbox was stuffed with various offers and investment intentions, and my life underwent a thorough, earth-shaking transformation. Not just those tangible economic benefits, but more importantly, with strong support and incubation from Shanda Group, I secured 30 million in investment, just like that, "confusedly" yet without hesitation, I started my entrepreneurial journey.
During this time, doubts never ceased. Many said it was marketing, a packaged myth. But as someone who personally experienced the super-individual era spawned by Vibe Coding, I have too many heartfelt insights to share. More importantly, people around me have already successfully replicated this path, proving again: this road isn't crowded yet, it's full of infinite possibilities. This is a path with extremely low entry barriers that can lead to tremendous success.
As a beneficiary of this wave, I want to write down all the Vibe Coding experience, promotion experience, logic of making AI products, and even the intuition of "finding opportunities" from vast information accumulated during this period, and "open source" them too. This is my way of giving back to the community and providing closure to this crazy journey.
This 10,000-word article took a long time to write. I want to incorporate every second after going viral into the text. Although words can sometimes be thin, I hope through this article, I can bring a sliver of inspiration to those who are lost, programming, or looking for opportunities—then it will be worth it.
One: New Project Introduction
After completing BettaFish, I received a piece of feedback that deeply impressed me: "The report is generated quite beautifully, 30-plus pages, detailed data, nice charts. And then? What can I do with it?" This sentence stuck in my heart. The pain point of sentiment analysis has never been "can't see data," but seeing data without knowing what to do next. A 30-plus page report looks convincing, but it can only tell you "what happened," unable to answer the more important question: what happens next? It's a "rearview mirror," while most people need a "telescope."
So what MiroFish wants to do is simple: turn the endpoint of analysis into the starting point of prediction, forming a true closed loop from raw data to intelligent decision-making. MiroFish is also fully open sourced on GitHub, now with 4.5k stars. You can understand it as a "digital sandbox"—input a news item, a policy draft, or a market signal, and it automatically builds a micro parallel world inhabited by thousands of "digital humans" with personalities, memories, and behavioral logic. They communicate, debate, and influence each other. You stand from a god's perspective, observing what results emerge from these interactions, and can even throw in variables: what if the news was released a week earlier? What if public opinion suddenly reverses? Let different "what-ifs" run in the sandbox first, then make decisions in the real world.
In the future, we will continue iterating this tool. Financial market trend predictions, evolution of hot events—like "can the president win the election," "can Moore Threads be bought these days"—these prediction tasks that can actually be monetized in reality are all target scenarios for MiroFish.
For specific project introductions and operating logic, you can go directly to the project's GitHub homepage. I've put a very detailed and interesting explainer video there, as well as a website for online experience. I believe after watching, you'll have a more intuitive feel for "predicting everything."
Two: Experience Sharing
Vibe Coding creating super-individuals—I'm increasingly certain: it can really succeed, and will only get better.
In this chapter, I want to explain my workflow clearly first, then say why I judge this way. Many people ask if I can produce a detailed Vibe Coding tutorial. I usually say it can't be done—not because I'm hiding anything, but because it changes too fast, fast enough that the paradigm you summarize today might be invalid or even unnecessary next month. At this stage, it's more like you have to personally get involved, experience, and "taste" it—it's not something mature enough to be taught by the book.
So I'd rather share something more fundamental: in this current development stage, how should we work collaboratively with AI. This philosophy applies not just to coding, but to any scenario where you need AI participation. Learn by doing, do by learning. I've also put an exchange group on the MiroFish project homepage—those interested can join to discuss and share tool usage insights and discuss startup ideas. Later I'll release some videos showing more intuitively: when you have a raw idea, how to push it to a runnable, presentable MVP form within one day.
