Zhang Yueguang added a chatbot to an otome game

"Pure Comedy"

After fully experiencing the first beta of Zhang Yueguang's STARFIELD, I kept wondering what angle to write from, so I decided to cover it alongside Dokie — an AI PowerPoint tool that only started development recently, like, what year is this?

First, a disclaimer: I'm male, and this review comes from a pure gamer's perspective.

My first impression after playing: STARFIELD has nothing to do with AI. At its core, it's a low-quality otome game that one-to-one copies the narrative and home-building systems from established titles in the genre, with the only difference being a freely conversable chatbot added on top.

No idea how they managed to hustle Baidu, but Robin's money sure is easy to earn 👋😭👋

STARFIELD's premise: you play as a shop manager running a dream hotel aboard a train alongside an android named "STARFIELD," where guests check in to dream and resolve their emotional knots. Sounds like it's targeting victims of East Asian family dynamics.

The narrative has zero connection to AI — you just tap through a bunch of Live2D images.

The first three chapters are standalone, episodic. Chapter one features a young man burdened with guilt over his mother's death due to a past accident. Yep, definitely East Asian family-oriented storytelling.

Chapter two takes a more unique angle, from a cat's perspective, hoping its owner can find happiness.

Chapter three returns to the dysfunctional family well: a young female artist suppressed by her family.

As a player, my primary goal entering the game is to experience the story. New chapters unlock at certain levels, and there are two ways to increase intimacy: buying furniture and developing new dishes in the home system, or conversing with STARFIELD.

Because I discovered that talking with the male lead raises intimacy extremely slowly — only asking about STARFIELD's height, weight, and such would increase it — I mainly grinded intimacy by constantly buying furniture and developing dishes. So my total conversations with STARFIELD numbered less than 20 rounds.

But hilariously, after clearing a certain stage and reaching level 10-something, STARFIELD suddenly confessed to me, saying, "This sounds like the scene I've been longing for deep in my programming... let's stop leaving our relationship ambiguous."

I'm genuinely done. We're not that close 😭

And what year is it that people still love using parentheses so much, I'm literally ( ) ( ) ( ). Pure STARFIELD leftovers, retro Orange Light game vibes.

After the confession, STARFIELD becomes willing to answer some NSFW plotlines. Pure manual backend configuration.

And there's only one character. The moment you speak, it knows you want to advance the romantic relationship — the mask comes off immediately. Real yumejoshi are better off dreaming on STARFIELD, though STARFIELD isn't much better either.

Zhang Yueguang claims STARFIELD isn't building an AI chat tool but an IP, with all assets manually created. (But it really has no distinctive features, and why do all AI otome male leads look identical now?)

Then why didn't Love and Deepspace, which actually built an IP, adopt AI? The reason: if the AI suddenly spits out something cringe, the idol illusion shatters instantly and it's over.

Love and Deepspace lets players cosplay; STARFIELD weakens the narrative to let the AI character directly enter users' real lives. This means constantly initiating conversations with it, but with zero pre-set triggers — are users supposed to rack their brains coming up with questions?

Even galge doesn't do this 😭

I suggest everyone making AI otome games download a real otome game and play it. If that's too hard, search for top-tier "Qin Che" on Xiaohongshu. See how far proper otome games have evolved — anyone watching these little videos would find it hard not to cringe 🥵

If we're stretching it, Love and Deepspace builds major idols; STARFIELD at best can produce a crooked-mouth dragon king short-drama pretty boy.

This wave of AI otome games also loves setting characters as androids — first hinting they're androids, hoping players will be more forgiving.

But STARFIELD constantly reinforces this, endlessly reminding you "I'm AI, this is all within my calculations."

Feels like the guy is betting everything on an AI boyfriend IP, trying to give STARFIELD that lively artificial intelligence feel.

Zhang Yueguang said the company was making otome games from day one, so STARFIELD is Zhang Yueguang's own white moonlight?

My assessment: two years of development for this, might as well keep making camera apps.

I actually think the camera app they made shows more promise — definitely more usable than their otome game, ChatPods, and AI PowerPoint combined. How about making another Miaoya? Moonlight, take the fam for another ride 😭

No idea why an overseas product is cosplaying as an indie developer on Xiaohongshu, and why it stopped updating — they're not shutting down, right? 😭

At Baidu's last conference, Zhang Yueguang said all AI companion products combined barely cracked 10 million DAU, with difficulty breaking into mainstream audiences. His solution: gamification — you can't simply add AI features to otome games.

Therefore, he wants to shift characters from pure UGC to PGC, expand interaction from pure text to multimodal, and introduce game monetization models — selling costumes and items instead of tokens.

Zhang Yueguang said this was their "relatively mature understanding formed in Q3 last year" 😇

Here's a quote from @DeepSheep's assessment: "Teacher Moonlight takes common knowledge from anyone who's played games, distills it through very strategic methodology, and presents it comprehensively, systematically, persuasively, and professionally. Truly impressive 🙏 This is also the core capability scarce in today's AI VC circle — continuous learning!"

So this is the core competency of star entrepreneurs. I see the light 😭

To summarize: compared to ordinary otome games, both the game design sophistication and art quality are relatively low-quality copies. For example, wanting to attract furries by adding cat ears but not a tail. Didn't copy the full set.

STARFIELD's only advantage: the narrative has minimal male-gaze energy, clearly from a female creator. Turns out STARFIELD's female players easily accept its premise, and they're skilled at community operations with fandom-style thinking — probably can make a small-but-beautiful, one-hit-wonder game.

But I see zero ceiling for STARFIELD. Are they really planning to compete head-to-head with game companies at this art and story level? That's a bit too disrespectful to the gaming industry.

Finally, some irresponsible thoughts on the AI companion赛道.

I think the real opportunity now is 3D ChatGPT (in China, probably 3D Doubao). Most otome games are pursuing what looks more achievable — 2D and Live2D — but that's likely a detour.

Chatting with ChatGPT about work and life is a definite use case; chatting with a 3D character with a story and background is definitely stickier than a robot, but ultimately you need lag-free video calls with precise intent understanding.

For women, the character should be a feminist man who understands feminism. For men, whatever, just pretty and thirst-trappy.

I used to wonder if AI companions must all be romantic partners. Currently it seems romantic relationships monetize better, and only romantic relationships make visuals matter. Wonder if any AI companion product will try bestie/friend relationship models.

Still hoping this赛道 produces a real contender. Don't let AI companion hype for two and a half years end with everyone still unhappy and back on Soul.

(Article images generated by ChatGPT, purely human-written, some perspectives inspired by Ginger & 🥚)