2026 AI Gaming Panorama: Four Layers, Three Myths, One Consensus Gap | A Talk with 405 Games' Xiaoning
Lots of fragments, no smash hit yet.
👦🏻 Podcast Interview: Koji
🥷 Edited by: Crossing
🧑🎨 Layout: Zeoooo
🚥 "AI + Interactive Entertainment / AI + Gaming" is at a fascinating juncture: fragmented demos are popping up everywhere, the pace of technological progress is almost disorienting — AI-generated mini-games, interactive short-form content, AI NPCs, world models, real-time multimodality — yet breakout hits that achieve genuine mass-market consensus remain scarce.
For this episode, Crossing teamed up with "405 Gaming," a podcast focused on AI interactive content and games, and invited host / veteran game industry professional Xiaoning Liu for a "panoramic scan" of AI gaming in 2026: we break the industry down into "four layers," identify three of the most common misconceptions, and discuss what that most critical consensus gap actually is — and why it's more likely to emerge from wild, organic soil than to be "greenlit into existence" by a major studio.
If you're an AI entrepreneur, investor, or someone tracking next-generation content platforms, interactive entertainment, and AI games, this episode aims to give you a framework: what's settled? What's still hallucination? What are the most important variables to watch in the next phase?
Listen on WeChat:
Listen on Xiaoyuzhou:
🎬 Video podcast now live on Koji's WeChat Channels, Xiaohongshu, Bilibili, YouTube, and other platforms
Opening
👦🏻 Koji
Xiaoning, you used to be a producer at a top-tier gaming studio. What prompted you to leave and start a podcast?
👩🏻 Xiaoning
I was at a leading game studio, working as a producer on a fairly well-known UGC game. After I left last year, I was talking with my former boss and realized we were both fixated on the same question: with AI's arrival, what changes and impacts would it have on interactive entertainment and gaming? At the time, we had a strong sense that the gaming industry's response to this wave of generative AI was noticeably slower compared to internet companies.
We really wanted to create some cross-pollination between AI folks and game folks. So we started the podcast as a conversation starter, and we've kept it going ever since.
👦🏻 Koji
"405 Gaming" is a weekly podcast — nearly 50 episodes in one year. Speaking of AI plus interactive entertainment plus gaming, which sub-areas are you personally most focused on?
👩🏻 Xiaoning
Initially we positioned the podcast as "AI gaming," weighing games extremely heavily. But we gradually shifted to "AI interactive entertainment." Because we kept asking ourselves: what exactly is our core boundary? Eventually I realized it boils down to just two things:
First, interactivity. If a user is just watching a drama, an animation, or looking at images, that's fundamentally different from "interactive" in our view.
Second, fun. People invest time, attention, and money ultimately to satisfy their emotional experience — and specifically, a joyful one.
If we're thinking three, four years out, I believe games aren't competing against other games. They're competing against Douyin. When users spend their leisure time, they don't categorize whether something is a "game" or not. They look at whether it delivers satisfaction and fun in the moment.
So we narrowed our boundary to two words: interactive and fun. And that actually broadened our focus, positioning us at AI plus interactive entertainment.
👦🏻 Koji
Speaking of interactive and fun, games do come to mind first. But thinking more carefully, Douyin has similar properties too. Swiping up and down, commenting, arguing in the comment section — those are all forms of interaction?
👩🏻 Xiaoning
Yes. We mentioned in one episode that if you map interaction depth on a single axis, the tipping point might be Douyin filters. A step further would be danmu (bullet comment) interaction. And games are definitely the deepest endpoint of interactivity.
The Industry's Four-Layer Landscape
👦🏻 Koji
When we talk about AI plus interactive content now, what kind of landscape has this industry developed into?
👩🏻 Xiaoning
I think it can be broken into four layers.
Layer one is AI as a tool serving the creation of entertainment and interactive content — efficiency gains in the production pipeline.
Layer two is AI as a creative entry point, enabling content creation that wasn't possible before.
👦🏻 Koji
What's the difference between these two layers?
👩🏻 Xiaoning
Layer one is more about improving efficiency within existing production workflows. It doesn't necessarily mean a complete, standalone creation. For example, current AI coding helps us with more infrastructure-level efficiency gains.
Layer two is where you're actually using AI to generate a short film or other content. Here, coding itself is hidden — what we care about is the result.
Layer three is AI as an interactive object. This is what we see the most of right now, like AI NPCs.
