Star Point · Decision Simulation Run your own simulation

ByteDance merges TRAE and Coze into Doubao: Will developers churn?

Tech & AI Simulation date 2026-09-19

Decision question: Decision: In late August 2026, should ByteDance merge the two product lines—TRAE (an AI coding tool) and Coze (an agent-building platform), which already have independent user bases—into the new office AI product 'Doubao Work,' and unify them under the 'Doubao' brand externally?

Core judgment

A path of underlying integration with front-end layering should be adopted, abandoning a one-size-fits-all brand unification.

Top recommendation

Immediately establish a special emergency task force led by the TRAE team to conduct one-on-one communication with top developers at the highest risk of churn; if necessary, restore some third-party model access options to stabilize the situation.

Where they stand

Oppose 64.6% · Support 35.4%

Based on 65 simulated statements by 17 virtual roles — not a real poll, and not the actual positions of these organizations or people

Camps and reasons

  • Oppose: Professional developers and enterprise customers follow entirely different trust logic; stuffing an AI-native IDE into a mass-chat brand will directly dilute professional credibility and raise procurement thresholds.
  • Support: Agree with the strategic resolve to concentrate resources on key directions, but emphasize that technical architecture independence and smooth migration plans are prerequisites for advancement.

Biggest risks

  • Developer ecosystem churn: Forcing brand merger will cause core developers of TRAE and Coze to defect to competitors like Cursor due to collapsed trust, breaking the data Recirculation flywheel.
  • Enterprise customer trust erosion: Mass-market labels covering professional identifiers will increase the cost of internal review questions within enterprise IT departments, directly increasing resistance to B-end procurement.
  • Competitor window period exposure: If model scheduling and experience re-adaptation during the transition period encounters accidents, it will leave opportunities for Tencent WorkBuddy to poach targeted users and seize the market.

Key uncertainties

  • TRAE sub-brand existence or elimination plan (watch: In the officially released final architecture diagram, whether TRAE is retained as an independent sub-brand keeping its technology and distribution entry point, or completely folded into the Doubao menu.)
  • Stability during the mandatory replacement period of the Seed model (watch: Whether code generation accuracy and developer activity levels see a cliff-like decline during the transition period when TRAE integrates the self-developed Seed-Code model.)
  • Coze ecosystem equity contract (watch: Whether written guarantee agreements targeting third-party developers are issued before and after the merger, explicitly stating that distribution logic and monetization ratios will not be unilaterally adjusted unfavorably.)
  • Competitor WorkBuddy's market actions (watch: Whether Tencent WorkBuddy launches targeted developer recruitment plans or enterprise-side subsidy policies to grab professional users during ByteDance's transition period.)

In their own words

The quotes below are what the virtual roles said in the simulation, reproduced verbatim; they are simulated dialogue, not real statements by these organizations or people.

“You wouldn't expect to rename VS Code to 'Chat Tool'.” — Hong Dingkun (virtual role)
“If they merge? Then we'll just jump ship to Cursor.” — TRAE developer (virtual role)
“Without securing an independent ecosystem, nothing else matters.” — Coze developer (virtual role)
“The quality of code depends on the stability of the foundation, not whose logo is hanging on the roof.” — Tan Dai (virtual role)
“We take this step by step; once the direction is set, our pace must be steady.” — Lyu Bo (virtual role)
“I strongly oppose violent refactoring without contingency plans.” — Internal engineer (virtual role)

In late August 2026, ByteDance faced a key decision: whether to merge its AI programming tool TRAE (with millions of monthly active professional developers) and the Coze platform (which hosts numerous third-party intelligent agents) into the newly established office AI product "Doubao Work" and unify them under the "Doubao" brand. In late July, the Feishu and Doubao teams had completed organizational integration, with Zhao Qi overseeing products and Tan Dai responsible for commercialization (Source: material). Facing the strong momentum of Tencent WorkBuddy in the office Agent track, resource concentration became a group strategic consensus. This simulation covered 19 key stakeholders, including executives of the involved parties, teams from various product lines, internal and external developers, and competitors. Among a total of 73 posts, after excluding 8 from the observation Seat (observation gallery), the 65 posts entering stance statistics showed that 64.6% opposed merging TRAE and Coze into Doubao Work and unifying the brand, while 35.4% supported it. The overall public opinion trend exhibited a clear characteristic of "strategic Agree (agreement) but execution anxiety"—most stakeholders acknowledged the motivation to integrate against competitors but expressed strong concerns about developer churn, downgrade of enterprise trust, and chaotic experience during the merger caused by crude unification of the mass-market brand. Label explanation: "Material fact" in the report refers to content from user-provided materials and the relationship map; "verified online" refers to external data found via online search this time (with source URLs attached); "simulation viewpoint" refers to the statements and figures of virtual characters in the simulation, which are fictional deductions provided for inspiration and reference only.

