
On November 18, ahead of G-STAR 2026, Unity held the Unity AI Game Jam. Hosted for the first time in two years since its 2024 edition to coincide with Korea's largest game show, G-STAR 2026, the 24-hour overnight game jam featured OpenAI as a sponsor and mentor. OpenAI supported aspiring and active game developers in efficiently implementing AI to complete their games within a short timeframe.
Having made leap-and-bound advances over the past few years, AI has now become indispensable. How will it transform the game development process, and how far into game creation can AI reach in the future? Following the AI Game Jam, we spoke with Beom-joo Kim, APAC Advocacy Lead at Unity, and Tae-young Kim, Account Director at OpenAI.

What was the main reason for hosting an 'AI Game Jam' at G-STAR instead of a typical tech seminar?
Beom-joo Kim, APAC Advocacy Lead = When it comes to using AI for content creation—especially game development—there is a mixture of endless experimentation and conflicting opinions. Our research shows that over 90% of game studios are already using AI, yet surveys like GDC's show that more than half hold negative views about AI. Media reports also reflect this divide, showcasing games that achieved huge success with AI alongside titles criticized for using it. Truly diverse viewpoints coexist.
In this environment, we pondered how best to communicate with game creators. We concluded that it was essential to observe and engage directly on-site through a practical event like a game jam, where creators apply AI to actual game production. We wanted to see firsthand how they combine the Unity engine with AI, what new processes or ideas they generate using new technologies, and what support we need to provide them. Since data alone leaves many things unknown, a game jam felt like the most suitable format, so we pitched the idea internally. Unity HQ responded very enthusiastically, calling it a fantastic idea and providing extensive support. When we searched for a partner, OpenAI readily agreed to join, making this event possible.
Tae-young Kim, Account Director = From OpenAI's perspective, our headquarters considers the game industry, gaming clients, and the gaming ecosystem to be extremely important. Unity graciously invited us to join this great opportunity. We saw it as a chance to experience firsthand how AI is applied in real game development and to get a tangible feel for the current state of the gaming scene through the process of building games directly. We felt that having on-site mentors solve problems together with participants, as requested, carried immense meaning for us. We were thrilled to participate because we believed it would be a great opportunity to encourage more people to try game development and broaden the foundation of the indie and solo developer ecosystem.
It has been a while since Unity held a full-fledged game jam at G-STAR. How were the participants structured, and what was the atmosphere like on-site?
Beom-joo Kim, APAC Advocacy Lead = People from very diverse backgrounds participated. About 500 people applied, and we selected 54 participants to run the event. About 44% were students, while others included active company CEOs, aspiring founders, and job seekers.
In terms of enthusiasm, game jams naturally attract passionate people, but this time the atmosphere was slightly different because participants developed in earnest using AI agents and the Unity CLI. Traditional game jams primarily involve independent work followed by integration, where people code, draw art, and compose music separately before coming together to say, "I finished my part; let's combine them and see if it works." In contrast, this time programmers assisted with art, and artists created and experimented with process automation themselves. Alongside the familiar passion, the drive to attempt new approaches really stood out.

You mentioned that role boundaries blurred during the game jam. What specific differences emerged in actual development methods and final projects?
Beom-joo Kim, APAC Advocacy Lead = In the past, even when running a game jam over 48 hours or three days, completing a game was honestly not easy. Even if 10 teams started, many failed to finish, and because everyone was focused solely on finishing, there was little time to verify how the game would look to users, how it would play, or whether the tutorial was clear.
This time, however, tasks weren't divided into "this is art work, this is programmer work." Instead, participants actively crossed boundaries to collaborate, which allowed implementation itself to wrap up quickly. Once implementation was done, they moved on to post-completion polishing—adjusting balance, figuring out how to add more fun elements, and refining the UI for better convenience.
Looking at the games completed yesterday, I didn't feel like "these are novel even if polish is lacking because it's a game jam." Instead, I felt that many games were both novel and remarkably polished. In typical game jams, cinematic and presentation elements tend to be weak because teams are busy implementing core mechanics. This time, more than half of the games included intro cutscenes, staging, or story explanation videos, allowing first-time players to quickly understand the setup and world, and become immersed—a special aspect that set this apart from previous game jams.