Layer four is where AI changes entertainment relationships. What we often call "creation as consumption" is essentially reconstructing who creates and who consumes.
Some recent experiments don't directly use AI as an interactive object, but instead place AI within social relationships to facilitate or modulate atmosphere. It's changing entertainment-social relationships. So all four of these layers currently coexist.
The Eye-Opening Works
👦🏻 Koji
Within these four layers, can you think of some representative products or works?
👩🏻 Xiaoning
For layer one efficiency gains, AI coding is definitely an important representative.
For layer two as a creative tool, things like generating video, generating images, generating games — all fall into this category.
For layer three as an interactive object, companion products are a major representative. Starting with Character AI, and more recently Eve which just entered open beta, plus AI NPCs in games.
The fourth layer around entertainment relationships is still relatively exploratory, but you can see some attempts, like miHoYo's Stellar Valley.
👦🏻 Koji
Can you expand on that?
👩🏻 Xiaoning
Stellar Valley is a life simulation game. The NPC Nalo has an Agent attached to her — she's a coffee shop owner. In the game, you can choose to chat with Nalo one-on-one in a private room, or enter a public room where the coffee shop has you plus maybe three other players, but Nalo is still Nalo.
What you discover is that in a multi-player plus one Agent scenario, the Agent's role starts to shift.
👦🏻 Koji
So is the core experience of this game the chatting?
👩🏻 Xiaoning
Chatting is just one component. Since it's a life simulation game, you still need to build houses, live on the island, and do other things. But it provides a coffee shop scene that allows players to engage in deeper conversation.
And this chatting isn't limited to talking with the Agent — it's actually facilitating social interaction between players.
👦🏻 Koji
Has there been any AI-plus-interaction content that's recently made you go "wow" — whether a product or a game?
👩🏻 Xiaoning
Stellar Valley which I just mentioned would be one. In that coffee shop scenario, the role AI plays has already changed. Answering questions is no longer the sole focus. On one hand, it's demonstrating how to chat in this coffee shop — new players see how it communicates with others, and learn how to participate.
Second, it can rescue the conversation when things go cold, modulating the atmosphere. If you factor all this in, it's no longer a simple NPC role — it's social infrastructure.
In different scenarios, what AI needs to pay attention to changes. When I chat with someone in Character AI, a good experience means I say one sentence and it says ten, or at least speaks more than me — I give a little, and it gives me more feedback.
But in social settings, the question becomes: when should AI shut up?
So the problems people are solving have already diverged.
👦🏻 Koji
I recently came across a Nintendo game called Tomodachi Life that I found fascinating. It also has everyone living on an island. First you can create yourself and your friends, making them look very realistic, or even create weird animals and monsters.
You can assign several people to live together — one house supports up to 8 cohabitants. In this shared house, you can also set up drama plots. For example, set it so all 8 people are in love with the same middle-aged woman, and they'll start acting according to your setup. You might check back half an hour later and discover they've performed an incredibly dramatic soap opera.
This process can also be recorded as short videos for secondary distribution. Inside, you're simultaneously the world's creator, the director, and the audience — lots of room for creativity.
👩🏻 Xiaoning
That's because there are precedents. The first founder who mentioned precedents to me was an entrepreneur named Patrick. Their product, Bside, was actually exploring this direction. But after trying it myself, my biggest impression was that it still struggled to keep people playing for extended periods.
The first hour or two felt great, but the positive feedback loop afterward wasn't strong enough. Actually, many town-simulation-style games are exploring this direction too. Last year everyone was watching Aivilization, but it faced the same problem — people just felt like it wasn't quite fun enough.
There are plenty of fragmented small experiments, but products that can sustain repeated, longer-term engagement remain rare.
👦🏻 Koji
Speaking of AI plus interactive content, does anything else come to mind that caught your eye?
👩🏻 Xiaoning
Stardew Valley: Starlight that I mentioned earlier counts as one. Another thing everyone's watching is AI-generated mini-games. One point I personally care about is how it might change what we mean by "creation" in the first place.
No matter how you approached interactive entertainment before, you had to deliver a complete experience. But on current creation platforms, what you're creating is an entry point — and everyone else remixes from there. The degree of completion, and what kind of experience you're actually creating, are both shifting.
👦🏻 Koji
Sounds like there aren't that many complete AI-plus-interactive-content products that really wowed you?
👩🏻 Xiaoning
If we're talking fragments, there are plenty. For example, a team recently built something where you could walk into Along the River During the Qingming Festival and interact with it.
