---
title: "The ToB Veteran Who Fastest Hit 100M ARR — Why Is He Now Making 'Playable Content'?"
author: "Unique Research"
sourcePublication: "Unique Research Substack"
originalPublishedAt: "2026-09-14T16:16:07+00:00"
canonical: "https://ffcap.cn/en/research/the-tob-veteran-who-fastest-hit-100m"
source: "https://uniqueresearch.substack.com/p/the-tob-veteran-who-fastest-hit-100m"
language: "en"
---

# The ToB Veteran Who Fastest Hit 100M ARR — Why Is He Now Making 'Playable Content'?

[![](https://substackcdn.com/image/fetch/$s_!Pztc!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F63587ea3-156b-4f0d-a602-039636f5cff5_2048x1152.jpeg)](https://substackcdn.com/image/fetch/$s_!Pztc!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F63587ea3-156b-4f0d-a602-039636f5cff5_2048x1152.jpeg)

**Original by Unique Research | Published: 2026-09-14 11:50 (Shanghai)**

* * *

For decades, we’ve been lowering the barrier to “producing content.” The next truly important step is lowering the barrier to “participating in a world.”

If you only looked at his resume, Old Sun doesn’t seem like someone who would go into AI entertainment.

He was the first employee for Feishu People commercialization, building the sales and pre-sales system from scratch and setting the record for the fastest 100M ARR in the HR SaaS industry. Later, he led B2B business at a large model unicorn, building teams and driving business from zero, helping the company achieve nearly 100x valuation growth in two years.

This is a classic ToB veteran.

So when he started his own company and the direction became AI interactive content, my first reaction was: why?

He had already figured out the enterprise services business model. Clients, sales, solutions, delivery — he knew all of it.

Entertainment is a different game entirely: user attention is unpredictable, content costs are high, and viral hits are hard to replicate.

Moreover, interactive content isn’t a new concept.

_Black Mirror: Bandersnatch_ did it. Domestic interactive dramas did it. Chengguang has been doing it for years.

Why now?

After talking with him, I realized what Old Sun is really betting on isn’t “AI video,” or even just “video” — it’s the broader concept of “content.”

After 10 years in commercialization, he’s started paying attention to a question that seems less commercial: if the barrier to creation becomes low enough, what happens to content supply?

And ReelFork is his product validation of this question.

* * *

Old Sun’s career over the past decade seems to span many industries.

In 2011, he worked on big data and MarTech. Later came mobile internet, quantitative trading platforms, then enterprise SaaS and AI.

But in his own view, what changed was the industry; what stayed the same was one thing: he’s always been looking for scenarios where new technology can truly be applied at scale.

Over the past few years, the first thing AI changed was efficiency.

Writing code faster. Processing information faster. Making reports faster. Generating copy faster.

Essentially: the original tasks didn’t change, they just got done faster.

But this wave of AI is starting to show another interesting shift.

Many AIGC products still attract massive user usage even when their early productivity value isn’t fully mature.

The reason is simple: the new experience itself is enough to make people want to try.

This made Old Sun start thinking: as AI keeps pushing the experience barrier higher, where will the next true experience leap be?

He eventually set his sights on entertainment.

Because entertainment is inherently an industry that depends heavily on experience.

Users are willing to spend time, spend attention, and even pay for emotion, engagement, and novelty.

So the opportunity he sees isn’t just AI making content faster — it’s: will AI start changing “what content is”?

* * *

This question brings us back to the supply side, which he knows best.

When he worked on enterprise software, Old Sun was used to asking: how many clients can one system serve? How many opportunities can one salesperson cover? After automation, can the original costs continue to drop?

Essentially, he was looking at: when production costs fall, what happens to supply?

After AI entered the content industry, the logic is actually the same.

Over the past few years, what people first saw was: writing code faster, making images faster, generating videos faster.

But Old Sun is increasingly concerned with another question: not how much content one creator can produce in a day, but whether people who originally couldn’t create will start creating?

These are two completely different questions.

In the past, if a person had a story, a character, or even just an idea that flashed through their mind, and wanted to turn it into real content, there was a whole professional process in between: scriptwriting, shooting, editing, design, and more.

So the entry point for content production has always been narrow.

What AI might truly change isn’t just the efficiency of this pipeline — it’s compressing the distance between “I have an idea” and “I made it.”