If we take BettaFish and MiroFish as examples, both were developed by me in about ten days. But many think ten days means writing code the whole time, which isn't true. What I spend the most time on is actually market research and technology selection. Simply put, first figure out "why do it, who is it for, how to do it." Once this step is clear, execution becomes a process of "directing AI to work"—you're not sitting there coding from scratch, but like a project manager, like a director, clearly breaking down tasks, clearly stating goals, then pushing rapid implementation.
I also strongly feel how exaggerated the model iteration speed is. When writing BettaFish, I stepped in many pits—some problems required repeatedly writing prompts, repeatedly teaching AI how to debug, to slowly get things running. But by the time I made MiroFish, Claude Opus 4.5 and Gemini 3.0 Pro had just been released, with capabilities leaping again compared to a few months prior. Many previously torturous steps suddenly became "smooth matters." You could even say MiroFish's task difficulty is an order of magnitude harder than BettaFish, but the Vibe Coding experience felt lighter. This is also why I keep saying "don't count on a set of never-expiring tutorials": reality develops too fast. What should be done most is forming your own experience and judgment, not memorizing a set of fixed actions.
Specific to the implementation process, my current approach to interfaces is: first draw a rough sketch in Figma, then use some AI tools to continuously iterate, polishing it into a more beautiful, more product-like interface. Then in Google AI Studio, quickly replicate it into a clickable, interactive frontend demo—even if it's just frontend dummy data, that's fine, because this step's core value is turning abstract requirements into something "visible and tangible." Once the interface and experience are determined, add these pages to project documentation, then break tasks into modules, batch developing with AI IDEs. Here I have to praise Gemini 3.0 Pro—its frontend presentation, aesthetics, and interaction understanding are indeed strong, like a smart designer plus programmer. But my personal feel is that it's more "inspirational type," relatively weaker in engineering robustness and patience. So I use it for frontend page initialization, beautification, and interaction details needing aesthetics, while more complex engineering structures, interface design, and stability improvements, I'd rather hand to Claude, this "seasoned engineer," to chew through.
Another highly recommended practice: run multiple agents in parallel. Many tools now support agent mode. If you don't care much about cost, I strongly suggest doing the same task with multiple agents simultaneously, then you pick the optimal solution. I often simultaneously open 8 agents doing the same work when developing a module. Token consumption is indeed huge, but efficiency gains are real. More importantly, you'll very intuitively feel each model's capability boundaries—what it's good at, what it fears, where it tends to be "confident but going off track," and how you should work with it. Knowing tool boundaries is the first step to mastering them.
The more high-speed this is, the more you need a "brake system." I always focus on two things: first, use git well for code management; second, write good documentation. Errors aren't scary; what's scary is not knowing where the error is, fixing A affects B, and finally the whole project goes out of control. Many times, documentation isn't just written for humans, but also for AI. It's both an organization of your own thoughts and a detailed prompt and technical architecture design document. I often throw documentation directly to Cursor, Claude Code, Google AI Studio, and various full-stack development agents on the market, letting them run a version first. We all know this version is likely unusable, but it's very valuable: you can reverse-engineer its capability boundaries from "what it did well, what it did ridiculously," and faster find how you should express requirements, how to break down tasks, how to write constraints, and more clearly determine what you really want.
Another frequently discussed question: Is Vibe Coding stable? Is the code safe? My personal feeling is, if you just "send task, get result, test passes and that's it," this works well for small demos. But when what you're making can be called a "project," deep human-AI collaboration and deep code review are still very necessary. I not only audit AI-written code line by line, but also try to follow its execution process, observe its thinking process: why did it change this way, what are the assumptions in its mind. Many bugs' root causes aren't a wrong line, but going off track at some key point; correct that misunderstanding, and many subsequent problems disappear together. Even better, in reading through its process, you also gain ideas and inspiration, conversely improving your project documentation, letting you and AI enter a state of "deep collaboration, mutual progress": you increasingly understand where the pits are, it increasingly understands your boundaries and preferences.
