Why Did ByteDance's Coze Suddenly Go Quiet? Outdated AI Workflows and Shifting Internal Priorities

From Must-Learn Tool to Nobody Cares: Coze's Turning Point

Anyone following AI has almost certainly heard of Coze. Before AI coding took off, Coze—with its self-developed node-based workflow—was one of the few visual AI application-building platforms in China, and at one point was a required course for anyone getting into AI. According to analysis by a Bilibili creator, back then you could hardly find AI learning content online that didn't involve Coze, and this near-monopoly quickly built up user awareness.

But today, Coze's voice has clearly grown quieter. The creator states outright that ByteDance has almost given up on Coze internally, and the project team may even face dissolution. This judgment isn't unfounded—Coze's core selling point, the "node-based workflow," has come to look outdated as AI coding capabilities rapidly evolved. Users no longer need to manually drag and drop nodes to wire up logic; code generation and automation tools are replacing that kind of operation.

Node-based workflow (visual workflow) is a way of orchestrating AI tasks by dragging and dropping nodes and connecting their input/output relationships. Typical nodes include LLM calls, conditional branches, and API requests. It lowers the barrier for non-programmers to build AI applications, but flexibility is limited by the platform's preset nodes and expressive logic. As large models' code generation capabilities improve, users can describe logic directly in natural language and have AI generate executable code, bypassing manual wiring—this is the fundamental reason workflow tools face replacement.

ByteDance's Anxiety: Getting Beat by Workbody and Losing Its Bearings

Coze's decline becomes clearer when viewed within ByteDance's overall AI strategy. The creator notes that ByteDance is currently in a state of obvious anxiety, describing its passive position as "getting beat by Workbody and losing its bearings." To respond to competition, ByteDance internally merged Doubao with Feishu, attempting to integrate conversational AI with office scenarios.

That said, ByteDance is also pushing Tree Work, aiming to compete with Workbody. But based on the creator's observations, Tree Work hasn't gained much momentum, and internal confidence in it isn't strong. This raises a practical question: should limited energy be bet on Tree Work, or on the Doubao-Feishu merger line? ByteDance is still balancing internally and hasn't given a clear answer yet.

Workbody is one of ByteDance's main competitors in the AI coding and automation space (the article doesn't specify its exact product form; it may refer to a type of AI coding or agent platform). Against the backdrop of rapidly iterating AI coding tools, Workbody has gained market attention with more natural interaction and stronger code generation capabilities, putting pressure on ByteDance's related products. This prompted ByteDance to adjust resource allocation, merge Doubao with Feishu to integrate conversational AI with office scenarios, and push the Tree Work project to try to compete.

Coze's Two Pivots: From Workflow to Coding, Neither Hit the Rhythm

Coze's problems aren't just external competition—they're also about its own repeatedly wavering positioning. The creator points out that Coze's early popularity came from node-based workflows. When AI coding hadn't yet become mainstream, this interaction approach genuinely lowered the barrier to building AI applications, and domestically Coze was basically the only one offering a similar experience.

But the workflow model quickly revealed its ceiling. As AI coding tools improved, users preferred to generate logic directly in natural language rather than manually wiring nodes. Coze then pivoted toward coding, but the creator's assessment is blunt—"doing coding was clearly over as soon as it started," meaning the pivot didn't create competitiveness. Neither pivot hit the rhythm, leaving Coze in an increasingly awkward position within ByteDance's system.

AI coding refers to technologies and tools that use large language models to automatically generate, complete, or modify code (such as GitHub Copilot and Cursor). Its core advantage is that users can get runnable code simply by describing requirements in natural language, greatly lowering the programming barrier and improving development efficiency. Once AI coding capabilities mature, logic that previously required manually wiring visual workflows can be implemented directly by model-generated code, weakening the value of intermediate orchestration tools—this is the context for Coze's pivot to coding failing to create competitiveness.

The Chain Reaction in Org Structure: Merger or Dissolution Is Only a Matter of Time

In ByteDance's system, if a project can't find a strategic position for a long time, organizational adjustments usually follow. The creator judges that Coze's team has already undergone obvious changes within ByteDance, and within the next two to three months it will either be merged into another department or the project will be completely dissolved. This time-window judgment includes personal speculation, but the direction is consistent with Coze's current declining presence.

From a product-logic perspective, Coze's predicament is representative: when underlying models and coding capabilities iterate rapidly, intermediate-layer visual orchestration tools are easily bypassed. Users want results, not the building process. If Coze can't redefine its value—for example, pivoting to enterprise-level agent management or binding deeply with the Doubao ecosystem—marginalization is almost inevitable.

Takeaways for Practitioners: Don't Treat Workflow Tools as Long-Term Skills

Coze's rise and fall is a reminder for AI practitioners. The core message the creator conveys in the video is that a tool going viral doesn't equal long-term value. Node-based workflows were efficiency tools at a particular stage, but once AI coding matures, the transferability of such skills drops rapidly.

More noteworthy is the strategic logic of big companies like ByteDance: resources concentrate toward areas that can fight tough battles. The Doubao-Feishu merger and the push for Tree Work both show ByteDance reordering its priorities. Coze isn't the first project to be adjusted, and it won't be the last. For ordinary users and developers, rather than betting on a specific tool, it's better to watch the trend of underlying capability changes—after all, tools become outdated, but understanding the boundaries of AI capabilities won't.