In the Age of AI, Who Is Making Games?

This Chinese New Year, K Yan didn’t visit relatives or friends.He signed up for an AI subscription and spent the entire holiday glued to his computer, creating a simple early-learning game for his daughter to help her recognize numbers. The moment the game ran successfully, he finally made up his mind to quit his job and devote himself entirely to indie game development. It had been exactly five years since he last left the gaming industry.

A few months later, during the Dragon Boat Festival, Wang Jiewei—in an unprecedented move—spent the entire three-day holiday fine-tuning the AI game he had created. It’s hard to imagine that just a month earlier, he had never played video games.

Muffin is also deeply passionate about using AI to develop games. Every night for the past few months, after finishing his demanding day’s work, he hasn’t been in a hurry to rest; he always makes sure to assign development tasks to Codex so the AI can continue working overtime before he can finally fall asleep with peace of mind.

By 2026, AI will have completely lowered the barriers to entry in game development, and more and more people are flocking to the field.

01

Push open that door

Someone who never plays games at all wants to make games?

When Wang Jiewei first brought up this idea, it came as no surprise that it took his friends by surprise.

From his undergraduate studies at Southwest University of Science and Technology to his graduate studies at Shanghai Jiao Tong University, he majored in transportation engineering. After graduation, he joined China Communications Construction Group, where he participated in major infrastructure projects such as the Harbin–Dalian High-Speed Railway; he later transitioned into the data industry, working on digital government and smart city projects for more than a decade;In 2023, he founded a data asset company, where he spends each day researching ways to help enterprises transform their data into assets that can be used as collateral for financing on financial statements.

But in his 39 years of life, he had never stayed up all night playing a game, and had hardly ever even voluntarily launched a commercial game. He knew nothing about game genres, didn’t understand gamer slang, and had no idea what the popular blockbusters young people were talking about were. Yet it was precisely this “layman,” who knew absolutely nothing about games, who decided to make his own game.

This idea didn’t come out of nowhere. Prior to this, Wang Jiewei had already used AI to develop a series of service systems for industry clients, including a data quality assessment tool and a data asset operations platform; the third-party valuation of one of these systems reached over 2.9 million. As he successfully implemented these projects, he continually expanded his understanding of AI.

One of the systems developed by Wang Jiewei

One weekend morning this past May, a thought suddenly popped into his head: Since AI can build systems, why not try creating a game for a wider audience?

Stories always begin with a moment of inspiration, and this was true for Muffin as well.

Muffin works as the head of marketing operations for a sub-brand at a financial company in Hong Kong. Her days are filled with meetings, live streams, and business negotiations, and she frequently flies to Singapore, the United States, and other destinations to organize client events. There is virtually no boundary between her work and personal life.

But his busy life did not dampen his dream of becoming a creator.Ever since he was a child, Muffin had wanted to tell the stories he imagined to others through his creative work. He had studied painting, but because his grades in academic subjects were strong, he ultimately did not pursue the path of taking art entrance exams; after putting away his paintbrushes, his creative aspirations were buried deep within his heart, where they remained dormant for more than a decade.

It wasn’t until the advent of AI that this idea was reignited.

With the explosive growth of AI capabilities, companies have begun to widely incorporate AI into their workflows. After getting started with it, Muffin truly experienced the power of AI: “In the past, creating an ad landing page required going through the entire process—from design and development to front-end and back-end work and testing—before seeing the final product; but now, using Codex and Claude on my own, I can put a page together in just a few minutes.”

Just as he was actively exploring more AI skills, he came across examples of others using MCP to develop games. At that moment, an idea that had been dormant for more than a decade resurfaced: Why don’t I make a game myself?

For Wang Jiewei and Mai Fen, using AI to create games was a sudden idea, but for K Yan, it was a decision that had been in the works for a long time.

K Yan was certain from a young age that he would one day work in the video game industry.He started playing games in second grade and still vividly remembers the immense impact he felt when he first played RPGs like *Xianjian* and *Xuanyuan Sword*: “I never knew games could be this amazing”—their storytelling and immersive experience were almost on par with movies and TV shows. Deep down, a seed was quietly planted in the young boy’s heart.

Driven by this passion, he resigned from his management position at a foreign company in 2013 and returned to Chengdu, where he worked as a game designer at two medium-sized game companies before joining Longyuan as lead game designer. He also participated in the early development of the company’s flagship title, *Wuhua Mixin*.After eight years of hard work in the gaming industry, K Yan had achieved quite a few impressive results, but behind the glamour, a sense of emptiness had quietly taken root; he felt that the games he developed always “lacked a grounding in reality.”

