The gist

A Chinese survival game (the local counterpart to Rust) is flooding servers with copies of one streamer, and the phenomenon reveals a structural shift in the AIGC era: KOLs are becoming public characters. A real person has 24 hours. Their digital double can appear simultaneously in thousands of player stories, stealing horses, raiding bases, getting revenge. This introduces three new concepts: concurrent personality presence (one identity living in countless narratives at once), content SDK (packaging a creator's personality as a callable creative interface), and digital personality licensing (likeness rights are no longer enough; what's needed is a dynamic permission protocol that looks like software permissions).

A game overrun by copies of one person

For context: "Uncontrolled Evolution" (失控进化) is essentially the Chinese counterpart to Rust, a survival game where players gather resources, build bases, form alliances, and raid each other in an emergent, player-driven world. Lan Zhanfei (蓝战非) is one of China's most popular gaming streamers, known for content built on real accidents, self-deprecating humor, and the kind of contrast that makes great clips. He's not a distant celebrity; his community already treats his experiences as raw material for memes and remixes.

This game recently went viral. Streams, short videos, even feeds that don't usually cover gaming kept surfacing clips featuring Lan Zhanfei.

One Lan Zhanfei in a game is an official KOL partnership, deliberately deployed. Ten Lan Zhanfeis chasing someone across a wasteland is where things get weird.

In TikTok comment sections, players say: Lan Zhanfei stole my horse. Lan Zhanfei killed me. I got surrounded by three Lan Zhanfeis. Lan Zhanfei raided my base. What makes this funny is that it's incredibly real, yet impossible to tell which one is the real Lan Zhanfei.

Concurrent personality presence

When a game company hires a KOL, they used to buy their time, traffic, and content output. Stream a few sessions, post some videos, attend an event. Even with deep integration, full 3D modeling, custom voice lines, the KOL remained something to be displayed. Players could own a celebrity skin, but rarely used that celebrity's identity to actually do things.

Lan Zhanfei entered the player behavior layer. He can steal horses, raid bases, deceive people, beg for mercy on his knees, or get hunted down after killing someone.

The real Lan Zhanfei might be traveling somewhere right now. His digital double is stealing horses and raiding bases across servers nationwide. A real person has 24 hours. A digital personality can appear in thousands of player stories at the same time. The same person is a tribe leader on one server, a victim on another, and on a third server just scammed someone and is being hunted down.

This is a capability that was rarely priced seriously before: concurrent personality presence.

Copying a face many times isn't enough to produce this effect. Many games have done higher-spec celebrity integrations that still ended at the skin level. They were missing a complete chain: the person must be playable (high image elasticity, can be hero or punching bag), the character must appear frequently (everywhere means worldbuilding), and the game itself must continuously generate events (the system provides rules, players create stories).

My horse getting stolen is just a normal experience. Lan Zhanfei stealing my horse is approaching complete content. Character, action, conflict, and absurdity, all compressed into one sentence.

Celebrity skins solve whether players can own him.
Content SDKs solve what players can do with him.

Content SDK: a creator's face as a callable interface

A software SDK packages capabilities so developers can call them across different scenarios. Applied to content, a KOL's face, voice, mannerisms, catchphrases, existing memes, and public recognition can be packaged as a creative interface.

A truly functional content SDK needs three things: personality assets (someone people recognize and want to use), a generation environment (each call to this personality encounters new events), and a permission protocol (defining how far players and AI can take this person).

AIGC here acts more like a content compiler. The game generates events first, players experience conflict, Lan Zhanfei's image provides unified character recognition for that event, and AI helps players complete editing, voiceover, image modification, continuation, and repackaging.

In the past, a player who experienced something hilarious might let that story die because they couldn't edit video or organize footage. That gap is closing fast. Every player can simultaneously be actor, screenwriter, footage collector, and editor.

Marketing is changing as a result. Marketing used to happen outside the product, drawing users in. Now marketing assets enter the gameplay directly. The process of playing the game simultaneously manufactures the next round of promotional material. When a player says "Lan Zhanfei stole my horse," they're describing their game experience and completing a natural brand impression at the same time.

Digital personality licensing: likeness rights aren't enough anymore

When a face can be called infinitely, who does it belong to?

Traditional contracts specify usage platforms, time periods, regions, and asset scope. With generative content, both parties can't predict in advance what players will actually create.

Future digital personality licensing will likely need to handle a dynamic permission set: can voice be cloned, can new dialogue be generated, can the character be parodied and rewritten, can players roleplay as them, can generated content be used commercially, can these assets train models, how to handle takedowns in disputes, and how to split derivative revenue.

What creators sell will gradually shift from a face and a performance to usage rights over part of a digital personality. This looks a lot like an API with permission management: who can call it, which capabilities, how many times, who owns the output, who's liable for risks.

There's a trickier problem. When thousands of players continuously use the same character, the community starts to reshape that character. The person provides the original face and personality, the game provides the behavioral environment, players add new experiences, platform algorithms filter which versions spread most easily, and AI learns from those popular versions to generate the next batch. The Lan Zhanfei the public knows best may already be substantially co-created by players and algorithms.

Legally, the personality and likeness still belong to the person. Culturally, the personality is being co-produced by the community. How much control a person has left over their public image will become a deeply difficult question.


Today, players are just wearing Lan Zhanfei's face to steal a horse. The problem still seems funny. Going forward, when digital characters have his voice, expression patterns, personality, memory, and autonomous conversation ability, one KOL can enter a hundred thousand players' independent storylines simultaneously.

When a person can be infinitely copied, called, and reinterpreted,
are we buying a face, or usage rights to a piece of personality?