Another particularly interesting one I played on a platform was a Snake game. Snake is simple, but this version had lots of characters on it, and you had to eat them to form a sentence, then have AI judge whether it was a good sentence. That was the hardest Snake game I've ever played.
There are tons of fragmented things like this. But if we're talking about complete-experience products, there really are very few right now.
👦🏻 Koji
A product or game that could grow into something massive — we haven't really seen that yet?
👩🏻 Xiaoning
Right. I wouldn't say massive, but complete-experience products are indeed relatively scarce. Though fragments are everywhere. Another example: a historical simulator called Chongzhen that launched recently, where you time-travel back to become the Chongzhen Emperor.
The AI Angry Birds hasn't appeared yet
👦🏻 Koji
When we talked early this year, you mentioned that the current moment felt like a year and a half before Angry Birds came out. How do you understand that now?
👩🏻 Xiaoning
That was probably at one of your open mic events. What I mainly wanted to express was that in AI interactive entertainment, the consensus-forming product hasn't emerged yet.
I used Angry Birds as an example because it's so concrete. It was the first product that made everyone realize that interactive entertainment on large touchscreen phones could deliver a new kind of experience. But if you trace the entire history of mobile games, Angry Birds was the beginning, not the end. It wasn't even the most commercially successful work. But people remember that moment.
Because after it, the industry formed a consensus: this direction is worth investing more resources to explore. And under the proposition of AI interactive entertainment or AI games, that moment hasn't happened yet.
👦🏻 Koji
What conditions do you think would need to be in place for that moment to arrive?
👩🏻 Xiaoning
When Angry Birds appeared, it was roughly two years after the first iPhone launched and one year after the App Store opened. So it needed some preconditions.
First, the infrastructure was already there. Like large language models today — everyone is starting to use them, they don't feel foreign anymore.
Second, content distribution methods already existed. People could access this kind of content, and some consensus had formed around infrastructure and habits.
So the reason I made that analogy was that I felt some infrastructure is already in place now, but many things are still being explored, and the technology is still developing.
👦🏻 Koji
Angry Birds was a mass-market game that everyone was installing and talking about overnight. Its lifecycle was short, and it had few paying users, but it proved that phones could deliver genuinely fun games.
👩🏻 Xiaoning
Exactly. Just imagine: if you played Angry Birds on a computer, would it still be fun? Because a mouse is really designed for precise clicking — it's not suited for dragging, pulling, physics-simulated actions. That mode only works on phones, where the intuitive feedback is especially strong and it becomes extraordinarily fun.
So what we mean by AI's "Angry Birds moment" should be a game with a completely new experience — one that simply doesn't work without AI, or where AI amplifies some core experience by several orders of magnitude.
👦🏻 Koji
Do you see any similar shoots emerging now? A game that's become significantly more fun because of AI, by several orders of magnitude?
👩🏻 Xiaoning
Because the current focus is mainly on large language models, the industry's consensus attempts are still concentrated on AVG-style games. Everyone will definitely have that kind of experience.
But the problem is, AVG games are still in a very narrow domain. When most people play games, they don't interact through natural language — they experience through multimodal audiovisual stimulation, click feedback, tactile feel, and so on.
So I don't think it will necessarily emerge as completely new gameplay or in traditionally understood game forms. At this point in time, besides language dialogue-based experiences, we really haven't seen particularly clear shoots. Because if we had, consensus would probably have already formed.
👦🏻 Koji
From early 2026 until now there have been quite a few breakthroughs. Whether it's Agents becoming capable of long-horizon task planning, or advances in coding — what boost has this given to AI plus interactive content?
👩🏻 Xiaoning
The first change is the level of attention from game companies. I was chatting with an industry person the other day who said that around July or August last year, the feeling was still "our jobs aren't that easily replaced, AI is just somewhat an efficiency tool for certain environments." But now everyone is thinking holistically about how much AI can actually do.
Probably because AI in August last year and AI in April this year are simply not the same thing anymore.
👦🏻 Koji
What about world models? Whether it's Google's Genie 3 or corresponding domestic models — do you think they've brought new tools or new works to game makers?
👩🏻 Xiaoning
I think it's too early. But the coolest thing about world models is that they make generating an "enterable, influenceable" world possible. This makes us feel that Westworld or the Oasis from Ready Player One are getting closer, which is incredibly exciting.
