After years of making content, I have seen many viral hits. Whenever one appears, the team’s reaction is usually the same. Take screenshots. Check the numbers. The group chat gets excited. Then someone immediately says: "Let’s make the second one right away." There is nothing wrong with that. But over time, I have increasingly wanted to ask one more question at that moment: And then what? A video going viral and truly owning an asset are two entirely different things. I now use a very simple test: Hide the view count, then look at what is left of the project. If there is nothing left but an impressive number, it is more like a traffic event. If, once the view count is removed, there are still characters, relationships, a world, audience understanding, production methods, commercial capability, and even a system that can keep operating— then it is beginning to resemble an asset.
The most dangerous thing about a viral hit is that it can make you think you have already found the answer. Especially with AIGC. When a piece of content suddenly takes off, it is easy to attribute it to something specific: Was this visual style effective? Was this model effective? Was this subject effective? Was this prompt effective? So you copy it immediately. Make a second piece. A third. A tenth. But much of the time, the real difficulty is not reproducing the result. It is knowing: Why did this result happen? Landing on a result by chance and knowing how to achieve a result again are two different capabilities. The real asset often takes shape in the space between them.
As AIGC has rapidly lowered production costs, I have become increasingly concerned with what is left behind after each project. Because generation can easily create a bad habit: Use it once, then move on. This piece needs a character, so generate one. The next needs a different setting, so generate another. Make the voice again. Write the prompt again. Rethink the character relationships again. Every piece is fast. But every piece starts from zero. If that is the case, AI is only helping you start from zero faster. Real efficiency is not making things faster. It is allowing past production to become part of future production.
That is why the things worth preserving for an AI character go far beyond a reference image. Character definition. Voice. Relationships. Character Memory. Visual style. Recurring settings. Behavioral boundaries. Stories that have already happened. The understanding audiences have already formed. All of these should be retained. Then the next generation no longer starts from zero. It continues to grow within a world that already exists.
I now ask every content project to try to leave behind at least five things. First, characters. Has the audience remembered someone? Second, relationships. Have the characters formed relationships that can continue to develop? Third, a world. Are there rules, settings, history, or background that can continue to expand? Fourth, methods. After this success, what exactly did we learn? Fifth, audience understanding. Why do people follow? What do they expect to see next? These five things may be more worth reviewing than a single view count. Because views belong to yesterday. These things belong to the next production.
The most easily overlooked part is the method. After producing a viral hit, many teams save only the finished piece. But what may really need to be preserved includes: Prompts. Models. Parameters. Reference images. Shot templates. Failed versions. Reasons for revisions. Review criteria. Why this version was ultimately selected. Together, these things answer one question: How did the success happen? If one success does not make the next production easier, then its organizational value is actually quite limited.
Another asset that is especially easy to overlook is what lives in the audience’s mind. As an account matures, audiences gradually form some stable associations. They see a character and know who that person is. They see two people standing together and know their relationship. Before a sentence is even finished, they already have a sense of what may happen next. In a sense, a truly mature IP is cultivating what exists in the audience’s mind. That is why I increasingly see IP as more than "an image." IP is more like an understanding accumulated over time. With every piece you publish, you add something to the audience’s mind. Or you undermine something.
That is also why I increasingly like to distinguish between two terms: File and System. A File is a finished piece. An MP4. Its value lies in: How well this one piece was made. A System is different. Characters. Relationships. World. Voice. Visual rules. Story history. Generation methods. Workflows. Audience feedback. Commercial data. Copyright and licensing. Together, these things make up a system. The value of a System lies in: Whether it makes the next piece easier to do well. If a project has made one hundred videos and its system is still the same as it was on day one, those hundred pieces were likely just consumed. Nothing was truly accumulated.
That is why I have always been cautious about so-called "viral formulas." It is not that formulas are useless. It is that they can lead people to focus all their attention on "replicating the result once more." A viral formula asks: How can we raise the chance that this attempt succeeds? An asset mindset asks: How can this success make the next success easier? The former is concerned with results. The latter is concerned with accumulation. I care more about the latter.
This becomes even clearer when applied to MCNs. If the experience behind every account exists only in the mind of a single operator, then no matter how large the company becomes, it is still essentially a group of people each making content on their own. A real institutional capability should mean: A pitfall one person has encountered does not need to be encountered again by the second. A method validated on one account can directly benefit the next group of accounts. It means moving experience out of any one person’s mind and into the system. This may be one of the capabilities MCNs need to rebuild in the AI era. Because AI can rapidly raise individual productivity. If an institution cannot provide stronger systemic accumulation, why would an individual still need one?
So when we talk about scaling AIGC, the question I now most like to ask is: What exactly are you scaling? If what you are scaling is random generation, then more scale may simply mean producing more content clutter. If what you are scaling is validated character systems, methods, standards of judgment, commercial capabilities, and audience understanding, then it has meaning. Quantity is not industrialization. Repeatability, control, reviewability, and accumulation are what begin to resemble industrialization. I would even ask: Is this project becoming smarter over time? By that I do not mean the model. I mean the project itself. By the hundredth piece, does it know more than it did on the first? If not, then it has only produced one hundred times. It has not grown one hundred times.
Finally, there is one kind of asset that I think is hardest to replicate. Time. A character and its audience have traveled through a year together. Dozens of things have happened. The comments are full of shared in-jokes. Longtime followers know parts of the past that new followers do not. You cannot gain these things overnight simply by switching to a stronger model tomorrow. Others can certainly generate a character more visually striking than yours. But they cannot have, in one day, the year that you and your audience have lived through together. So the real moat of IP may not be that others cannot make what you make. It is this: They have not traveled that stretch of road alongside your audience.
So when an account goes viral now, I am still happy, of course. Data matters a great deal. Commercialization matters a great deal too. But once the excitement settles, I care more about several other questions: Who was remembered? What relationship was established? What understanding did the audience retain? What did the team learn? What can carry into the next piece of content? The real asset is not how much traffic today brings you. It is how much from today’s work will still belong to you tomorrow. A File is a result. A System is a capability. After a viral hit ends, I still ask the same question: And then what? If there is no answer, it may only be a beautiful firework. If there is an answer, the things that truly hold value may only just be beginning.
Copy this essay for someone thinking about AIGC content, IP, organisation or digital identity.