As a result, content supply might no longer depend only on a small number of professional creators.

More ordinary people might also start becoming part of the content supply.

* * *

But the question quickly arises: once these people who originally couldn’t create actually start creating, what will they create?

If the answer is still: “Open a blank page and write a script from scratch,” then the barrier hasn’t really disappeared.

So what ReelFork wants to do isn’t give users a more powerful creation tool — it’s to move the starting point of creation one step forward.

Not letting users create a world from zero, but letting them enter an already existing world.

You like a character — you can continue interacting with them.

You’re not satisfied with an ending — you can try another path.

You suddenly think of a new possibility — you can continue developing it.

Many people, after watching a drama, have actually had similar thoughts: “What if they hadn’t chosen that way?” “What if they hadn’t separated — what would happen next?”

In the past, most of these thoughts would only stay in their heads.

Because from a thought to actually making content, there was still a whole professional process in between.

What ReelFork wants to do is compress this distance as much as possible.

You don’t even need to think through the entire story at the beginning.

You just need to say: “I want to see another possibility.”

The rest, AI helps you continue unfolding.

At this point, the original content is no longer just a consumer product — it becomes a starting point for creation.

Here, Old Sun proposes a standard he values highly: **causal agency**.

Not whether the user clicked or input something, but: did my action truly become the cause of what happened later?

If I choose not to trust her, and her attitude changes because of it, and the subsequent plot changes as well — then my action has entered the causal chain of this world.

But if my input just switches me to another pre-recorded segment of content, then it’s essentially still content selection.

So the difference between “watching” and “playing” can be compressed into one sentence:

**Watching** is when your input determines what you see next. **Playing** is when your input determines what happens next.

And when your input truly enters the causal chain of the content, interaction truly begins.

* * *

But one step beyond causal agency, there’s another question: can the content remember what you did?

The first time you help a character, the next time you meet, their attitude toward you should be different.

You changed the relationship between two people — the subsequent story should be affected.

You triggered an event — it shouldn’t just disappear into thin air.

At this point, the content is no longer just a fixed story — it starts having its own state.

Old Sun feels this might be where interactive content gets truly interesting.

Interaction isn’t about having a few more buttons or a few more branches — it’s about whether the content has a state about “you.”

“What you did, whether it remembers, and whether these things affect what happens next.”

Traditional games, of course, have had state systems for a long time.

The problem was never “there was no state before.”

The real limitation is: most of these states needed to be designed in advance by humans.

Creators can write 100 nodes, 1000 nodes, but it’s hard to pre-write everything a user might possibly do.

The value of AI is precisely here.

Before, it was: the author writes in advance “what might happen.”

Now it can become: the author defines a world, and after the user enters it, things truly start happening.

AI doesn’t necessarily just make more branches.

What it might truly do is: let this world be able to respond to behaviors that weren’t originally designed in advance.

* * *

At this point in the conversation, you can understand why Old Sun is unwilling to simply define ReelFork as an AI video product.

Because if the goal is just: generate more videos, more branches — then ultimately it’s still the logic of a video platform.

What ReelFork wants to explore is another form of content: content itself can be a character, a world, a relationship, or a state that continuously changes.

Creators are responsible for building the world.

AI lets this world continue operating.

Users enter it and make it change through their own actions.

And every interaction from users might also become new content.

This means the boundary between users and creators starts to blur.

A person might start out as just an audience member.

After watching, they suddenly think: “I want this story to happen differently.”

And then they become a creator for the first time.

This is also what Old Sun truly wants to validate:

Not letting original creators make 100 more videos a day, but letting someone who originally couldn’t create — because of a suddenly emerging idea — immediately create something.

If this holds true, the supply logic of content platforms might also change accordingly.

In the past it was: a small number of creators produce, most people consume.

In the future it might become: consumption itself continuously generates new content.

Content attracts users to enter.

Users interact with content.

Interaction generates new content.

New content attracts the next user.

As a result, content platforms might for the first time have a new flywheel: **content → interaction → creation → new content**.

This is also why what Old Sun is ultimately building isn’t “AI video,” or even just “AI entertainment.”

What he truly wants to try is an interactive content platform.

* * *

Looking back at Old Sun’s career over the past decade from this perspective, you’ll find he hasn’t really left what he’s familiar with.