From OpenAI's perspective, what were your impressions observing and participating in this overnight AI Game Jam?
Tae-young Kim, Account Director = Since it hasn't been long since we established our Korean office, this was our first time supporting an overnight game jam event like this. While we have occasionally run half-day hackathons for companies or universities, we didn't have much experience with events where participants had to complete projects under a tight overnight deadline.
However, the level of AI utilization among participants on-site was exceptionally high. Evaluating the finished results combined with the Unity game engine, I was deeply surprised and impressed by the outstanding quality. I was profoundly struck by how the entry barrier for the gaming ecosystem and solo developers has significantly dropped, enabling people to accomplish much more with fewer resources, and I felt strongly that we should continue planning events like this in the future.

Korean developers are adopting AI at a very rapid pace, as shown by Unity CLI usage ranking 3rd globally. What do you think is the reason for this? Did this also influence hosting this game jam?
Beom-joo Kim, APAC Advocacy Lead = Unity HQ is well aware that Korean game creators take on many new challenges. Just like pioneering graphic MUD games in the past with titles like 'The Kingdom of the Winds', whenever a new platform or technology emerges, Korean creators deeply contemplate how to use it to expand game quality and worldbuilding. In my role at Unity, I meet many creators in Korea and abroad, and I clearly sense that Korean developers don't reject new situations or tech; instead, they actively embrace them. So there was genuine anticipation to see what kind of results an AI Game Jam in Korea would produce.
As mentioned earlier, the primary reason for hosting the AI Game Jam was to see how AI is used on-site and shape our support plans. The second reason was that we truly held high expectations because Korean game developers acquire and apply new technologies faster than anyone else to make better games.
From OpenAI's perspective, based on this game jam and regular use cases, do Korean developers have unique strengths in how they utilize AI compared to other countries?
Tae-young Kim, Account Director = The biggest strength is that Korean game developers are 'early adopters.' That aspect feels overwhelming to us. I've heard that developers in some other countries can be very conservative or even hostile toward AI. In contrast, the Korean developers we observed on-site acquire technology and advance their skills at an exceptionally rapid pace. While quickly testing new technology, they place great importance on how much it actually benefits their projects.
The models we provide aren't just a single frontier model, but are divided into multiple product suites. Beyond Codex, which was used heavily during this coding jam, I was deeply impressed to see them actively leverage the Responses API—which handles reasoning and responses—and the Decision API, which enables fast in-game decision-making and action. That shows Korean developers' AI familiarity and adoption speed are truly world-class.
Beom-joo Kim, APAC Advocacy Lead = I also found their active use of tools like the Decision API very encouraging. Until now, the trend was to use AI to make existing game formats faster and more economically. In this game jam, however, it was impressive that many teams pitched ideas aimed at using diverse AI capabilities to discover novel fun factors that we hadn't experienced before.
What was your primary focus while providing on-site mentoring regarding AI utilization?
Tae-young Kim, Account Director = Four technical engineers from our headquarters came in person to conduct mentoring. The main reason we supported the event this way is that while Korean developers are extremely comfortable learning and using AI tech, there are still areas where things don't go smoothly. In particular, once people get used to a method, there is a tendency to stick to familiar approaches. AI advances so fast that a new model came out just two days ago, and applied model products keep expanding, yet developers can end up sticking with what they know.
So we strongly wanted to provide hands-on help directly on-site. Our primary focus was helping participants learn practical use cases and situational best practices through mentoring, apply them in-game, and walk away with hands-on experience they can continue to leverage on their own after the event.

What specific areas did "sticking to familiar approaches" refer to, and how did you guide participants on-site regarding newly released models and features?
Tae-young Kim, Account Director = Because technology develops so rapidly, a gap can form between knowing a new feature exists and properly weaving it into actual development workflows. We strive to convey these updates regularly and quickly to our major enterprise clients and customers. It is also important to tailor AI usage to the developer's specific goals and work environment. When a new model like the 6.2 model comes out, new API products follow and Codex skills expand, so we believe it is vital to deliver these updates in a way that allows developers to adopt them easily and effectively.