But the unsolved problems they face are also extremely numerous, and there's a long road before they're practically usable. Also, the biggest imaginative space for world models actually isn't in games. Not long ago, DeepMind seemed to be using an offline version of EVE Online to train AI. So game worlds have instead become a great scene and source material for AI training.
👦🏻 Koji
Right, especially for training robotics, training embodied AI — it can generate synthetic data and also provide an environment.
👩🏻 Xiaoning
Exactly. I think that's the more starry-sea direction, but it's still very far from game implementation.
👦🏻 Koji
What about real-time multimodal? Being able to generate sound, video, or entire scenes in real time — are there any practical attempts yet?
👩🏻 Xiaoning
There are definitely attempts. Before, playing a game meant entering a pre-crafted world. Now the imagination is that you can enter a world and real-time simulate its story, interact with it in real time.
People more or less like a sense of control, and the imagination this enables is beautifully spacious. But it's not what I'd call practical implementation. Because games and interactive entertainment are fundamentally about delivering a stable, continuously interactable system, and this system needs to be carefully designed. So like world models, it represents a beautiful future, but hasn't landed yet.
So the industry is currently in a state full of attempts but lacking consensus.
Big Tech AI Strategies
Tencent, ByteDance, miHoYo, NetEase — facing AI, the four companies are taking four different paths.
👦🏻 Koji
Should we gossip about the big companies? What are ByteDance, NetEase, Tencent, miHoYo and others currently doing in AI plus interactive content and games? What are their respective advantages?
👩🏻 Xiaoning
These companies are quite different. Tencent and ByteDance both have model training foundations within their corporate systems. Reflected in games, Tencent is relatively more conservative, while ByteDance reacts very quickly.
miHoYo, because of its "tech otakus save the world" ethos, has its own faith in technical foundations and is willing to make big investments in technology. You can already see attempts in Stardew Valley: Starlight and new products.
NetEase has probably changed the most. It wasn't that proactive about AI applications before, but this year public information shows very active embrace.
👦🏻 Koji
miHoYo has been consistently proactive on the technology front. Any other clues you've seen?
👩🏻 Xiaoning
From public information, Cai previously tried a lot of things, even approaches that touched the underlying model layer — that's a very rare level of investment for a pure game company.
Brother Wei also stated in speeches that they will continue making big investments in technology.
It's not just big companies. The other day Game Grape had an article about G-bits — they had employees stop their regular work for a week to participate in an internal Vibe Coding competition. All game companies are exploring AI's boundaries and capabilities in their own ways.
👦🏻 Koji
Do you think if AI's Angry Birds moment arrives, it would emerge from these Chinese gaming giants? Or is it more likely to be born from non-big-company entrepreneurial environments?
👩🏻 Xiaoning
I think the probability is higher for it to grow in the wild. Because big companies sometimes overlook small creative ideas. If you tried to make a tiny experience at Tencent, you might not even be able to launch it — getting project approval would be hard. But outside, you might blow up on Xiaohongshu or somewhere else.
👦🏻 Koji
Do you think there are still classicists out there — people who insist on handcrafting and believe AI won't change anything?
👩🏻 Xiaoning
There definitely are. But people will objectively collaborate. For example, a programmer might feel AI writes code worse than themselves, but they can use AI for art and design; an artist might feel AI draws worse than themselves, but can use it to write programs.
Selectively letting AI handle what you're not good at, turning it into a collaborative relationship — that's a worthwhile experiment too.
👦🏻 Koji
Have you observed major game studios actually reducing headcount because of AI efficiency gains?
👩🏻 Xiaoning
They definitely have in many foundational roles. People feel it acutely.
👦🏻 Koji
What advice would you give to game industry workers in grassroots roles who might feel threatened by AI?
👩🏻 Xiaoning
I'd tell them to use AI more, to embrace new things. From talking with friends around me, I've found that the more you use it, the less anxious you feel.
👦🏻 Koji
Why does using it more reduce anxiety? Is it because after using it, you discover it's not that powerful?
👩🏻 Xiaoning
It's because after using it, you know what it can and can't do — where its boundaries lie.
When you find the points where it can help you do better, it stops being a threat. It's only a threat when you constantly feel like it's competing for your job.
Three Blind Spots of AI People Who Don't Understand Games
👦🏻 Koji
I get the sense that startups in the AI plus interactive content space today rarely come from gaming backgrounds. People who make games seem to feel they can only make games their whole lives, that they can't do anything else. You mentioned before that you feel there's some kind of gap between people doing AI and people making games?