My judgment about this era is actually quite simple: knowledge is being equalized. Many things that used to require ten or a hundred times the effort to even reach the threshold will soon be brought before the masses—it's only a matter of time. So what's truly scarce for individuals will slowly emerge—your imagination, creativity, taste, and your ability to apply knowledge to real problems. Returning to open source, reviewing BettaFish's experience, I increasingly feel the most reasonable input-output allocation is: upfront research and ideas take the majority, promotion is also important, actual coding isn't the heaviest part, and documentation determines whether others dare to use it and are willing to stay.
I summarize open source experience into 5 points, which are also some of my rough understandings:
Vibe Coding topic selection is especially important. Before starting to write code, you must go through comprehensive research—think first, then act. For individual developers, come from vertical fields, go to the masses. Enter from vertical fields to accumulate professional reputation, then face the masses to attract broad users. Like BettaFish positioned as sentiment analysis, but the tagline says "usable by everyone." Don't pursue being big and comprehensive. Practice subtraction, continuously correct your positioning in practice, and don't be a perfectionist—verify quickly. Marketing promotion doesn't need to be much, but definitely prepare materials in advance that let "others help you promote," like clear, intuitive demo videos. Code is cold, stories are hot. Learning to tell the story behind the code is a required course for independent developers.
"Super-individual" isn't just reflected in programming. Before when promoting projects, I wanted to find someone who understands programming and AI and can edit videos and make animations—found almost no one. Later I used some AI video editing tools and quickly made a preliminary version of the promotional video (though ultimately not used for some reasons, that experience was shocking). Today it's not mature enough, but you can completely imagine one or two years from now: one person completing all actions of research, design, art, development, marketing, will become increasingly common; costs getting lower and lower, but return ceilings getting higher and higher. This is also why I say "it will get better"—tools are accelerating, collaboration paradigms are accelerating, human boundaries are also being pushed open.
In the previous 1k star article, I kept emphasizing: Vibe Coding for products, for open source, is a feasible, replicable path, and it's not crowded yet. Today I want to add another layer of evidence with a nearby example: my roommate made an open source project called DeepAudit, using multi-agent for vulnerability mining, now at 5k stars, with quite a few enterprises coming to support. He and I often exchange tool usage and workflows. After BettaFish went viral, I had many new insights. After a long talk with him, he almost "decided to go for it"—product conception, development, promotion all in one go, with similarly rich input-output ratios. He's said to me more than once: this path really works, we chose right.
These are things I didn't know clearly when making BettaFish, but now increasingly clear and certain: broadly learn, broadly master various AI tools, imagination becomes the primary productive force, technology is no longer the only barrier, an idea hitting the pulse of the times has its value unprecedentedly amplified; speed determines life and death, market opportunities are fleeting, the ability to turn ideas into reality and push to market in days using AI tools becomes key. This is my understanding of Vibe Coding, and this is the super-individual era I believe in.
Three: Story After Going Viral
Then I want to talk about what many are probably curious about—the story after going viral. After the BettaFish open source project surged 20k stars in one week, I intuitively experienced for the first time: how urgent the domestic market's demand is for young people who have AI in hand, are full of creativity, and can make things. Precisely because this experience was too dramatic, many would assume a premise: you must have spent a lot on marketing, or someone behind pushed it. Others criticized that doing open source while spending so much effort on marketing is like "doing traffic." But the fact is simple—I only publicly promoted BettaFish twice: once when the project just completed, once when it broke 1k stars. That real explosion afterwards, honestly, I didn't anticipate either.
Here's a look at the BettaFish I raised myself.
This time it's not an open source project.
The first promotion happened when the project just finished. At that time, borrowing the shell of "sentiment analysis system" from a year ago, it already had over two hundred stars. My goal was very simple: since it's done, try to let more people see it. Even if it doesn't reach my fantasy 1k, gaining a few hundred more stars would add one more plus on my resume. So I started promoting like a novice, hitting walls everywhere: submitting self-recommendations to various WeChat accounts and weekly repositories recommending open source projects, almost all sinking like stones. Finally I switched to a dumber but more effective method, posting self-recommendations on Bilibili, Xiaohongshu, and Linux Do forums, relying on real displays and repeated expression, slowly bringing some star growth. Not much, but enough to make me believe this project at least had people willing to click and take a look.
