Doubao AI Agent Rivals Meta's Muse After Six-Day Sprint

Meta's Muse needed less than two weeks to redraw the AI agent map. After the personal AI agent hit 2.8 million downloads in its first 12 days, ByteDance's Doubao AI agent team reportedly dropped everything over the Mid-Autumn Festival holiday to build a direct competitor. The reported timeline: six days, end to end.

According to a leak reported by BlockBeats and carried by KuCoin's news flash on September 28, the Doubao product line received an urgent task during the holiday period: system integration finished by September 27 and an internal test version ready for review by September 30. Multiple upstream and downstream teams were pulled into the sprint, which sources described as a high-priority project with a total development window of just six days.

The leak, attributed to Chinese tech outlet "Big Company Daily" via ME AI, frames the Doubao AI agent project as an explicit answer to Muse rather than a routine feature update. Whether a six-day build can produce a durable competitor remains the open question — but the speed of the response says a lot about how seriously the industry is taking Muse's debut.

Six days, one holiday: inside Doubao's reported crash build

Mid-Autumn Festival is one of China's most important holidays, a time when product teams normally go quiet. Instead, the Doubao AI agent effort reportedly consumed it. The milestones laid out in the leak are unusually specific: integration work done by September 27, an experimental version ready for internal review by September 30.

Doubao is no underdog in its home market. QuestMobile's 2026 half-year report puts the app at 382 million monthly active users in China as of June 2026, nearly three times DeepSeek's 129 million, according to a September comparison by Second Talent. Alibaba's Qwen sat between them at 167 million. Doubao has already been shifting from chatbot toward task-executing agent, which may explain why ByteDance leadership believes a credible Muse rival can be assembled at sprint speed.

That home-field scale matters. Reuters, citing Apptopia data, reported that Muse reached 2.8 million downloads in 12 days — an impressive figure that still represents a fraction of Doubao's installed base. A Doubao AI agent shipped to an audience of hundreds of millions would arrive with distribution most startups can only dream of.

ByteDance has also been reorganizing for the agent era. An internal email on July 30 consolidated the company's AI office products around Feishu and Doubao, as reported by Vocal Media, folding agent ambitions into the products people already use daily. In China, the race had already moved from "whose model is smarter" to "whose agent sits inside the workflow": Tencent's WorkBuddy prototype went up in two days in January and drew 20.97 million monthly visits by June, while Alibaba merged three agent products into Qwen Office in July and passed 30 million registered users within a month.

Why Muse spooked the industry

Muse launched on September 8 in the US and Canada as a personal AI agent that goes beyond chat. It connects to user-authorized services such as email and calendar, operates a browser, and carries out multi-step tasks including shopping and travel — continuing to run in an isolated cloud environment even after the user exits the app, KuCoin's flash summary notes.

The early numbers were startling. Apptopia data cited by Reuters showed 2.8 million downloads in the first 12 days. On a like-for-like basis, Muse logged 1.8 million iOS downloads in the US and Canada versus 1.3 million for ChatGPT during its own first 12 days after its mobile launch, according to analysis by The Motley Fool. Muse briefly took the number one spot among free iOS apps, pushing ChatGPT to number two.

Engagement figures were even more lopsided. The same Apptopia data showed Muse with 642,000 daily active users in its first 12 days, compared with 231,000 for ChatGPT over the comparable period. Early reviewers gave the app a 4.9-star rating on iOS across more than 36,000 reviews.

Wall Street noticed. Meta's shares surged more than 20 percent after the September 8 launch, adding over $200 billion in market capitalization, Reuters reported on September 22. J.P. Morgan analysts said Muse could become the most widely used consumer AI application since ChatGPT, citing strong product-market fit, Meta's distribution reach, and a free tier that lowers the barrier to trial.

The monetization story matters here too. Muse is free at the basic tier, with paid plans at $20 and $100 a month for heavier use. Bank of America analysts told clients that early users were responding to the agent's ability to execute multi-step workflows with limited intervention — a sign that consumers may pay for agents that do things, not just answer questions. That thesis is precisely what the reported Doubao AI agent sprint is trying to preempt.

What a six-day rival would need to prove

Speed is the story of the leak, but speed is also its biggest question mark. A personal agent that touches email, calendars, and purchases lives or dies on trust, and trust is usually built slower than six days. Muse's edge so far is execution reliability: users rate it highly because it completes workflows, not because it demos well. A Doubao AI agent rushed out over a holiday would need to match that reliability on day one or risk a damaging first impression with hundreds of millions of users.

The internal test milestone of September 30 will be the first real signal. If the sprint produces a working build by then, ByteDance's bet is that iteration speed matters more than launch polish — ship the agent, then harden it with real user traffic. If the test slips, the six-day legend becomes a cautionary tale about announcement-driven development.

There is also the question of scope. The leak describes a product targeting Muse, Meta's consumer personal agent, but ByteDance's agent ambitions span office workflows through Feishu and the broader Doubao ecosystem. A consumer-facing Doubao AI agent would complement the company's existing push to make Doubao the default agent interface for Chinese users, much as Muse leans on Meta's family of apps for distribution. Apptopia reported that 95 percent of early Muse adopters were also Facebook users — distribution, not just model quality, is doing heavy lifting in this race.

For the wider agent ecosystem, the pattern is becoming familiar: a breakout launch, then a frantic response cycle measured in days rather than quarters. Earlier coverage on genznewz detailed how Muse's subscription-management features alarmed billing businesses, and how OpenAI's rogue-agent weekend raised the stakes on agent oversight. The Doubao sprint adds a third data point: the agent wars are now fought on holiday weekends, and the combatants are the largest consumer platforms on earth.

Whether the six-day Doubao AI agent turns out to be real, vapor, or somewhere in between, one thing is already clear: after Muse's debut, nobody in the industry is treating personal AI agents as a side project anymore.