When Assets Become Free

The Trap in the Phrase 'AI Makes Everything'

The phrase heard most often in the game industry these days is 'AI makes all the games.' Unity's AI tools, Microsoft's game generation model Muse, demos that output playable screens from a single line of text. Seeing these, people conclude that production is being automated. But this statement is flawed from the start in that it lumps together the act of making a game into a single task.

Game development is not a monolithic block. You write code, draw graphics, add sound, arrange levels, decide what to include and exclude, and choose what to show to whom and in what order. What AI has gotten good at is the front part of this—the part corresponding to production. Code and assets are rapidly approaching near-zero marginal cost. The problem is that resources whose cost converges to zero are no longer the scarcity of that industry. Scarcity doesn't disappear. It just relocates. This time, it's moving toward direction and curation.

Every Time Something Became Free, the Bottleneck Moved

Looking at history, this kind of migration isn't the first. When desktop publishing emerged, the costs of letterpress typesetting and printing plate production collapsed. Then anyone could create documents that looked like books, and scarcity shifted from typesetting skill to editing and design discernment. What became rare wasn't people who could handle fonts, but people who knew which fonts to choose.

Streaming followed the same curve. When music distribution costs became effectively zero, the act of releasing a song into the world itself became worthless. Value went to playlists and curation. The real power on Spotify wasn't the ability to create songs, but the programming that determined what to play from among tens of millions of tracks and in what context. Game engines were the same. As Unity and Unreal were released for free, the ability to code engines directly lost its scarcity, and what mattered became who would use those engines to make what.

The pattern is clear. When production means become common, what becomes scarce isn't the skill to handle those means, but the judgment to select and set direction among the overflowing output. Generative AI has simply brought this curve to games. It's not making everything, but making the production part common and pushing the bottleneck backward.

What Technologies Does This Combine With?

Here we raise Nova's question. When generative AI game tools meet what other technologies do they create new industry structures?

One is combination with data and analytics. When content production accelerates, a single studio can churn out dozens of variations per day. Then the ability to measure which variations survive becomes key. When generation tools connect with play data pipelines, games become not works completed once, but systems that continuously throw out variations and learn responses. Next is combination with distribution and recommendation algorithms. In a world where content pours out infinitely, real power accumulates not on the creation side but in the discovery layer that determines what to show to whom. Last is combination with live operations and community. If AI generates assets on the fly, the operating costs of live-service games that swap content weekly in response to user reactions plummet.

At the intersection of these three, the definition of games changes. Games shift from products that launch to flows that continuously branch and learn following a director's taste. Code and graphics are merely water flowing through that stream, and a separate hand emerges to dig the channel.

Newly Emerging Economic Actors

When connected this way, what new economic actors emerge? The first to appear is the curator-type director. Someone who doesn't code a single line directly but picks out what's fun from among hundreds of candidates AI spits out and unifies the tone. In game design, the most expensive resource is already not hands but discernment, and that gap widens further.

The second actor is operators who control the discovery layer. In the era of infinite content, the bottleneck is not production but attention. The position that determines which games get exposure is currently held by platform stores, but as AI-generated content explodes, a more sophisticated curation layer will wedge in between. Just as playlists did in music, new intermediaries with programming rights will emerge in games. Third are one-person studios that convert taste itself into assets.

Thinking of a one-person studio making cozy utility games in Busan makes this structure clear. In the past, you had to outsource or learn art, code, and sound yourself, and that cost was the barrier to entry. Now that production cost is collapsing. The remaining competitive advantage is just one thing: this studio's unique tone and worldview, and the restraint of knowing what not to include. In a world where assets are free, a one-person studio's asset becomes not the speed of creation but the standards of rejection.

Korea's Position to Capture

Korea is in a peculiar position. It already has large-scale content production capacity, but it's difficult to beat the U.S. and China head-on in generative AI foundation model competition. Then the deciding factor is not models but the direction and curation layer placed on top of them. What K-content proved to the world wasn't technology but the ability to program taste. The sense of knowing what to show in what tone and in what order—that precisely overlaps with the resource becoming most expensive in games right now.

There's a counterargument. If AI eventually learns even curation and direction, even taste will be automated and this bottleneck will also disappear. There's merit to this point. However, curation is not about getting the right answer but about choosing with responsibility. The entity that decides what to exclude and bears the consequences of that remains human until the end. Models only expand options; someone must want what world they desire for it to begin. What's automated are choices, not desire.

The future doesn't come from a single technology. It's not that generative AI makes all games, but at the intersection where generative AI, data, and distribution meet, the definition of games is redrawn. When code and graphics become free, what becomes expensive is not the hands that create but the eyes that choose. The bottleneck hasn't disappeared. It's just moved to a more difficult position.