What were the most impressive examples or features among the completed games from this game jam?
Beom-joo Kim, APAC Advocacy Lead = There were many impressive aspects. First, as the burden of implementation itself lifted overall, many entries elevated quality aspects like graphics, sound, and animation to a significantly higher level. Another sight never seen in previous game jams was that when participants came for mentoring after completing their games, instead of asking about functional or technical troubleshooting as they used to, they asked numerous questions about art, design, balance, and elements that directly impact user experience. We plan to display several games at our booth, and many were highly polished titles with core gaming fun intact—graphics quality, timing, feel, and balance—making it hard to believe they were built in such a short time.
Another striking point was that even though teams equally used AI and received assistance from AI for concept art or 3D modeling, the results never turned out lookalike. Depending on the prompts used and the genre targeted, styles varied from cartoon to realistic, and color palettes and motion nuances differed considerably. Students often ask me, "If we use AI, won't all games look the same? How can we differentiate?" Looking at the projects from this game jam, we confirmed that even when using the same engine, AI agents, and APIs, entirely different outcomes emerge based on each team's planning and approach.

While past game jams centered around technical and practical execution, developers are now shifting toward a 'director' perspective directing the overall game. Do you think this shift creates the difference in polish?
Beom-joo Kim, APAC Advocacy Lead = Yes. As the burden of implementation diminishes, creators contemplate more deeply what experience they are offering players—whether it's a high-difficulty game delivering a sense of accomplishment, or a cozy, relaxing title. While AI handles implementation, creators must continuously playtest to judge whether the balance is right and fits the intended genre. Now, even at game jams, the conversation isn't "Congratulations on implementing it," but rather "Now that it's implemented, let's balance it and sharpen the edge and UI readability." High-level decision-making takes center stage. We saw many creators engaging in higher-level thinking rather than having their workload reduced or simplified by AI.
Tae-young Kim, Account Director = To add to that, no matter how advanced AI becomes, building the game and determining what makes it fun remains the creator's role. That is why the developer's role will become even more crucial in the future. Designing the game and shaping its fun elements belong to the creator, while AI acts as a helper that rapidly turns those ideas into tangible results. When combined with developer production experience and domain knowledge, AI immensely helps validate ideas and deliver outputs quickly for concepts that were previously hard to attempt due to time and headcount constraints. To use an orchestra analogy, whereas developers previously did their best in their specialized domain to present just that part, they can now step up as conductors to produce broader and richer content.
When collaborating with game studios or clients on a regular basis, what is their biggest consideration regarding AI adoption?
Tae-young Kim, Account Director = On a broad level, there isn't a huge difference from solo developers. Because competition is fierce and new technologies keep emerging, major game developers also adopt technology extremely quickly. However, a gap in familiarity with newly released tech always exists. Therefore, we try to present many use cases showing how new features can assist across the gaming vertical, in-game mechanics, productivity, and AX (AI Transformation).
Regarding this, we assist across three main pillars. First is AX (AI Transformation), transforming the development environment itself using coding tools like Codex. Second is productivity. Many game studios and IPs maintain and run games with legacy codebases; we help them with code migration or refactoring, and support them in rapidly drafting art concepts and mockups using image models. Finally, the area discussed most actively is the in-game realm.
Since we have strengths in reasoning models and multimodal features (voice and vision), we are expanding in-game applications. Multiple game companies are exploring ways to use our Audio API for seamless NPC interaction or building voice-activated AI companions and assistants. We are also supporting efforts to enhance NPCs via the Decision API so they make rapid decisions and act, interacting with players like real human beings.

Beyond this game jam collaboration, what are your expectations for future partnerships or synergies between the Unity ecosystem and OpenAI models?
Beom-joo Kim, APAC Advocacy Lead = What both of our companies jointly pursue is finding ways to support the Korean game creator ecosystem so they can stay ahead, and our visions align on that front. However, concrete collaboration plans remain to be discussed, though we believe possibilities are open whenever opportunities arise.