👩🏻 Xiaoning
I think any two industries probably have some mutual misunderstanding or unfamiliarity.
👦🏻 Koji
Most of our listeners are still AI entrepreneurs and practitioners. If someone is interested in AI plus interactive content or AI games, what blind spots need to be addressed when you say AI people don't understand games?
👩🏻 Xiaoning
For example, people often talk about "AI-generated games." The word "generate" actually underestimates the industrial complexity of games. Videos and images are indeed generated, because once generated, they're finished products ready to be viewed.
But games are interactive experiences. They need to stably deliver an entire interactive system. The feedback mechanics, operating systems, level design inside — all of it needs debugging and design. It's a polishing process, not a one-shot generation.
👦🏻 Koji
For games, I often say it's a precision instrument. The manufacturing difficulty of a precision instrument is real.
👩🏻 Xiaoning
Right.
👦🏻 Koji
What's the second aspect?
👩🏻 Xiaoning
Second is fun. Fun isn't always designed in one go. When we make games, we talk about making an eye-catching demo versus making a product people come back to repeatedly or play for a long time — these are completely different dimensions of difficulty.
So it usually starts with a good demo, then testing, feedback, adjustment — it's an iterative polishing process, not a one-time design.
Also, I think people sometimes overestimate the role of natural language in interactive entertainment, especially on the interaction side. When we play a game, the tactile feedback of clicking and other experiential elements have nothing to do with natural language.
When we talk about an AI product, we care a lot about whether it's stunning at first glance, whether the first try is fresh and exciting enough.
But I think as an interactive entertainment platform, as interactive content, what matters more is: Why do users keep coming back, why do they stay and keep playing, why do creators keep creating here, and how should content be distributed and discovered.
These, I think, constitute the core value of a content platform — what we call retention. For many content products, "generatable" might be the end goal they're chasing; but for games and interactive experiences, "generatable" is only the starting point of what can be designed.
"The Douyin of the AI Era" — Does That Hold Up?
Loopit, AIPPY, Rezona, Riffle… everyone's saying they want to build the next interactive content platform.
👦🏻 Koji
This reminds me of Loopit, AIPPY, Riffle and others that have been very hot in the capital markets lately. What do you think of this phenomenon where everyone wants to build "the next-generation AI-era interactive content platform"?
👩🏻 Xiaoning
People pile in because they've seen real possibility — the creation entry point has changed. Before, creating an interactive experience was the privilege of internet product managers; ordinary people could barely participate. Now the creation threshold has been massively lowered, and what comes out does look fun at first glance. But two problems haven't been fully solved:
First, if everything is generated, how does human value get involved? We both generated a Snake game — how is our unique value seen? If the creator's value can't stand out, why would they keep producing here? On Douyin, you might watch ten thousand people do the same dance ten thousand times, because each person dances differently — you're there for that person. But in a purely generative interactive model, how does individual unique value manifest?
Second, who exactly is the consumer? After content supply is massively expanded, what kind of person wants to independently consume this content on a separate platform? Why isn't this just a subset of Douyin? How does it need to change content formats or distribution methods to break free from being a Douyin subset? These are questions people are exploring but haven't fully answered yet.
But regardless, the next-generation interactive content platform definitely won't look like Douyin.
👦🏻 Koji
But I'm thinking — historically, have any products that set out from day one to be "the next [something]" actually succeeded? It feels like if your day-one goal is to become the next whatever, you usually don't make it.
👩🏻 Xiaoning
Founders probably don't truly want to build the next Douyin deep down — this is just rhetoric that makes it easier for capital markets to understand.
Since mobile internet, the two greatest business models have been games and advertising.
The games business model pursues extremely high ARPU through sufficiently immersive and repeated interaction; the advertising business needs to solve why users keep coming back, with DAU needing to reach at least 30 to 40 million upward.
If users don't keep coming back, neither of these two points holds up in a platform.
👦🏻 Koji
So we've seen many viral hits in AI plus short films already. For example, while we've been recording this podcast, I saw something called Zombie Scavenger — a young person made a 3-minute short with 3,000 RMB and 10 days, and it's been well-received domestically and internationally. But AI plus interactive entertainment rarely sees this kind of massive viral hit.
👩🏻 Xiaoning
I think it's just not the right time yet — video content is further ahead. What amazed me most about vertical dramas was their growth speed. A month ago when I watched season one, the visuals were still rough; half a month later they'd already updated four or five seasons with very good quality. The technology is developing very fast.