Executive brief

Core judgment: Under the base case, an integrated path of unifying the underlying layer while maintaining a segmented front-end should be adopted, abandoning a one-size-fits-all brand unification. The technical architecture can be merged, but TRAE and Coze must retain their independent brands and distribution channels, relegating Doubao to the background as an intelligent foundation.

Stance landscape: Most key stakeholders oppose direct brand unification; the minority of supporters all attach stringent execution prerequisites. All parties share a consensus on preserving the developer ecosystem, with disagreements centered on whether a mass-market brand can cover professional tools.

Greatest risk: Forcing brand consolidation before the technical underlayer has smoothly transitioned and before the developer protection covenant is formalized in writing risks a triple resonance of professional developers leaving, enterprise customers waiting and seeing, and competitors exploiting the void.

Turning-point signals: If the official team ultimately folds TRAE completely into the Doubao menu rather than retaining it as a sub-brand, or if code generation accuracy plummets during the model replacement period, or if no written rights protection for third-party developers is issued, or if competitors launch targeted poaching subsidies, it indicates that the situation has deviated from the base judgment.

Primary recommendation: Immediately establish a special emergency task force led by the TRAE team to conduct one-on-one communications with top-tier developers at the highest risk of churn. Restore partial third-party model access options if necessary to stabilize the footing.

This executive brief is a summary of the full report's conclusions; see corresponding chapters in the main text for the basis and sources of each judgment.

Distribution of key stakeholder stances: Deep fragmentation at the execution level under strategic consensus

Chapter 1 focuses on the stance distribution of 19 key stakeholders regarding the proposition that "ByteDance should merge TRAE and Coze into the Doubao workflow and unify the brand." Across 5 rounds of simulation, a total of 73 posts were recorded, of which 65 entered the stance statistics (excluding 8 observer remarks). The results show that the opposition camp holds an absolute majority, with 42 posts (accounting for 64.6%); the support camp had 23 posts (accounting for 35.4%), and supporters generally attached strict execution prerequisites. The overall public opinion presents a characteristic where there is consensus on strategic direction but deep division at the execution level. (Source: simulation)

Key points of divergence and key uncertainties: Misalignment between brand perception and technological foundation

(Source: simulation) The simulation record shows a significant split among 65 valid posts regarding the proposition that ByteDance should integrate TRAE and Coze into the Doubao workflow and unify their branding, with 64.6% opposing and 35.4% supporting (Source: simulation). This opposition is not a dispute over strategic direction—all parties acknowledge the necessity of resource concentration—but rather focuses on execution-level disagreements across three dimensions: brand perception, developer migration paths, and the pacing of the transition period.

Brand Perception Misalignment: Can the Mass Traffic Umbrella Carry Professional Trust?

The core argument of the opposing camp lies in 'the incompatibility of two evaluation systems.' A simulation character based on Hong Dingkun explicitly points out that professional developers and mass chat users follow entirely different trust logics; placing an AI-native IDE under a mass brand would directly dilute professional credibility (Source: simulation). Statements from enterprise professional user roles further corroborate this: B2B procurement prioritizes data isolation, permission control, and long-term service commitments. If the brand tone is covered by mass-market labels, the cost of internal review skepticism will rise (Source: simulation).

The supporting camp attempts to resolve conflicts through 'smooth migration.' The Doubao team role proposes two bottom lines: first, the mental footprint of Doubao's existing 200 million daily active mass users must not be disrupted (Note: This data source comes from media reports; third-party monitoring agency QuestMobile data shows DAU was 145 million on New Year's Eve, indicating a discrepancy in metrics); second, the reliable image in professional fields must be reshaped before integration (Source: simulation). However, the Feishu team role, drawing on personal experience, warns that even in office scenarios, simply flattening professional identifiers increases trust costs for enterprise clients, and sensitivity in the Coding track would only be higher (Source: simulation).

Developer Migration: The Game Between Ecosystem Assets and Organizational Integration

TRAE Team and Seed Team have formed cross-camp consensus on 'the independence of the technical foundation.' The TRAE team role emphasizes that its monthly active users in the millions were accumulated by retaining decision-making control and obsessing over user experience; the developer ecosystem is not an asset that can be simply migrated by changing logos (Source: simulation). The Seed team role adds from the perspective of model iteration: once the loop of Coding data Reflow and Dogfooding is broken due to user loss caused by brand switching, ByteDance's AI foundation will develop loopholes (Source: simulation).