The second promotion happened a month later. During slow growth, the project also received strong support from many in the Linux Do forum, finally breaking through 1k stars. At that moment, I returned to the mindset I wrote about in the previous article: I felt BettaFish had probably obtained "the maximum stars its technical capability could match," considered complete. So I organized my open source insights from a year, review experience from this project into an article, originally intending to give myself and the project a graceful ending. Then came the first "out-of-control luck" in my life: after the article was stable for a week, it was suddenly forwarded and recommended by many big accounts, traffic exploded, gaining 20k stars in just a few days—that wasn't a slow climb, but like floodgates opening, watching stars jump from 5k, 10k, 20k, like watching a completely uncontrolled counter.
After going viral, what rushed in first wasn't applause, but emails. Big company offers, collaborations, investments, flooding my inbox one after another. I also almost at the fastest speed earned my first pot of gold in life. But what really impressed me wasn't who gave the best terms, but the reality this matter revealed: the market is frantically searching for people who can use AI as productivity.
I want to highlight three insights, perhaps giving some inspiration to those still exploring or wanting to do AI entrepreneurship.
First point: The earlier you explore the "super-individual" path, the higher the probability of success, especially for college students. You have error tolerance, fallback options, and time—you can fail repeatedly. The transition from LLM to Agent essentially gives AI "agency"—this is another industry-wide renovation on top of digitization. Almost all traditional tasks, after digitization, can be redone with Agent collaboration, further releasing labor and improving efficiency. And there's naturally a time gap here: many industries understand business processes very well but aren't familiar with Agent collaboration frameworks; while 20-something young people have low learning costs, lots of time, abundant energy, and haven't been固化 by path dependence. In the window of technological innovation, they can often use tools deeply in very short time. The integration of these two types also needs time—I roughly feel there's at least a two-to-three year window period. This is why you'll find many internet companies' LLM leads and AI leads are 20-something young people—the times are stuffing opportunities into young people's hands.
This first point is particularly evident in hiring. Recently I've been participating in company recruitment, feeling it personally. After AI's rapid development, enterprise hiring paradigms are undergoing a novel differentiation. Now many teams prefer hiring two types of people.
First type: people with particularly solid fundamentals—senior developers, senior algorithms people, those who can stabilize complex systems and chew through key problems. These people have always been scarce, and are what we traditionally pursue.
Second type: people who use AI tools particularly smoothly—emerging in just these two years. They might not be specialists in the traditional sense, might even be criticized as "jack of all trades, master of none," but the advantage is broad coverage and quick onboarding. Facing a difficult task, they can quickly break it down, combine tool chains, maximally deploy AI, and produce results first. You can understand this as a new capability—not the ultimate of single-point technology, but the proficiency of turning AI into productivity.
Using an intuitive analogy: if the first type uses 10 days to take something from 0 to 100 points; the second type can use 1 day to push something from 0 to 70 points. This value of running first is infinitely amplified in the AI era, because iteration becomes cheaper, faster—speed itself becomes a barrier.
Of course, what's most scarce, and most enviable, are those who have both: deep technical foundation, willing to embrace AI, and can truly master AI. When we see such people in hiring now, it's truly like wolves seeing meat—wanting to immediately secure them. But reality is, most people get trapped in their comfort zones, thinking the current state is fine too, and few are willing to actively change—which is precisely where the opportunity lies.