On the interactive side, there are starting to be some fragmented products, especially video-based ones. Because Seedance only got popular around Spring Festival, after all. We haven't seen a mass-market consensus form recently, but the long-term outlook is definitely optimistic — you can't deny its development speed just because consensus hasn't formed yet.
👦🏻 Koji
When you mentioned vertical dramas, it reminded me — I recently talked with an AI short drama entrepreneur who said AI short dramas and live-action short dramas are changing production relationships, which could potentially disrupt the industry. Originally, to shoot a real live-action short drama, the cost of shooting episode one might already consume more than half the entire crew's budget, because you have to fly everyone to a location. You have to block out all the actors' schedules going forward before you actually start shooting, so the cost of creating a new work is very high, and the risk is substantial.
But AI live-action short dramas have no physical constraints — you can infinitely make episode one and throw it out to test, and only continue if episode one works. This was impossible in the traditional film and television industry. Are there similar production relationship changes in the gaming industry or AI interactive space?
👩🏻 Xiaoning
This is analogous to the 3D pipeline in gaming. In China's game market, 2D was considered to have a relatively low ceiling, while 3D had a higher ceiling. But making 3D blockbusters used to require extremely high barriers — only major studios could afford it.
Now the underlying AI 3D pipeline isn't fully solved, but if AI 3D barriers come down in the future, we could also do infinite episode-one trial and error like short dramas. When that happens, people will compete to create something else entirely.
👦🏻 Koji
You mentioned there are many non-consensus areas in AI plus interactive content and AI plus gaming overall. Is there any consensus that's already formed, beyond just beautiful plans for the future?
👩🏻 Xiaoning
The first is that long-term consensus exists — that AI will definitely bring some changes to productivity and production relationships. I think that consensus is there now. On the application side, I think the only thing that really qualifies as consensus is still on the companionship side.
People do have this consensus, because after all, it's still language models. Language models bring the most intuitive language experience. But everything else, I don't think rises to the level of consensus yet.
👦🏻 Koji
Do you believe that one day there will really be an embodied robot at home providing companionship?
👩🏻 Xiaoning
I actually find that scenario a bit terrifying when I think about it, but I think it's possible. I don't know if I'll see it in my lifetime.
👦🏻 Koji
Because many people think this might be something a game company makes.
👩🏻 Xiaoning
Why?
👦🏻 Koji
Because what robot companies make might just be functional — but not necessarily emotional, interactive, content-rich, or IP-driven. And those latter points are what game companies are good at.
👩🏻 Xiaoning
But isn't it possible these two needs could be separated? For example, a robot that helps me mop the floor — I don't necessarily want to fall in love with it. They might be two different things.
👦🏻 Koji
Do you know of any domestic game companies working on this?
👩🏻 Xiaoning
Not at the embodied intelligence level, but many are trying small desktop interactive toys. For example, Kingnet tested something before.
👦🏻 Koji
If we say AI plus interactive content is a complete field right now, what do you think is the most easily overestimated thing in this field?
👩🏻 Xiaoning
The most easily overestimated thing? First, "one-sentence game generation." Generating a prototype really is a one-sentence job. But tuning it into something you actually want to play might take 100 sentences, 100 rounds of complex back-and-forth.
👦🏻 Koji
And by the fifth or sixth round of interaction, the LLM's context window and instruction-following usually start to break down.
👩🏻 Xiaoning
Second, the weight that natural language carries in all this.
Third, the overestimation of "infinite choice equals fun." Agency and control in interactive entertainment both need to be designed. Take battle royale — the reason it could explode into something like PlayerUnknown's Battlegrounds is that ever-tightening "blue zone."
The blue zone constrains the uncertainty of the open world, giving players a clear directional goal. Only then does the fun system actually work. And the open worlds we love — aren't they all meticulously designed open worlds?
👦🏻 Koji
So do you think there'll come a day when someone scrolls Douyin and sees content completely different from everyone else in the world? Like Douyin generating content just for them?
👩🏻 Xiaoning
Possibly. But would that actually be fun? We're also afraid of "information cocoons" — everything generated just for me isn't necessarily optimal.
Often the strongest stimulation comes from a common denominator. Humans are more alike than different. A lot of these micro-trends, these tiny viral moments — they're essentially reliable dopamine hits. Life already has too much uncertainty.
👦🏻 Koji
That's also like memorizing English vocabulary.
👩🏻 Xiaoning
How so?