Tae-young Kim, Account Director = While details remain for future discussion between both companies, this event ended with positive results, and OpenAI gained immense insights. The outcomes showed a dramatic difference compared to the output levels of various hackathons we've held before. We've conducted similar events where individuals built games using Codex, but partnering with a major game engine provider demonstrated incredible high-quality synergy. If opportunities to generate synergy arise in the future, I believe we can fully plan and collaborate.
As noted throughout, technology is progressing rapidly. From the perspectives of Unity preparing next-gen engines and OpenAI leading AI, in what direction do you expect game development technology to evolve?
Beom-joo Kim, APAC Advocacy Lead = An analogy I often use when a new medium arrives and shifts how we view existing ones is the shift in publishing from manuscript copying to the invention of the printing press. As printing costs dropped and more people published their own writing, the market expanded from limited genres like law codes or religious texts into concepts like poetry, sonnets, and periodicals. This enabled exchanges of diverse emotions and experiences, sparking explosive growth in publishing culture itself.
Similarly, games evolved from electromechanical pinball machines into arcade culture, and then combined with PC, mobile, and the internet to evolve in ways unimaginable in the past. Moving forward, as new technologies are integrated, there will be opportunities to develop ideas, genres, and interactions previously unthinkable. Ultimately, if people harness technology with creativity, this won't be a crisis, but an evolution toward building gaming experiences on an entirely new plane.
Tae-young Kim, Account Director = Focusing on games, the core capability will remain defining the desired experience specifically, delivering necessary context to the AI, and judging whether the result aligns with the intent. AI will act as an assistant, enhancing productivity and in-game experiences—through more natural speech, photorealistic image quality, and various other avenues.
Conversely, as AI improves, developer competencies in planning, judgment, and prompting—supplying the context needed to achieve desired outputs—will inevitably become relatively more vital. I believe developer capabilities in these areas may become even more important than the advancement of AI itself.
Beom-joo Kim, APAC Advocacy Lead = To add a concrete example, we seem to be entering an era where people who didn't previously make games are starting to build them. Around me, filmmakers and dancers are now able to make games. I expect cross-disciplinary cultural and genre fusion to break existing molds and drive fresh attempts that advance the gaming medium further.
Finally, do you have any words of advice for aspiring developers, students, and industry professionals looking to create games using OpenAI and Unity?
Beom-joo Kim, APAC Advocacy Lead = I'd like to share two main points. First, I hope people don't disregard traditional development methods and continue studying them. You need to understand basic C# syntax and concepts, color psychology and foundational terms in graphics, as well as animation concepts and technical steps.
On-site, I often hear people ask, "Others are making awesome games with a single AI click—why can't I?" Looking at their prompts, the information is far too sparse. If you simply say "Make a running game," even AI doesn't know if it's running on a road, whether it involves animals or humans, or if it's a speed race or a coin-collecting game.
The Unity engine already features a 3D render pipeline, particle system, and VFX system. If you know traditional development concepts, you can give far faster and more effective instructions like "Use the Unity VFX system to make a game with exploding fireworks." The superior quality achieved in this game jam was precisely because traditional game development skills were combined with cutting-edge generative AI technology.
Second, it's time to contemplate broader strategic goals as a creator. Many say "I want to make a puzzle game" or "I want to make a strategy game," but often haven't considered whether their end goal is running a game company, remaining an active solo developer, or expanding their own distinct world. As implementation becomes easier and faster, standing out among similar-looking AI games will require establishing your own genre conventions, art style, worldbuilding goals, and building a community.
Tae-young Kim, Account Director = I share a similar perspective in a broader context. Rather than feeling you must master everything from the start to build a massive project, I encourage taking a small idea you've wanted to make and attempting to build it together with AI.
For prospective developers, AI acts as a 'helper' lowering the barrier to entry for creating that daunting first output, so I encourage you to try implementing everything you imagine. For experienced developers, it has become a tool to unleash accumulated expertise and experience more broadly, expanding reach into previously untouched domains to deliver higher-quality outputs. Recent news has highlighted success stories of non-developers achieving hundreds of millions of won in revenue as solo developers through vibe coding alone; I hope you leverage our underlying products and technologies effectively to take on new challenges.