Before, he did enterprise services, scaling sales, clients, and business processes.

This time, he’s focusing on another kind of supply: those “what if at that time...” thoughts that originally only stayed in people’s heads.

If technology can truly turn these thoughts into content quickly, then the way content is produced might change.

Perhaps in the future, after a user finishes watching a work, they won’t just say: “That was great.”

They’ll also say: “I want to try — what would happen if it were me?”

And when this “what if” can truly happen, content is no longer just seen — it starts being played into existence.

* * *

**Old Sun:** Actually, there’s a common thread in the things I’ve done: I’ve always been looking for scenarios where new technology can truly be applied at scale.

I started doing big data and MarTech in 2011, then went through mobile internet, quantitative trading platforms, and started doing AI applications in 2024.

On the surface, the industry span seems large, but the underlying logic hasn’t changed — for technology to truly create value, it ultimately has to land in a specific scenario that’s large enough.

Doing enterprise software over the past few years made me very confident that AI would first create a huge productivity value in enterprise services.

But at this wave, I’m increasingly optimistic about entertainment.

Because entertainment is a very special scenario: users are naturally willing to spend time, spend attention, and pay for emotion, experience, and engagement.

More importantly, AI is for the first time truly changing the way entertainment content is produced.

So it’s not “suddenly jumping from enterprise services to entertainment” — it’s that I feel: the most worthwhile thing to do in the last AI wave was efficiency, and the more worthwhile thing to do in this AI wave is experience.

And entertainment is precisely where experience is most concentrated.

**Old Sun:** Many business-side methods are actually universal.

In the early days of Feishu People commercialization, I built the sales and pre-sales system from scratch, creating the fastest record for 100M ARR in HR SaaS. Later, at Aishi Technology, I was responsible for the B-side, and also built a complete system from sales, pre-sales, CSM to product operations.

These experiences have made me always value one thing: for a new business, the most important thing isn’t doing everything well, but quickly finding the variable that truly determines whether the business lives or dies.

The biggest difference between ToB and ToC is still the way requirements are validated.

ToB can be quickly validated through clients, sales, and business results; ToC more often discovers requirements through product experience and user behavior.

So at ReelFork, I’ll pay more attention to: will users truly like it, truly play it — not what they say they like verbally.

And actually, even ToB isn’t completely “clients tell you the answer” anymore.

Models, Agents, Skills — these capabilities iterate too fast, and the impact on existing workflows and management methods is huge. Many clients themselves don’t know what the future way of working will be.

So a big change in this wave of entrepreneurship is: you can’t just be a requirement responder — you must also become a requirement guide.

ToC just has a shorter feedback cycle and more direct feedback.

Users won’t tell you: “I need a Stateful Content platform.”

They’ll only say: “That choice just now actually affected what happened later — I want to try another one.”

The product team has to reverse-engineer from this behavior what users truly want.

**Old Sun:** I’d say: ReelFork is a place where you get “playable content.”

Traditional content is for you to watch.

Games are for you to play.

What we want to build is the thing in between that has never truly been built before.

You don’t need to spend hours learning a set of game rules — you can truly change a story in minutes.

You enter a story, not just watching how it happens, but also deciding: what will he do? Will she trust you? What happens next?

So for users, it’s not an AI tool — they don’t even need to understand the technology behind it.

It’s just one sentence: this story will be different because of you.

**Old Sun:** Because simply improving production efficiency ultimately gives you more “watchable content.”

That certainly has value, but it’s not what we most want to do.

And ReelFork isn’t just branching narratives either.

We ultimately hope to support more content forms: text games, interactive dramas, light games, mini-programs, character interaction, and more.

Behind them all is actually the same thing: letting users go from content viewers to participants in the content world.

So what we care more about isn’t: “How many videos can AI generate?”

It’s: can AI let a content world continue happening because of user behavior?

That’s what we truly want to do.

**Old Sun:** The biggest bottleneck actually isn’t that users don’t like interaction.

It’s: interactive content is too expensive and too difficult to create.

A regular video only needs one storyline produced.

But interactive works need to pre-design multiple branches, multiple outcomes, character states, different endings, and the logical relationships between different states.

As branches increase, production costs rise rapidly.

So interactive content in the past was essentially still a very “handicraft-style” industry.

