Written by | Hao Xin

Edited by | Wu Xianzhi

The AI industry's so-called 'consensus' has emerged with remarkable speed.

The era of Vibe Coding is fading into history, giving way to diverse iterations of Vibe Work.

OpenAI has integrated Codex into ChatGPT, ByteDance's TRAE SOLO has evolved into TRAE WorkSOLO, and Moonshot AI recently introduced the Kimi Work feature.

The essence of the Work mode represents another alignment with the Coding direction. These moves collectively signal a shift: the value of AI-generated code is transitioning from the creative process itself to the tangible value of creative outputs in real-world applications. Workflow efficiency, productivity, and reusable outcomes are becoming the new focal points.

Amidst this trend, Ant Glimmer stands out as a unique, almost 'rebellious' presence in the current mainstream narrative.

Although Ant's leadership initially positioned the product as an efficiency-driven tool, the successive releases of 'Flash Apps,' 'Glimmer Circle,' and the 'Glimmer Flash App Creator Incentive Program' have progressively underscored Glimmer's consumer-facing (C-end) attributes.

Today, the most popular content within Glimmer Circle predominantly features light, interactive experiences with entertainment and social elements, rather than genuine efficiency tools embedded in professional workflows. Officially, it is described as a 'consumer-grade Coding Agent accessible to everyone.'

Further solidifying Glimmer's strategic shift from Coding to C-end is the launch of Qmuse on June 9. This new AI product by Ant focuses on web application creation and team collaboration, with strict limitations on commercial licensing. This suggests Qmuse aims to integrate Coding into workflows to serve business-end (B-end) enterprises.

In essence, Glimmer resembles an AIGC content consumption platform. Games, tools, web pages, and other user-generated creations can be viewed as content offerings. Within Glimmer Circle, other users can freely explore, like, and share these creations.

While other Coding platforms are dedicated to time-saving solutions, Glimmer is attempting to fill leisure time. Whether this approach succeeds will hinge on Glimmer's ability to strike the right balance between technology and content in the future.

'4399 + Orange Light Games'

Opening 'Glimmer Circle' evokes a sense of nostalgia, reminiscent of the early 2000s, with the familiar feel of 4399 web-based mini-games and the artistic style of Orange Light online novels.

Applications like Glimmer have indeed democratized creation. A literary game author who won a 10,000-yuan prize from Glimmer shared that crafting traditional literary games used to consume an entire afternoon for just one chapter and a few dialogues. Character illustrations and scene materials required custom commissions from artists, leading to high costs and lengthy development cycles. Now, a literary game application can be completed within an hour, and after a few more hours of refining the interface and narrative experience, it can be launched.

We observe that AI applications like Glimmer are reshaping how some users create content. Take the aforementioned literary game author as an example; she crafted the game text, dialogue settings, character roles, and level designs through conversations with AI. Moreover, subsequent game generation and iterations all occurred within a single chat interface.

However, as an AI content platform, Glimmer has yet to achieve a fully closed content loop.

The most popular content within Glimmer Circle remains 'one-time entertainment' such as games, quizzes, and interactive stories. Users may find them intriguing upon initial contact, but the homogenization is severe, with many 'click-to-score' and 'option-branch' applications offering similar gameplay.

After a few plays, users lose the motivation to revisit them. The lack of daily updates, progressive challenges, and other incentives for replayability results in a consumption pattern characterized by swiping, trying out, and leaving—a disposable experience. Therefore, it is challenging to sustain users' attention long-term, let alone trigger an 'addiction' mechanism.

If Douyin's underlying logic for retaining users is human-to-human interaction, Glimmer's current ecosystem remains stuck at human-to-application interaction. AI serves as the foundational technology supporting content consumption and interaction, not the endpoint.

It lacks the accumulation of social relationships, offering only basic liking and commenting features. The connections between users remain weak. Take the literary game author community as an example; they primarily use Glimmer as a creation tool and prefer to promote their completed works in fan groups, Xiaohongshu, and other platforms.

This makes it difficult for Glimmer to retain users. While Glimmer stimulates content production through a 100 million yuan 'Creator Incentive Program,' the sustainability of this subsidy-driven ecosystem remains a significant question mark.

Glimmer is not entirely inactive in closing the distribution loop. Users can indeed generate shareable links for Flash Apps with one click, allowing others to open and use them directly in an H5 interface. This at least achieves a 'plug-and-play' dissemination effect and avoids the conversion loss from social apps to app stores.

However, the closed loop created through such sharing remains only a 'half-closed loop.' The issue lies in the fact that the shared H5 version is essentially a lightweight demo with limited experience and functionality, significantly inferior to the full-featured version within the Glimmer App.

For example, Glimmer Circle supports secondary creation features, but the H5 interface typically only supports usage, not creation. If a user sees a fun Flash App in WeChat and wants to modify it or add a feature, they must switch back to the Glimmer App—and the switch itself constitutes an additional conversion funnel.

Glimmer Seeks Breakthroughs

At its first media briefing, Glimmer's relevant person in charge emphasized, 'Glimmer essentially solves efficiency-related problems. Our product proposition positions Glimmer's main axis on the efficiency side.'