Anxiety among Coze developers centers on distribution and monetization rights. In the simulation, the Coze developer role worries that after unifying the brand, original channels will be taken away and revenue-sharing mechanisms changed, viewing independent ecosystems as a bottom line (Source: simulation). Although Coze developers' concerns about code privacy and model boundaries differ in specific demands from those of TRAE developers, they point to the same risk: ecosystem partners are extremely sensitive to the stability of platform policies.

From external facts, alternatives to TRAE are already very mature. Online comparative reviews show that TRAE uses the exact same underlying VS Code architecture as Cursor, and its basic version is permanently free, making migration costs extremely low (Source: verified online, url: cloud.tencent.com). This means that once developers develop distrust, the resistance to 'voting with their feet' is minimal.

Key Uncertainties: Four Variables Determining the Outcome

No party currently has a definitive answer, but the evolution of the following four variables will directly determine the success or failure of the integration, requiring close monitoring:

  • Fate of the TRAE sub-brand: What to watch—whether TRAE is retained as an independent sub-brand preserving technical and distribution entry points, or completely folded into the Doubao menu in the officially announced final architecture diagram.
  • Stability during the mandatory replacement period for the Seed model: What to watch—TRAE is currently in a transition period of integrating the self-developed Seed-Code model (Source: material); need to observe whether there is a cliff-like drop in code generation accuracy and developer activity during this period.
  • Rights contract of the Coze ecosystem: What to watch—whether written guarantee agreements for third-party developers are issued before and after the merger, clarifying that distribution logic and monetization ratios will not be unilaterally adjusted to the detriment of developers.
  • Market moves of competitor WorkBuddy: What to watch—whether Tencent's WorkBuddy launches targeted developer recruitment plans or enterprise-side subsidy policies during ByteDance's transition period to snatch away lost professional users.

Risk warning: Chain reactions that may be triggered if the merger is pushed forward under current conditions

Against the proposition that 'ByteDance should integrate TRAE and Coze into the Doubao workflow and unify their brand,' the 65 valid posts in the simulation record show an overwhelming majority of opposition (Source: simulation). Ignoring this deep rift and forcibly pushing for brand unification under the current status quo will not only fail to achieve the strategic goal of resource concentration but may also trigger a chain reaction of damage across three dimensions: the developer ecosystem, enterprise customer trust, and competitive defense. The following section evaluates the core risks triggered by forcing this integration one by one.

Developer Ecosystem Attrition: Trust Collapse from Professional Tools Being 'Absorbed' by a Mass Brand

The migration risk for the TRAE developer community is the primary threat. Since its launch in January 2025, TRAE has rapidly accumulated a large professional user base due to its positioning as an AI-native IDE. By the end of 2025, total registered users exceeded 6 million, with monthly active users surpassing 1.6 million (Source: verified online, url: csdn.net). This group's loyalty is built on code privacy, model capability boundaries, and decision-making control, rather than brand halo. A simulation role modeled after Hong Dingkun sharply pointed out that placing a professional IDE under a mass chat brand directly discounts trust, using the analogy that VS Code would not rename itself as a chat tool (Source: simulation).

"Developers want certainty. Once the brand is diluted, they will switch to Cursor."

More critically, TRAE is currently in a sensitive transition period involving data feedback loops and model replacement—multiple application departments have been mandated to use ByteDance's self-developed Seed-Code model (Source: material). A simulation role modeled after the Seed team warned that if brand switching causes user churn, breaking the data feedback flywheel will create vulnerabilities in the entire coding infrastructure (Source: simulation). Against the backdrop that competitors like Chaitin Technology have launched open-source AI programming platforms supporting private deployment and multi-model access (Source: verified online, url: cnblogs.com), forcing integration acts as a catalyst for developers to leave.

The monetization panic among Coze's third-party developers is equally Cannot be ignored (cannot be ignored). Third-party developers on the Coze platform rely on distribution channels and monetization rights for survival. Simulation roles modeled after Coze developers repeatedly expressed concern that the greatest fear after brand unification is having distribution channels taken away and monetization rights cut, stating directly that if the independent ecosystem is not preserved, they will vote with their feet and switch to other platforms (Source: simulation). This panic is not unfounded but is an amplified expectation based on precedents observed after the Feishu team was integrated.