After his daughter was born, he left the gaming industry to spend more time with his family and joined an educational software company, where he stayed for five years, working his way up from an entry-level employee to vice president. But his desire to make games never faded—if anything, it grew stronger: “Perhaps it’s only by seeing the world outside that you can truly understand what you really want.”

The opportunity he had been waiting for presented itself at the company’s annual meeting. During the meeting, the boss announced that he wanted all of the company’s business operations to embrace AI as soon as possible, and a live demonstration of AI-generated results was even given. It was then that K realized just how dramatically AI’s capabilities had changed.

After the annual meeting, he set to work using AI to create a little game for his daughter to help her learn numbers and practice addition and subtraction. Once the game was finished, he was certain: working full-time on games using AI was a viable option. So he decisively handed in his resignation: “I’m 38 now—if I don’t take a chance like this, I might never get another opportunity.”

02

The “Headaches” of Middle Age

Even though they all use AI to create games, people with different backgrounds always end up creating something unique.

Wang Jiewei knew he wasn’t good at visual design and didn’t know how to code, so he modeled his game after the “Jianghu” chat rooms that were all the rage during his middle school years. Using VS Code and the DeepSeek API, he issued commands to the AI in natural language to create a text-based H5 game set in a martial arts world.

When building this world, he didn’t predefine too many specific stories or details; instead, he incorporated his own understanding of “life values” into it: the world centers on character growth as its main storyline, and players gradually gain experience through combat, equipment crafting, and map exploration, continuously increasing their level and power.Interestingly, at levels 20 and 40, characters face a choice between two distinct paths; choosing one means permanently missing out on the other—because Wang Jiewei believes that life is just like that: growth is always accompanied by irreparable regrets.

He summarizes his development approach as “development based on feedback” and “incremental iteration,” constantly incorporating player feedback and carving out time on weekends, during lunch breaks, and in the evenings after work to make adjustments and optimizations.This requires investing a great deal of extra effort outside of his day job, but he relishes it: unlike his previous work on system-related projects, game development doesn’t require client research or multiple layers of review—he has the final say on all ideas, and he thoroughly enjoys this free creative process.

He even regarded this as a rare source of spiritual solace in his middle age—something he referred to in his local dialect as “jiao tou”: “At our stage in life—what we call ‘lao deng’—it’s truly rare to still have something you genuinely want to do.”

When the game was first released, he was too embarrassed to share it with others: “I felt ashamed—after all, I’m not a professional.” It wasn’t until the user base grew to 300 that his mindset gradually shifted, and he began to think, “It’s actually not bad.” He said that in the past, the government and enterprise systems he developed were very expensive, but their actual user base didn’t necessarily reach that level.

Exactly two months after the game launched, the number of players surpassed 1,000. At the time, he was on his way to work; he opened the data dashboard he had built on his phone, and the moment he saw the number, he immediately took a screenshot to capture that moment, which was especially precious to him. Wang Jiewei said that while this figure didn’t exactly constitute “success,” he was already very satisfied.

The Backend Data Situation at That Time

03

“Creating is a joy in and of itself.”

Muffin, who wanted to share her original stories, brought the novel she had written into the game and began developing *Animal Little Detective*.

Late last year, inspired by the children’s book *The Little Tigers’ Adventures*, he spent less than two months writing a detective novel: three teenagers embark on adventures together, encountering a series of mysteries that seem supernatural but actually have logical explanations; the five cases are interconnected, forming a complete full-length story.While writing the novel, he had already refined the logic behind the cases and the methods used to commit the crimes, and he drew numerous illustrations; the core framework and artistic style were essentially finalized at that time.

Early Concept Art

But novels and games are, after all, two different languages. When to provide clues, how to guide players’ thinking, and where to place feedback—every detail is worth repeated adjustment and refinement. He knows full well that “this is the game’s framework, and it determines whether or not it’s actually fun to play.”Only after these elements have been refined does the official development phase begin. Using AI to import scenes into the game engine, build maps, design animations, and debug the UI, they gradually transform the story on paper into a world players can step into.

Future Versions with Enhanced Graphics

Muffin has a very clear vision, but he never sets a schedule for himself—no milestones, no deadlines—he just keeps working on it every day. As long as the internet connection holds up, Codex and Godot run in the background on his computer year-round: he issues commands and assigns tasks to Codex, and Codex uses MCP or the command line to automatically call upon Godot to handle development issues.Before going to bed at night, he assigns the AI another sufficiently large task so it can continue working overtime, then reviews the results upon waking the next day—because Codex has a quota limit, “so he has to use up the quota before it resets.”