👦🏻 Koji
When life isn't going your way, you go memorize vocabulary. Because it's the easiest way to feel like you're making small, certain progress. I memorized 20 more words today — that's unambiguous progress. Dopamine releases, and you feel like you're regaining control of your life.
👩🏻 Xiaoning
Exactly. Positive emotional feedback for humans depends on this sense of control. So it can't be something completely uncertain.
👦🏻 Koji
Are there specific questions you're particularly committed to tracking long-term?
👩🏻 Xiaoning
First: whether there's room for short-form content in interactive entertainment, and what form it might take. Do traditional games' day-two retention metrics have more room to breathe in the AI era?
Second: AI 3D. If small teams gain access to 3D pipelines, what happens? The speed and methods by which AI 3D solves topology and rigging.
Third: watching for AI's "Angry Birds moment" — when and in what form of content it arrives.
👦🏻 Koji
Do you think there's an opportunity for a creation platform built around, say, a dedicated editor?
👩🏻 Xiaoning
You mean something like Roblox?
👦🏻 Koji
Yeah.
👩🏻 Xiaoning
I think in a large platform, the editor is the necessary foundation, but probably not the deciding factor. For a platform to work, the core is solving who you serve, what need you address, why users come back, and how content gets distributed.
These operational-level questions, I think, are what really answer how a platform gets built. Roblox actually grew quite slowly — it exploded through specific timing and circumstances. That growth path would be hard to replicate today.
TapTap Maker: Polarizing Reactions
People with design experience: "Incredible." People without: "Might as well use Claude Code."
👦🏻 Koji
There are actually quite a few companies building dedicated editors now, right? Using editors to create all kinds of AI interactive content.
Have you seen any you think are doing well?
👩🏻 Xiaoning
I don't think anyone's clearly pulled ahead yet, but the exploration directions are very different. Take TapTap Maker — I tried it when it first launched, found it not great, and set it aside.
Later I recommended it to some developer friends. They came back saying TapTap Maker has a very high ceiling, that it's incredible — someone recreated a Minecraft world in it, others built complete roguelike games.
I picked it up again and finally understood them. It really gets game language. But it's extremely polarizing. People without design experience find it worse than Claude Code.
While people with design experience, after a few more alignment rounds and tweaks, find it uncannily good at grasping game language and needs — it produces exactly what you want.
👦🏻 Koji
So what specifically does it do?
👩🏻 Xiaoning
My guess is they've built a lot of templates and scaffolding. They also made the Spark Editor before, with creators feeding and training content for the AI. So the AI understands game creators' needs and expression patterns better, and gives better feedback.
👦🏻 Koji
Are there any games made through TapTap Maker that you've seen and thought were genuinely good?
👩🏻 Xiaoning
Depends on your standards. If you look at efficiency gains and completion level, then absolutely — and they publish directly on TapTap. For example, something that used to take a team two or three months to recreate Minecraft, now one or two people can do in a week or two.
This circles back to what we said is underestimated: the impact of prototype production speed changes is severely underestimated. The dramatically lowered barrier to trial-and-error, the easier establishment of iteration loops — even if AI can currently only do prototypes and demos, its impact on the industry will far exceed the demos themselves.
👦🏻 Koji
Besides TapTap Maker, have you seen other editors you find interesting or worth sharing?
👩🏻 Xiaoning
There are many similar ones. Including Yoroll, which I mentioned before — they're doing AI interactive film-games, another different exploration direction.
👦🏻 Koji
AI interactive film-games?
👩🏻 Xiaoning
Right, this probably follows similar logic to AI short dramas. Before, interactive film-games required live-action shooting. Because of branching paths, they're more complex than short dramas — you could count the projects on one hand, with extremely high barriers and costs.
Now with AI generation, costs can be reduced by a hundredfold.
👦🏻 Koji
Have any decent small hits emerged in interactive film-games yet?
👩🏻 Xiaoning
Not yet. I spoke with LinearGame before — they've produced several titles, but I haven't tried them. But beyond pure video, interactive video should also be a direction people explore.
👦🏻 Koji
Then back to AI NPCs — quite a few games have started incorporating AI NPCs now. Are there any you think are particularly well done? Especially lifelike, really injecting new soul into the game?
👩🏻 Xiaoning
I do think the clear category leader is still Stardew Valley.
👦🏻 Koji
OK. Why are they good? Is it just plugging in a language model with some character presets?