What AI truly changes is: the content space that originally required a lot of manual pre-production starts becoming a content space that machines can unfold in real-time.

That’s the key.

It’s not that AI invented interactive content for the first time — it’s that AI for the first time gives interactive content a chance to go from “a few boutique projects” to a type of content that users can directly participate in and that has vitality.

**Old Sun:** I think this is a very important standard for judging AI-native content: does the user have causal agency?

Not whether the user clicked a button, but: is my action the cause of what happened later?

For example, you click once and another video plays.

That’s just switching content.

But if I just chose “don’t trust her,” causing her trust in me to drop, and the entire subsequent plot changes — then the user has truly entered this content world.

So AI-native content shouldn’t just be: “AI-generated content.”

It should more be: content can understand users in real-time and continue changing because of user behavior.

That’s also why we keep emphasizing Stateful Content.

Because truly “stateful” content will remember what you did before.

**Old Sun:** The most common misunderstanding here is thinking that all interactive content must have all paths pre-produced.

In the future, we’ll clearly distinguish two types of content.

The first type is pre-produced branches.

Key plot points, key nodes, brand content — these can still be designed and controlled manually.

The second type is state-driven dynamic content.

What the user does, AI unfolds the next step in real-time based on the current world state.

So in the future it’s not: 100 states = 100 pre-produced videos.

It’s: rules + state + AI generation.

This also means the role of creators themselves will change.

In the past it was “write all the plot.”

In the future it will increasingly become: define this world, and the rules by which this world operates.

That’s precisely where AI is especially suited to help people.

**Old Sun:** I think the latter will become increasingly important.

Future directors don’t necessarily need to control every specific outcome.

They’re more like designing: what this world allows to happen, and what can’t happen.

For example, the most important value of D&D isn’t that the DM writes all the plot in advance — it’s establishing a complete enough set of world rules, then letting participants truly enter it.

So what we truly give users isn’t just “the right to choose.”

It’s: causal agency.

The director defines the world and boundaries.

The user decides what they do.

AI is responsible for keeping this world operating within the rules.

The truly important ability for future creators might no longer just be writing a very complete storyline, but whether they have the ability to build a world that can continuously grow.

**Old Sun:** I think there’s a big misunderstanding here: interaction shouldn’t equal making users constantly do exercises.

Users might indeed just want to lie down and scroll after work.

So ReelFork won’t require them to make a choice every 5 seconds.

What we truly focus on is: at what nodes is it worth giving causal agency to the user?

For example, a moment that determines character relationships, a moment that changes the entire story direction, or a particularly strong suspense point.

At these times, users will very naturally have a thought: “If it were me, what would I choose?”

At this point, interaction isn’t interrupting viewing — it’s adding a layer of engagement on top of viewing.

And I think another change AI brings is also important: it will unleash more ordinary people’s creative ability.

In the past, many people’s creative talent had no chance to be seen, because the barrier to expressing a complete story was too high.

Now that AI has lowered the production barrier, more and more people can enter content production.

In a sense, I think this is also a kind of “people’s history.”

Good content should originally come from all corners of life, different social classes, not always be controlled by a small number of people.

**Old Sun:** Because creation and consumption in ReelFork aren’t actually two isolated stages.

If it’s just: creation tool → export video → go play on TikTok, YouTube, or other platforms — then we’re essentially still just a production tool.

But what ReelFork truly wants to build is a complete content loop: creators create a world → users enter and interact → user behavior changes state → new content continues to be produced → new users continue to enter → users and creators continue to Fork.

So Cinema isn’t an extra player.

It’s a very critical part of the entire Stateful Content loop.

Without the consumption side, user behavior won’t feed back to the content.

And without this feedback, it’s hard for interactive content to truly form a flywheel.

So in the early days, we’ll first focus on solving creation supply.

The most core asset of a content platform is ultimately still content.

But we also won’t go the traditional platform way of “first pull in a bunch of creators, then wait for users to come.”

AI gives us the opportunity to first work with the first batch of creators to produce a batch of representative content at relatively low cost.

First get this chain working: good content → user experience → user feedback.

Once we have the first batch of paradigms, then let more creators enter.

**Old Sun:** I think what will truly be scarce in the future is definitely not generation capability.

Models will get cheaper and cheaper, and generation capability will become more and more common.

