Lingguang Flash Apps on X: Real Demos, Real Links, and What They Show

LingGuang (灵光) from Ant Group has been showing up all over X with demos that look a lot like “vibe coding”: you type a plain-English prompt, and an interactive mini-app (often called a “Flash App” or “flash program”) appears right inside the chat.

The interesting part isn’t “AI can generate UI.” It’s that the outputs are increasingly app-shaped: runnable flows, interactive widgets, and small tools you can share.

This page is a curated roundup of real posts, quoted as plain text (no embeds), styled like tweet cards, and linked to the original source.

Related reading:

• /blog/lingguang-vibe-coding-app-review

Real Flash App demos people are sharing on X

MieYuan

@lylailing0410

"Zero-code built an Animal Chess game. Started as a small experiment, ended up playing it for way too long. 零代码手搓了一个斗兽棋小游戏。 原本只是想试试规则，结果一局接一局，根本停不下来 #NoCode #GameDev #IndieDev #BuildInPublic #instantgame #CreativeCoding #aicoding #灵光"

Source on X

Ant Group

@AntGroup

"Since its launch on November 18, LingGuang users have created over 3.3 million flash programs by leveraging the AI assistant’s native coding capabilities. The flash program feature enables users to generate customized, interactive programs through straightforward natural language prompts. Watch the video showcasing a selection of flash programs developed by our users."

Source on X

m1a0/acc

@m1a0acc

"今天国内很多 AI 大 V 都在吹捧蚂蚁灵光这个 App，尤其其中一个闪应用的功能，也算有大厂正式下场搞 Personal App 了。 但实测下来创建一个 AI 原生 App 的完备度挺低的，基本相当于编程小白 vibe coding 的卡片式 demo 搭配一个自动部署服务。尝试几个实际场景小工具的 prompt 都不可用，只好要求一再简化，还是简陋的不行，生图都靠模拟😓，最令人费解的是，创建对话中上下文记忆理解非常弱鸡，追问答非所问看得一脸懵逼。排版很精美在有价值的输出面前不值一提。 相同甚至远复杂于此的 prompt 在@myshell_ai 的 text2app 里一遍就过，交互清晰，功能可用。"

Source on X

D-Coder

@Damn_coder

"China just made a huge statement in AI... LingGuang feels like China’s answer to the new “AI assistant era.” It chats, analyzes images and video in real time, builds interactive charts, runs code, generates 3D models, and even creates flash-style programs directly inside the chat. No plugins. No extra tools. Just one interface doing everything."

Source on X

idoku

@dokupeng

"昨天试了灵光,今天试了gemini 3,对比感觉灵光像个demo...不知道Google是不是也召集精英搞的封闭式开发,憋出大招."

Source on X

Kirill Volodin

@Kirill_Volodi1

"Ant Group's new AI, LingGuang, built a million mini-apps in just four days. The future is arriving faster than we can download it."

Source on X

What these demos suggest (beyond the hype)

The recurring theme across the best posts is that LingGuang is trying to make outputs feel less like “chat answers” and more like small, runnable software:

• The strongest demos show a complete loop (prompt → app → interact → share).
• The most impressive moments are when the app is usable without manual fixes.
• The most common failure mode is also clear: early “Flash App” outputs can look polished but still be fragile (missing edge cases, weak memory/context, or prompts that only work after repeated simplification).

If you’re evaluating it, the question to ask isn’t “did it generate something?” It’s:

• Does the result work on the first try?
• Does it keep working after 2–3 edits?
• Can a non-developer actually complete a real task with it?

How to try LingGuang safely (official sources)

If you want to test LingGuang yourself, start with official distribution points:

• Web app: https://www.lingguang.com/chat
• iOS (App Store CN): https://apps.apple.com/cn/app/%E7%81%B5%E5%85%89-%E5%85%A8%E6%A8%A1%E6%80%81ai%E5%8A%A9%E6%89%8B/id6751496092
• Android (Tencent MyApp): https://sj.qq.com/appdetail/com.antgroup.leopard.android
• Android (Xiaomi): https://app.mi.com/details?id=com.antgroup.leopard.android

If you want a deeper breakdown (availability, positioning, what it feels like to use), see:

• /blog/lingguang-vibe-coding-app-review