The project has been underway for less than a month, and he has already completed nearly half of the first of the five cases. He said, “If I can finish the entire game in a year, I’ll be very satisfied.”

Game Scenes

His passion for creation makes every moment of game development a joy for Muffin: setting up the first scene, designing the first UI interface, and solving map-related issues—each step fills him with a sense of accomplishment.Even his wife says he’s been playing video games less lately: “Creating is a form of expression—it’s more enjoyable than simply passively consuming what others give you.”

04

One Night, One Game

K Yan is more efficient; in the more than half a year since he quit his job, he has already released three games on TapTap.

K-Yan Social Media Accounts

He spent a full three months polishing the first version, but at the time, AI capabilities were far less stable than they are now. The bugs were so numerous they were a constant headache, and the art assets never quite looked the way he wanted—he had to redesign the character skins three times just to get the look right. On top of that, the gameplay itself was too hardcore, and the post-launch metrics were dismal. K even admitted that he felt it was a waste of the time invested in development.

Later, when GPT-5.6 was released, many of the technical challenges that had previously held him back were easily resolved, and he adjusted his development approach accordingly: instead of obsessing over perfecting every detail, he began using AI to quickly build a demo, which he would then release after a brief review and some minor tweaks—all to quickly validate whether the gameplay would resonate with an audience.

Following this line of thinking, he launched two new games in quick succession, one of which—titled *Military Strategist: The Han Dynasty’s Weiyang Palace* (hereinafter referred to as *Military Strategist*)—performed particularly well.

What’s more, it took just one night for this game to go from a concept to a playable demo. Even K Yan himself was deeply impressed: he had originally heard from netizens that a game could be made in a single night, but thought it was just “bragging.” He assigned the task to the AI on a whim, and sure enough, by the end of the night, it was actually done.He then spent another week refining the details and adjusting the game’s parameters before completing the final version of *Military Strategist* for release.

“Military Strategist” is a text-based choice strategy game inspired by a sudden flash of inspiration K Yan had one morning: Since his first game didn’t perform well, he decided to create a “Three Kingdoms” game—a theme everyone is familiar with.

In the game, players take on the role of a military strategist and must continually offer various suggestions to Liu Bei in order to secure resources and influence the course of the story. However, the Liu Bei in the game is not entirely rational; he places greater importance on morality, reputation, and the welfare of the people, among other factors, and may therefore reject the player’s carefully calculated “optimal strategy”;Players can win over colleagues to secure support or wait for the right moment to resubmit their proposals. As long as the right choices are made, there is a chance to rewrite the historical tragedy and achieve alternative endings, such as subduing Lü Bu or saving Guan Yu.

K chose Liu Bei as the central character because he has always felt that “Liu Bei isn’t someone who always makes the right choices.” He’s heard too many people speculate: If someone had been there to advise him back then, might history have taken a different course? So he turned that hypothetical scenario into a game.

Thanks to this ingenious idea, *Military Strategist* quickly climbed to No. 31 on TapTap’s New Releases chart after its launch, attracting more than 7,000 players to try it out, with many spending several hours playing.Some players have also told Kyan that they watch hundreds of rewarded ads in the game each day; this revenue has largely covered the cost of Kyan’s AI platform subscription.

K admitted that he was well aware that *Military Strategist* had a short development cycle and hadn’t been fully polished; there were still many areas in the game that left something to be desired, and the revenue it could generate was quite limited. However, he said, “Just seeing how much everyone loves this theme makes me very happy.”

Recently, K Yan released a sequel to the same series titled *The Imperial Uncle: Wei Wu’s Unfinished Legacy*.

05

“AI can never replace humans”

During the game development process, all three creators were deeply impressed by the capabilities of AI.

Wang Jiewei, who has no technical background, had never worked in operations before. With the help of AI, he successfully achieved a zero-downtime deployment, an accomplishment that once again expanded his understanding of the limits of AI’s capabilities.

Muffin was equally delighted when he first discovered that AI could directly control game engines like Godot. In the past, it took an entire team to make a game, but now he can bring his ideas to life all by himself. AI has truly lowered the barrier to entry for creativity by a significant margin.

K gave the AI’s communication, comprehension, and task-execution capabilities a near-perfect score: “You could say it’s even better than my former colleagues.” In the vast majority of cases, it can accurately carry out requests—something that is actually quite difficult to achieve in human-to-human collaboration, given that information loss is almost inevitable in team communication.

Of course, AI is not perfect; it still has many limitations.