Second point: We severely underestimated the determination of traditional and semi-internet industries toward AI transformation. Many think numerous industries are still proceeding as usual, collecting data manually, making judgments by experience. But after BettaFish went viral, I chatted with quite a few entrepreneurs and department heads, feeling very strongly: not just internet, China's overall tolerance and acceptance of AI is actually very high, and demand for AI/Agent talent isn't just concentrated at a few big companies. Almost all enterprises are experiencing a kind of "AI anxiety": they know the wave is here, not acting means being left behind, so they're very willing, even urgently, conducting AI transformation and efficiency improvement. For young people, as long as you're willing to go deep into a field, in China's vast market, whether working or starting a business, there's a place.
Third point: After going viral, I felt investors' overflowing enthusiasm for post-00s AI people and for "super-individuals." Honestly, this path is hardest to walk, because it requires you to shoulder uncertainty long-term; but it also has the most potential, because it transfers organizational ceilings partly to individual ceilings. The younger you start acting, the earlier you start treating AI as your collaboration team and productivity multiplier, the earlier you become proof that "everyone can be a super-individual."
Precisely because going viral came too fast, choices were almost forcibly placed before me: one path was taking a high-salary internship, smoothly entering a big company, living a more stable, more certain life; the other path was challenging my own cognition, catching this momentum, attempting an AI startup I'd never seriously thought about before. Many only see the opportunity, but I want to say, what hit me first after going viral was actually pressure—and unreasonable, overwhelming pressure at that.
Originally just treating BettaFish as a graduation thesis, the writing was unavoidably rough. Computers are truly magical things: programs run a hundred times without problems on your own computer, but on others' computers all sorts of strange issues pop up. So issues rose crazily, and doubts came along. During that time, I'd open my eyes in the dorm every day and start fixing bugs, becoming customer service for the whole internet, pushed forward by the project. Roommates said what a great opportunity, you should be a bit happy; but I had none of the relaxed excitement from breaking 1k stars, instead feeling like being washed away by flood—BettaFish washed my originally stable life rhythm into complete disarray.
I didn't dare open my inbox, didn't dare reply to private messages: afraid of missing opportunities on one hand, afraid of facing opportunities on the other, because clicking in means you have to start making decisions. First week I lost ten pounds, looked visibly haggard. Later I slowly sorted things out: first stabilize the project, first solve solvable problems, then face external voices and invitations. I started replying to emails one by one, meeting people to chat, for the first time using a very immature but very real way, to learn commercialization, understand the costs behind opportunities, and seriously think about where the future should go.
That month was actually very painful, but growth was also extremely fast. I was indeed ignited by the entrepreneurship thing, that impulse of "let me just go all in" was very real, even hard to suppress for a time. But just when I was about to firmly decide to start a company, rejecting all offers, Shanda "temporarily held me back." They didn't use "advise you not to start a company" to stop me, but gave me a harder-to-refuse option—a big enough stage, long enough patience, and a sentence I still remember clearly. From here, the story entered the next chapter.
Four: From Joining Shanda to 30 Million Investment and Incubation
As you can tell from the title, I ultimately didn't take that solitary traditional entrepreneurial path, but made a now seemingly incredibly correct choice. Shanda Group founder Mr. Chen Tianqiong showed great interest after seeing my project, proposing to personally chat with me. Honestly, at that time my inbox was in an explosive state, almost no time to reply, but Shanda's team showed extremely sincere dedication, tirelessly contacting me. So on that night preparing to go all-in on entrepreneurship, I still decided to first chat with Shanda.
In that conversation, I genuinely felt Shanda's attitude toward "embracing super-individuals" was more serious than any company. This attitude is summarized as a voice I hear most often at the company: AI Native organization. Mr. Chen Tianqiong afterward told me directly, he was very clear BettaFish's technical level wasn't particularly outstanding, but he valued more the complete planning I showed in the project from data collection, analysis to prediction, and that action capability of "seeing and defining real valuable problems, and attempting to solve them in AI's new way."
After hanging up, I didn't hesitate a bit, packed my bags and went to Shanghai alone.