Enterprise Customer Trust Erosion: C-end Traffic Umbrella Cannot Cover B-end Procurement Logic

Brand tone misalignment will directly increase procurement resistance at the enterprise level. Doubao's core perception remains that of a consumer-facing conversational assistant; its peak Daily Active Users (DAU) during the Spring Festival were monitored by QuestMobile to be 145 million (the original source text states '200 million', which is inaccurate as this figure comes from media sources rather than third-party monitoring data). However, enterprise client procurement evaluation follows a completely different coordinate system. A simulation role modeled after Tan Dai emphasized that enterprise procurement does not look at brand hype but at whether it can withstand key business scenarios; no matter how high the C-end DAU is, it constitutes a separate coordinate system from the B-end trust structure (Source: simulation).

A simulation role modeled after the Feishu team corroborated this judgment, pointing out that enterprise clients value data isolation, permission management, and service sustainability. Simply reducing the brand and erasing professional identifiers impacts most severely the partners who have invested real capital in the ecosystem (Source: simulation). When TRAE and Coze's professional identifiers are folded into a mass-brand container, the questioning cost for enterprise IT departments during internal reviews will rise significantly, and the professionally established credibility will be directly called into question.

Exposure of Competitive Window Period: Internal Chaos Hands Knives to Rivals

Experience fractures during the transition period will become a breakthrough point for Tencent WorkBuddy. One of the backgrounds for ByteDance's organizational adjustment is precisely to counter Tencent WorkBuddy's offensive in the office Agent track (Source: material). A simulation role modeled after Tencent WorkBuddy explicitly stated that if brand unification affects distribution light and monetization rights, ecosystem attrition is inevitable, emphasizing that enterprise users buy reliable delivery, not brand stories, and the market will vote with their feet (Source: simulation).

A simulation role modeled after internal engineers issued a stern warning: overlaying brand restructuring onto the transition period of Seed model replacement means model scheduling, evaluation benchmarks, and developer trust all need to be re-harmonized. If accidents occur and windows are ceded to competitors, the consequences are unbearable (Source: simulation). External observer roles also pointed out that developer loyalty for TRAE and Coze is tied to product usability and stable monetization; if API strategies or revenue-sharing mechanisms change, the risk of attrition is substantial (Source: simulation).

In summary, the greatest risk of forcing integration lies not in strategic direction errors, but in execution rhythm mismatch—making rash moves on the 'skin' of the brand before the technical foundation has completed a smooth transition and developer protection contracts are formalized in writing Extremely easy (easily) triggers a triple resonance of developer exodus, enterprise customer hesitation, and competitors taking advantage of the void.

Recommended actions: Step-by-step integration path based on scenario simulation

Based on the analysis of stance distribution, core divergences, and potential risks in the preceding text, this chapter provides decision-makers with a set of structured action recommendations across scenarios, centered on the proposition that 'ByteDance should merge TRAE and Coze into Doubao Work and unify the brand.' The simulation records show that among 65 valid stakeholder posts, opposition voices constituted an absolute majority (Source: simulation), meaning that any integration move cannot detach from the fundamental base of developer trust. In light of the established fact that the Feishu team has already been merged into the new Doubao product team as of July 30, 2026 (Source: material), and the current situation where Doubao Pro officially launched on June 24, 2026, adopting tiered pricing to penetrate productivity scenarios (Source: verified online, url: cls.cn), we recommend abandoning 'one-size-fits-all brand unification' and instead adopting a strategy of 'backend integration, frontend stratification,' advancing according to the following three scenarios.

I. Base Case: Parallel Architecture Integration and Brand Isolation (Recommended Path)

Trigger conditions: Internal assessment confirms that the transition period for replacing the Seed-Code model in TRAE is not yet fully stable, and the distribution rights contracts for third-party developers on Coze are still being executed.

This is the path most aligned with current realities. At the organizational level, referencing the precedent of Feishu's merger, the R&D teams for TRAE and Coze can be physically integrated into the new Doubao product team led by Zhao Qi, achieving data Return flow and unified scheduling of the technical foundation; however, in terms of external branding, it is imperative to retain TRAE’s sub-brand identity as an independent AI-native IDE, and Coze must also retain its independent entry point as an agent platform. Stakeholders modeled after the Doubao team also advocated for a smooth migration route, emphasizing that professional IDEs should not be forced into the shell of a chat assistant (Source: simulation).