Wang Jiewei’s biggest headache is the AI’s memory issues: “It’s like a child at home—you have to keep reminding it of important things over and over again.”Furthermore, the AI’s output is often unrealistic or prone to omissions. For example, during game development, it frequently happens that “the AI remembers weapons and combat techniques but forgets about internal energy or additional bonuses.”

Therefore, he believes that AI is currently still only suitable for handling specific tasks; under no circumstances should all work be entrusted to AI, and developers themselves must keep clear records so they can verify the situation at any time.

Muffin also noticed the AI’s memory limitations: “In long-term projects, it’s difficult for the AI to ensure consistency between different parts of the content. Even when the project context isn’t particularly extensive, it requires constant reiteration of rules and reference documents; over time, it tends to forget or even misinterpret the specifications.”

Next is the issue of cross-platform compatibility. At this stage, AI relies on MCP and command-line interfaces to integrate with game engines, which essentially involves cross-platform calls, making deep integration difficult to achieve. In his view, game engines with built-in agents may emerge in the future, allowing development to be carried out directly within the engine using natural language—a process that would be far more efficient than the current model.

Since existing agent applications have virtually no search capabilities and cannot quickly retrieve historical content, K Yan decided to use Feishu for project management alongside his AI tools: to submit requests, track bugs, and document progress—all to make it easier to search and verify information later. He even wrote a custom system specifically for this purpose that integrates with Feishu, allowing the AI to search for information within it on its own.

On top of that, what frustrates him the most is the output of art assets. Today’s AI tools not only struggle to produce satisfactory results, but they also lack consistency: “As you’re working, it starts to veer off course, and the entire style changes.” It takes a lot of effort to tweak the prompts, try different approaches repeatedly, and burn through a lot of credits just to get it back on track.

These technical shortcomings may gradually improve as AI evolves. However, the three creators agree that no matter how AI evolves, it cannot replace humans.

In Muffin’s view, judgment is the core capability that AI will never be able to replace in humans: “AI cannot determine whether a story is moving, nor can it tell if a game is fun—these are decisions that must be made by humans.”

K Yan went so far as to say outright that AI lacks a human touch. It can give you ideas, but its underlying logic is that of a Q&A tool, so what it produces will be nothing more than a textbook—not a real game. Therefore, you still have to take charge of the gameplay yourself.

K said, “AI will never replace or become a human. It lacks human senses; it has no hands, feet, eyes, or mouth. All its imitations of human emotions and behavior stem from programming, so it can only be an imitator and a performer.”

06

Floods and Freedom

For the gaming industry, the impact of AI is now inevitable and is already taking place.

K Yan observed that a large number of AI-generated games have already emerged on TapTap, noting that “at least 80% of the titles listed on the Creative Workshop are made by AI.”However, most of these games have a strong fan-made feel, and their details do not hold up to commercial scrutiny. While they often feature a novel and interesting angle that quickly grabs players’ attention, the actual production quality is not particularly refined.

Muffin also believes that the mini-game market will soon enter a phase of oversupply, with a flood of reskinned, imitative, and uneven-quality products flooding the market.Mini-games and browser games have simple technical structures; once an idea is conceived, an AI can generate a playable version in a matter of hours and a complete game in a few days. For instance, his browser game *Planet Exchange*, which he developed in three days, was played by more than 2,800 people within a week of its launch on TapTap, and its player base has since surpassed 5,000.

On the other hand, however, more people are gaining unprecedented creative and expressive freedom through the medium of games.

In the past, game development was monopolized by professional teams, and turning an idea into reality required going through project approval, development, and testing—a process long enough to wear down the initial enthusiasm.Today, even an ordinary person can turn the story in their mind into a playable game in their spare time. Those niche ideas, once overlooked by the mainstream market, finally have a chance to be seen.

As Muffin said, there will certainly be a wave of small but delightful games in the future; they may only account for a very small market share, but that will be enough for them to survive.

“I hope everyone can fully take advantage of the current AI boom to express whatever they want to express. A decade or two ago, you couldn’t have even imagined that, even with such a busy day job, you’d still have the chance to showcase the world you envision in your heart.”

K also said that, as someone who has been deeply influenced by video games since childhood, he particularly loves this current era where “everyone can express themselves, and every idea has a chance to find like-minded people.” “Even if you make a really bad game, as long as you put your heart into it, there will definitely be people who understand what you’re trying to convey, and there will definitely be people who resonate with you.”

In his view, making games is fundamentally about finding like-minded people; “Even if it’s not for the money, the act of creating and sharing should be a source of freedom and joy in and of itself.”