However, creating applications within an AI application inevitably draws comparisons to another product—Macaron. Chen Yusen, the head of MuleRun, represents one perspective: 'Ant Glimmer can be understood as a high-end version of 'Macaron.' Essentially, they differ in their choice of production relations. Glimmer's logic is to create for itself and view it as part of a social network, while MuleRun operates on a marketplace logic of creating for others.'

First, we must answer a question: Is handcrafting applications within an AI application a pseudo-demand?

In reality, users do encounter scenarios where existing apps are unsuitable, yet they lack the coding skills to build their own. For example, creating a calculator for splitting bills during a group meal or a Pomodoro timer with countdown reminders. AI generation can fulfill these minor, non-standardized needs within 30 seconds, far faster than searching, downloading, and registering in an app store.

Many users 'craft' applications not for use but for fun. For instance, creating a 'meow when you click the kitten' mini-game and sharing it with friends. The joy of creation and social feedback are real—as evidenced by Glimmer's data showing users modifying their creations over a hundred times consecutively.

Glimmer currently attempts to position 'handcrafted applications' as the core of an independent platform, simultaneously pursuing both entertainment and efficiency goals. From a demand perspective, this is not a pseudo-demand: there is indeed a subset of users who enjoy 'crafting a bit.'

However, for now, Vibe Coding To C has inherently fragile aspects.

Users do not require temporary applications or novelty mini-games daily. The vast majority of needs are one-time: generate, use once, or play for a few minutes and never reopen. Consequently, the platform faces extremely low retention rates and user lifetime value.

Many handcrafted creations either come pre-installed in systems or have more professional free apps available. Users opt for AI generation merely out of laziness to download. However, if the generation quality is unstable or requires waiting, this convenience is negated. Meanwhile, user-generated Flash Apps rarely become their 'personal toolboxes.' Unlike notes, documents, or photos, they lack long-term value. Without asset accumulation, users face no migration costs, and the platform lacks stickiness.

Thus, Glimmer's current predicament lies not in whether demand exists but in its attempt to satisfy two weak demands simultaneously with a single product without providing complete distribution, retention, and monetization systems for either.

This approach essentially takes a feature that should belong within a larger platform or a gameplay mechanic within an entertainment community and turns it into a standalone app. A general store selling only 'temporary screwdrivers' cannot sustain a true chain brand.

An Open-Ended Future

The commercialization of AI Coding remains largely exploratory, with many uncertainties making it difficult to reach definitive conclusions.

An AI entrepreneur told us that they currently see no profitable space for C-end AI applications, arguing that many Coding products lack a compelling reason to exist for users. If an application is highly practical and widely used, considering profitability is fine. However, users currently treat AI as dispensable, with even ChatGPT lacking strong necessity.

Another product leader noted that some products' strategy is to convince users that they can help them make money, but the initial investment simply burns cash. 'When you offer services for free, hundreds of thousands or even millions of users may flood in during peak periods. However, once you start charging, you might not see ten users in a day.'

These individuals believe that the commercialization dilemmas of both C-end AI assistants and Kill Time-style Coding products can be summed up in one sentence: the cost structure resides on the production side, while revenue models can only rely on the consumption side.

This means every generation consumes computing power and tokens, with costs rising linearly with user activity. Meanwhile, revenue depends on users' limited attention spans, low ad click-through rates, or payments from a tiny fraction of heavy creators.

When users treat AI as an occasional toy, the monetization ceiling on the consumption side falls far short of the cash-burning speed on the production side. Consequently, many players have shifted their focus to Save Time productivity tools, packaging AI capabilities into narratives about 'helping users make money.' High-value demands generate high revenue, covering costs simultaneously.

The shutdown of Sora App serves as a cautionary tale. It possessed top-tier video generation capabilities yet attempted to enter the C-end market as an 'AI version of Douyin.' Users flocked to it for novelty generation, not to consume AI videos produced by others. When the novelty of free usage faded, the platform could neither cover its exorbitant computing costs through advertising nor charge users for video browsing. According to estimates, Sora's daily operational costs reached tens of millions of dollars, while user lifetime value was negligible.

OpenAI ultimately chose to shut it down not because the technology failed but because greater success led to greater losses. The platform economy model clearly malfunctioned.

Returning to Glimmer, its most dangerous predicament is not being defeated by a single competitor but being encircled by the market from both directions simultaneously.

Entertainment-side users seek immersive, addictive experiences and will flow toward more professional AI interactive storytelling platforms. Efficiency-side users require stable, reliable productivity tools embeddable in workflows and will choose more specialized Coding products.

A product caught in the middle, trying to satisfy both, may become a disposable toy—users try it once, play a bit, and then forget it.

However, Glimmer's advantages are also evident. Backed by Ant Group, it has sufficient resources to invest and support. Simultaneously, it possesses Alipay, a rare domestic transaction ecosystem with a billion-user scale. This means Glimmer does not need to build payment, credit, merchant services, and other infrastructure from scratch like other independent AI products. Once Flash Apps integrate with Alipay's payment, Zhima Credit, Mini Programs, and Life Account capabilities, they can evolve into tools and share customers.

Users can generate an event registration page with Glimmer and complete payments directly, create a rental agreement that automatically invokes credit-based deposit waivers, or generate a store coupon that seamlessly connects to the merchant's redemption system. This 'generation-to-transaction' closed loop is difficult for other similar products to replicate in the short term.

For Glimmer, this represents an open-ended future and offers a potential evolutionary direction for domestic Vibe Coding.