Specific action recommendations are as follows:

  1. For R&D management: Clarify a dual-track system where 'the intelligent core belongs to Doubao, and technical entry points belong to TRAE/Coze.' The Coding line represented by Hong Dingkun will continue to lead TRAE’s technical iterations, ensuring that the auxiliary development tools used by approximately 80% of internal engineers do not suffer experience degradation due to changes in reporting lines (Source: material).
  2. For the developer ecosystem: Concurrently with the merger announcement, issue written guarantee agreements to TRAE’s million monthly active users and Coze’s third-party developers, committing to no unilateral adverse adjustments to API strategies, code privacy boundaries, and revenue-sharing ratios.
  3. For enterprise customers: Connect the office capabilities of Doubao Pro (such as AI Spreadsheets and Deep Research, already launched) with TRAE’s coding capabilities at the scenario level, rather than through brand-level coverage, allowing B-end customers to perceive this as an 'expansion of professional toolboxes' rather than a 'descent of mass-market brands' (Source: verified online, url: cls.cn).

Signals to watch: Whether there are abnormal fluctuations in the weekly activity of the TRAE developer community; whether the month-on-month growth of new agents added to the Coze platform declines.

II. Best Case: Deep Integration Drives Office Agents to Surpass Competitors

Trigger conditions: The evaluation performance of the Seed-Code model on TRAE meets all standards comprehensively, and developer acceptance is extremely high; meanwhile, Tencent’s WorkBuddy slows its offensive in the enterprise sector, leaving a strategic window of at least one quarter.

In this scenario, deeper brand consolidation can be attempted. In the simulation, stakeholders modeled after the Volcano Engine team believed that if the smooth migration plan for developers is designed upfront and independent decision-making power for the technology line is not diluted, the merger can form true synergistic effects (Source: simulation). At this point, TRAE’s advanced features can be deeply embedded as 'Expert Mode' plugins for Doubao Enterprise Edition. Unified promotion can use the 'Doubao Work' matrix concept externally, but the TRAE logo should still be retained within the IDE interface.

Specific action recommendations are as follows:

  1. For the commercialization team: Leverage Doubao’s massive user base that reached a Fault line-leading position after the Spring Festival (according to multiple sources, DAU reached 200 million after the 2026 Spring Festival, but this data is not from third-party monitoring Caliber, so information discrepancies need to be noted) to demystify TRAE’s professional capabilities for the broader developer group, launching lightweight programming Agent subscription packages.
  2. For ecosystem operations: Host cross-product-line joint developer conferences, combining Coze’s agent building with TRAE’s code generation to demonstrate end-to-end capabilities of 'generating usable business systems with one sentence,' using actual delivery results to counteract brand mindset conflicts.

Signals to watch: Renewal rates and upsell rates for early access clients of Doubao Enterprise Edition; changes in market volume for WorkBuddy.

III. Worst Case: Trust Collapse Triggers Competitor Harvesting

Trigger conditions: Forced unification of the brand leads to public resistance from TRAE’s core developers, or serious code quality incidents occur during the mandatory replacement of the Seed model, amplified and spread by external observers.

If professional tools are forcibly folded into a mass-market brand container, it Extremely easy triggers the chain risks warned about earlier. Stakeholders modeled after TRAE developers issued severe warnings, believing that once professional independence is sacrificed, developers will directly switch to competitors like Cursor (Source: simulation). At this point, it is necessary to brake immediately.

Specific action recommendations are as follows:

  1. For top decision-makers: Suspend all brand unification publicity Oriented to C End and cut off the source of public opinion fermentation. Reiterate Liang Rubo’s principle of 'direction is set, steps must be steady,' shifting the assessment focus from brand integration progress to delivery continuity (Source: material).
  2. For frontline technical teams: Establish a special emergency task force led by the TRAE team to conduct one-on-one communication with head developers at highest risk of churn; if necessary, restore access options for some third-party models to stabilize the footing.
  3. For PR and marketing: Prepare defensive contingency plans to prevent Tencent WorkBuddy from launching targeted poaching subsidies. The unified external narrative should be 'upgrading underlying capabilities, keeping frontend experience unchanged,' avoiding falling into self-proving traps.

Signals to watch: Search index and registration conversion rates for competitors such as Cursor and Claude Code in domestic developer communities; density of reports by tech media regarding internal turmoil at ByteDance.

In summary, the essence of the merger is resource restructuring, not name elimination. Guarding TRAE’s AI-native IDE genes and Coze’s open ecosystem contracts, while letting Doubao recede to the background as the intelligent foundation, is the optimal solution that balances strategic concentration with professional trust.

Materials and Sources

Verified Online Sources

Source of Materials

This page is the finished report of this simulation. Open the interactive version to replay it round by round and talk to the roles, or start a simulation of your own.