A Few Strange Things About AI Interactive Content

@Wu Ruirui

The Secret Morning Report

On a morning just before Loopit's new funding round closed, partners at numerous VC firms and major tech companies' strategic investment arms received a data report.

The report ran seven pages. Its author had analyzed Loopit's downloads, retention, session length, and other metrics one by one: some sourced from SensorTower, the world's most prominent mobile app intelligence platform, others from their own scraping estimates. Word has it the author had even vibe-coded a data scraping tool specifically for this report.

Despite bearing watermarks reading "For [Fund Name] Use Only," the report instantly spread like wildfire.

You can only say the AI funding battlefield has grown so intense that even the primary market now has its Muddy Waters short-seller stories. Who's playing Muddy Waters here? I know nothing—readers can draw their own conclusions.

Starting roughly around mid-last year, the interactive content赛道 suddenly surged. Loopit was one company to emerge in this space. Other players include: Rezona, Aippy (incubated by listed company BlueCity), Sekai, Loot, and Gizmo (a Silicon Valley company).

Similar launch timing, similar technology and model—all generate playable content from a single user prompt, then distribute through content feeds. The only slight variation is content category: some focus on absurdist meme mini-games, others start from lifestyle content and interactive art.

Perhaps precisely because differentiation is so limited, these companies quickly found themselves locked in fierce rivalry, especially Loopit and Rezona. They debuted at similar valuations, doubled them simultaneously, then followed the now-familiar startup rhythm—breaking past the $200 million valuation mark within just over four months.

The story these companies tell is also remarkably similar: "the next Douyin." The logic goes that Douyin achieved perfection in both short video (content) and vertical scrolling (interaction format), but vibe coding enables UGC content generation with user participation built in—creating an opportunity to disrupt Douyin.

Only later did this narrative draw mockery from some quarters, prompting certain companies to soften it to the more generic: "next-generation content platform."

Pixel-for-Pixel Douyin Copying

In many investors' telling, whatever the story, these companies are essentially copying Douyin pixel for pixel.

Founders and teams, for instance, are ideally Douyin alumni who learned the algorithm, user growth, and monetization trifecta there. Creator community operations: where Douyin recruited music and dance students, these companies seek out interaction design majors. Where Douyin once built around cameras and secondary creation, today it's vibe coding and remixing.

Most of these companies target overseas markets, betting that TikTok's operational and monetization efficiency lags far behind domestic Douyin—creating opportunity for startups. Gizmo, a Silicon Valley product built by a former Snapchat engineering team, was described by TechCrunch as an "interactive TikTok," reporting 600,000 total downloads in six months with growth rates hitting 312% over two months.

However, in early March this year, Gizmo's core team was acqui-hired by Meta.

Hard to blame anyone for getting excited—who could resist a赛道 this reminiscent of mobile internet's golden age? For over three years, AI startups have been simultaneously disrupting others and being disrupted themselves. Once the narrative returns to content platforms, all methodologies and metrics become crystal clear again.

Because these are content products, entrepreneurs and investors have refocused on downloads, DAU, retention, session length per user... thoroughly mobile-internet metrics. This explains the secret report's obsessive insistence on breaking down and verifying every data point.

Good-looking data was indeed a major reason many investors pulled the trigger.

On Google Play, Rezona, Loopit, and Aippy show 5M+, 1M+, and 1M+ downloads respectively. Compared to AI tools still struggling to build ARR, these companies have embraced the most straightforward business model: advertising, with good data translating into highly predictable revenue.

But is looking at data the right move at this stage?

DAU can be bought. Especially on Android devices in Southeast Asian markets, user acquisition costs run just $0.50 per install. North America is pricier at $3-5. To produce a million-download report card in short order, the floor is roughly $500,000.

Retention doesn't buy so easily. Two months ago, retention data charts for Aippy, Loopit, Gizmo, and Rezona circulated widely. Day-1 retention only barely cleared 40% for Aippy, with the lowest at just 25%. Day-7 retention hovered around 10%, with some below 5%.

How to contextualize these figures? Against big-tech standards from the mobile era: day-1, day-7, and day-30 retention should hit 40%, 20%, and 10% respectively to justify resource investment. Douyin itself had day-1 retention of 25-30% in its early A.me days.

Poor retention stems from insufficient addictiveness. I've used all these launched products, and they're genuinely fun at first. There's a poke-the-cat-belly-without-getting-bitten game I played immersively for over ten rounds. Rezona's one-sentence generator is remarkably approachable—even with zero inspiration, I made a Super Mario Milk Dragon version by selecting gameplay and meme templates.

But I didn't open a single product the next day. Some content feels fresh, but grows repetitive with crude visuals; some games require learning rules first, and I only figured out how to play by round two; and making my own content... isn't one workday exhausting enough?

I asked a founder who'd left the interactive content赛道 why this happens—was the model not good enough? He said the model's planning and coding capabilities are actually quite strong; the problem is games without pipeline support simply don't sustain engagement.

Of course, measuring retention against mini-games may be somewhat unfair—they've always been low-retention. But isn't that precisely why most mini-games remain parasitic within WeChat and Douyin mini-programs rather than running as standalone apps?

So some founders have sought paths beyond mini-games. Chen Weipeng said in a previous interview, "If AI-made game quality doesn't exceed what ordinary people can do, disappointing users too early will force the platform down a PGC path." LinearGame took a distinctive angle: interactive games like The World Under Heaven and The Invisible Guardian combine mature video models + short drama user scale + gaming monetization power for higher certainty.

A Familiar Taste

Today some interactive content products tout "North America download chart #1" as a selling point. This reminds me of a social-sector founder who also hit #1 on North America download charts a few years back.

Back then, with strong growth momentum and decent day-1 retention, he pumped user acquisition to 300-500K daily new users, surpassing TikTok in total downloads. But he soon discovered long-cycle retention wouldn't hold—the user pool was leaking as fast as it filled.

Investors came asking: "Why stop spending?" Though he already knew that with retention stuck, $2 million was essentially the user acquisition ceiling. But there was no choice—fundraising had to continue, so he kept burning money.

That product has now been shut for over three years.

Sound familiar? Doesn't this resemble a narrative you may have heard lately: certain companies spend $1 million on growth, convert half, then raise $2 million from investors?

Those who lived through the 2021 new consumer frenzy will find this story even more familiar.

In that era when brands could be manufactured through Douyin, Xiaohongshu, and Taobao spending, startups "made" traffic to raise from investors. During negotiations, founders would lose money driving sales volume to command higher valuations.

This is mutual hostage-taking around traffic. Once begun, no one escapes easily.

But Douyin wasn't built on bought traffic. An early ByteDance employee turned entrepreneur told us that internally, ByteDance viewed growth and monetization as the same thing—growth was never for growth's sake. "ByteDance's capital efficiency was extremely high. Douyin could later acquire users for $20, while Kuaishou had to give up."

Interaction has indeed brought innovation on the content side. But beyond this, these companies must still answer how to defend against ByteDance's trifecta outside the product itself—though this question has grown somewhat cliché.

A ByteDance-alumni founder told me the company has already begun building AI interactive content, based on cameras and meme formats, as a feature within TikTok.

Moreover, why should AI-era content platforms retrace the mobile era's path?

The next Douyin won't be Douyin. You don't win big by riding coattails. But if the attitude is "we raised money, might as well enjoy it," then forget I said anything.

Cover image: Caravaggio, The Cardsharps, 1594, Kimbell Art Museum