👩🏻 Xiaoning
I think for NPCs in traditional large games, connecting a model and giving it a prompt was enough — it provided a PLUS experience.
But Stardew Valley is doing this seriously, and it has the multiplayer scenarios we discussed earlier. That makes the technical difficulty and application dimensions completely different. It's not just pure roleplay — they're treating the character as an IP. I'm really looking forward to its official release.
👦🏻 Koji
When's the official release?
👩🏻 Xiaoning
No idea, you'd have to ask miHoYo. I hope they hurry up.
When Claude Code Became the Most Fun Game
If you ask game designers to name "the most fun game I've played recently," more and more people's answers have changed.
👦🏻 Koji
The games industry is actually full of incredibly talented people, and I think many of them entered this industry out of pure passion.
Say there's a young person today who loves games, just graduated from college and wants to make them. They probably can't avoid facing the AI question somehow. What advice would you give them?
👩🏻 Xiaoning
Current commercial mobile game pipelines grind that passion down. Gacha games with 42-day major versions, 21-day minor versions, even weekly updates for live operations — you're like a cog embedded in a massive machine, and your passion gets worn away.
AI coming along isn't necessarily bad. If AI lowers the barrier to making big games, attention gets fragmented, industrialization scales back a bit — that actually protects people's passion.
For fresh graduates job-hunting, honestly, whether AI exists or not doesn't change much. If you're a programmer, you might panic more, but that's not games-industry-specific. If you're an artist, aesthetic judgment remains irreplaceable — actually amplified in the AI era. If you're a designer, I don't think AI brings revolutionary impact to core questions of what's fun and how to design interaction.
On my podcast I often ask people what games they'd recommend lately. After this past Spring Festival, more and more people's answer was: the most fun thing they've played recently is Claude Code.
If you asked me, I'd give the same answer. This raises a question: Is our relationship with AI really just tool-based? The relationship between creation and consumption, business models — they're shifting. If "creation is consumption," then the creative process itself can become more fun, even become a standalone experience.
👦🏻 Koji
Creation has always been inherently fun for people. I know selling courses, selling photography courses is extremely profitable now — especially phone photography courses for moms. Learning while shooting while editing — that process genuinely brings happiness.
👩🏻 Xiaoning
Right.
👦🏻 Koji
And I've always had this interesting observation: people love photographing their food. I think it's because three meals a day are, for most people on an ordinary day, the closest thing to a work, even an artwork, that they encounter.
👩🏻 Xiaoning
That's a fascinating angle.
👦🏻 Koji
Because with a dish, you're not just consuming its usefulness — you're also consuming its color, composition, plating. So in a given day, we don't really have other opportunities to see something resembling a work like this.
👩🏻 Xiaoning
This is a very interesting angle. When technology and products are still relatively early and non-consensus, a lot of things actually shouldn't be analyzed rationally.
If everyone feels it's fun — this sensory thing, your emotions genuinely getting satisfied from it — that matters. That really matters.
I was listening to your podcast with Yusen, and you mentioned that what first made you think Xing was really impressive was that his products fascinated you, made you want to use them. That gut-feeling, intuitive signal matters more than anything in the early days. No need to over-explain it — it's just instinct.
👦🏻 Koji
It's like when my kids first went to The Walt Disney Company. They saw Mickey and Minnie for the first time — even though those characters weren't in the cartoons they'd watched.
But the first time they saw, life-sized, someone dressed as Mickey and Minnie — they were three years old — they literally froze for three seconds, then let out a scream. A physical reaction: how could something this cute exist in the world? They were trembling and screaming.
I actually thought they were really, really beautiful too, but the kids' reaction was just so much more intense.
👩🏻 Xiaoning
Right, I think that kind of more intuitive experience is actually what people in interactive entertainment are trying to capture.
👦🏻 Koji
Thanks Xiaoning for coming to Crossing today. Looking forward to sitting down again in a year for another crossover episode — and hopefully by then, we'll have more AI interactive content that we can both say we've played, that we're genuinely excited about.
👩🏻 Xiaoning
I'm really looking forward to that day too.
👦🏻 Koji
Thanks, bye.
👩🏻 Xiaoning
Bye.
Crossing is looking for independent writers to cover AI product and model reviews.
If you've written articles like: Hands-on: PixVerse C1, Hands-on: LibTV, please contact zeo0811@gmail.com. Your email should include: ① a brief bio, ② AI review articles you've written.
We offer competitive rates. Looking forward to observing and documenting the AI era together 🎪