If it’s just “infinite generation,” it will very easily become a content garbage dump in the end.

What’s truly important isn’t: “How much content can you generate?”

It’s: can a good piece of content continue to spawn derivative works.

That’s also why we pay special attention to Fork.

A work isn’t a fixed endpoint.

It can be continued by users for secondary creation, adapted by different creators, and give rise to different characters, different perspectives, different gameplay.

So the true value of AI isn’t: “produce 1000 pieces of content for you.”

It’s: let a good content asset have the ability to continue growing 1000 kinds of content.

We always say one sentence: one IP, one universe.

Further on, what’s truly scarce is actually content relationships.

Why do users follow a character?

Why do they continuously follow a world?

Why are they willing to Fork a work?

Why are they willing to continue paying for a character?

In the future, more and more vertical content platforms will definitely emerge.

The value of a platform isn’t just: “there’s a lot of content here.”

It’s: here’s a group of people I identify with, a batch of characters, a few worlds I truly care about.

Once users establish relationships with content, the platform’s stickiness, willingness to pay, and loyalty will be very different from simple content consumption.

**Old Sun:** If looking at metrics, I’m now paying special attention to three levels.

First: are users truly playing.

Not how many times they clicked, but whether true causal interaction has happened.

Second: will users come back because of the causal relationships in the content.

For example, replaying, exploring multiple endings, or going back to a previous node to make a choice again.

These metrics prove whether interactive content works better than simple dwell time.

Third: can creators continuously produce.

Because the platform ultimately must form: good content → more creators → more content → more users.

So what I least want to see is everyone coming in for the first time and thinking: “This thing is pretty novel.”

What I truly want to see is users, after playing, developing an impulse: “I want to see what happens with another choice.”

At that point, I think the product has truly worked.

From an industry stage perspective, I think now is much like the intersection of the early mobile internet era and the early explosion of large model applications.

At the beginning, there will definitely be a lot of “original product + AI,” which is normal.

But when I judge whether a company has truly built product barriers, I look at three things: is there new user behavior; is there new content or product form; is there new network effect or asset accumulation.

If it’s just using AI to produce past content faster, once the model iterates, this advantage is very easily eaten away.

But if a product starts accumulating user behavior data, character relationships, Stateful Content, creator networks, and IP assets, then it’s actually not just eating model dividends — it’s building its own product layer.

So if ReelFork truly succeeds in three years, I don’t want it to just be an AI creation software, and I don’t want it to just be an interactive short drama platform.

I more hope it ultimately becomes: a new infrastructure for content consumption and production.

But that’s what we say internally.

For ordinary users, it doesn’t need to be this complicated.

They just need to know: there’s a lot of playable content here.

They can watch an interactive drama, chat with a character, play a text game, enter a story world, or Fork someone else’s work.

For creators, it’s: I give you a world, a story, a character — you help me turn it into content that can continuously grow.

So my ideal ReelFork isn’t an “AI content tool.”

It’s: letting content go from “a work to be watched” to “a world that can be entered, changed, and continued to be created.”

True AI-native entertainment shouldn’t just be “AI helps us make more content.”

It should be: content starts being able to understand users, remember user behavior, and continue happening because of users.

For decades, we’ve been lowering the barrier to “producing content.” The next truly important step is lowering the barrier to “participating in a world.”

I think ReelFork stands exactly at the intersection of these two trends.

When I first started a business in 2011, technology was helping enterprises process data; in the mobile internet era, technology was helping people connect with each other.

In the AI era, I increasingly believe that technology will ultimately return something even more ancient to ordinary people: creating worlds, and participating in worlds.

That’s also why I’m willing to start a business again now to build ReelFork.

* * *

_Source: 非凡产研 (Unique Research) WeChat Official Account_  
_Original URL: https://mp.weixin.qq.com/s?\_\_biz=MzU5Mjg5MjQ5Ng==&mid=2247522685&idx=1&sn=8fca80399ac136cef6ac2b1f0bbaa637_  
_Translated for Unique Research Substack publication_

---

Original publication: https://uniqueresearch.substack.com/p/the-tob-veteran-who-fastest-hit-100m
On-site reading page: https://ffcap.cn/en/research/the-tob-veteran-who-fastest-hit-100m