On my first day at Shanda, the first sentence I was told was: "Continue doing what you want to do." Here there are no cumbersome constraints, no rigid task arrangements, replaced by sufficient computing power support and an unreserved stage. Under this almost luxurious freedom, I started implementing that "prediction" function already conceived during BettaFish period but never had time for. In just ten days, I completed MiroFish development at Shanda. In the late night of December 16th, watching the running program, I excitedly recorded a somewhat rough demo video, directly submitting it to Mr. Chen Tianqiong's desk.
This time, Shanda chose me again. Accurately speaking, Shanda chose to embrace any kind of life force that might grow wildly in the AI era. Within 24 hours after submitting the video, we had our second conversation. Mr. Chen Tianqiong decided on the spot: the group invests 30 million RMB, fully supporting MiroFish's deep incubation. Overnight, identity changed, I leaped from Shanda intern to AI startup CEO. This gamble about the future ultimately returned me to the entrepreneurial main line.
In this month at Shanda, I saw what a truly AI Native organization looks like: here, you might come from any position, might have never led a team, might not even have a so-called "pretty resume"; but as long as you can define real problems, dare to be responsible for results and continue pushing forward, the group will provide patient capital, sufficient computing power, and comprehensive support. These aren't just slogans, but real actions. We are an organization full of creativity and action, and I'm standing at the center of this stage.
Five: Final Words, Super-Individual Recruitment Call
Now that the story has reached here, all foreshadowing has been resolved, finally moving from "my personal experience" to "what we're going to do together next."
First, our ultimate goal: rooted in powerful data collection, driven by true reality simulation, from serious prediction to趣味 simulation, making every "what if" see results, making predicting everything possible.
Introducing MiroFish once more seriously: it's a next-generation AI prediction engine based on multi-agent technology. We'll ingest seed information from the real world like breaking news, financial signals, or any raw material you care about, then automatically build a high-fidelity "parallel digital world" in the backend. In this world, thousands of agents with independent personalities, long-term memories, and behavioral logic interact, discuss, conflict, and spread like a real society, undergoing social evolution. You can dynamically inject variables from a "god's perspective": a statement, a shock, a key remark, a clause change... then watch how it chain-reacts, how it branches, ultimately pre-running the future in the digital sandbox, letting decisions return to reality after hundred-battle simulations.
The product vision is also clear: we want to create a collective intelligence mirror mapping reality. Traditional prediction often stays at model inference or single-point judgment, but the real world is more like collective emergence: how opinions influence each other, how emotions spread, how factions form, how turning points appear—these things can't be calculated with a sentence; they must be run out in an evolving environment. Macro-wise, we hope to become decision-makers' rehearsal laboratory, letting decisions, PR, and business strategies repeatedly trial-and-error in a near-zero-risk environment; micro-wise, we also hope to become individual users' creative sandbox—you can simulate novel endings, or let your imagination fly, exploring "what if we did this then, what would happen."
To make this happen, what we need isn't a group of "standard answers," but a group of "non-standardized" young partners. MiroFish is an extremely young team. We're looking for similarly restless young minds: hoping you're an AI master, not just a user; you don't need to follow rules, but must have near-fanatical acuity for AI; you don't need to obediently do as told, but must be accustomed to deep thinking first, then extreme-speed action. We reject mechanical execution and formulaic development, favoring wild ideas and unconventional solutions. You can be immature, but must be sharp; can be reckless, but must be sincere; can lack experience, but must be willing to be responsible for results, willing to keep getting stronger in high-speed iteration. If you crave personally building a cyber sandbox that can simulate everything, and firmly believe code can be both hardcore and fun—don't hesitate, this is your home field.
We're hiring full-time/interns, based in Shanghai. If you have great passion for the super-individual concept, can master AI, and are full of creativity, imagination, and action, welcome to send your resume to: mirofish@shanda.com. A young AI startup team, waiting for you to join.