Doubao's layoff denial can't hide an executive reshuffle, and the trading team's contraction lays bare the commercialization struggles of large models.
When Doubao's PR head urgently flashed precise numbers—"only 11 people adjusted, 3 resigned"—to extinguish the layoff rumor that the "dialogue team was cut in half," this precise narrative control at the PR front desk could not obscure the high-level executive purge quietly completed by its underlying entity. Beneath the carnival surface of a general-purpose conversational large model whose daily active users have broken through tens of millions and which appears to sit firmly in the top spot for domestic C-end AI, the change of legal representative and executive director at its core operating entity is by no means an ordinary personnel rotation. Public relations denials by major tech companies have always followed the rule: the fewer the words, the bigger the matter.
Doubao's PR head quickly responded to the rumor that the "dialogue team was cut in half," characterizing the move as routine division-of-labor optimization involving "11 adjustments and 3 departures among 50 people on the trading team." However, layoff rumors widely regarded on Maimai and social platforms as "organizational shock" are often just the tip of the iceberg, while the baton in the deep water has long since quietly turned.
According to Tianyancha App, Beijing Chuntian Zhiyun Technology Co., Ltd., the operating entity of Doubao, recently completed a round of intensive executive replacements: former executive director, manager, and legal representative Yan Lin fully exited, and Guo Chang took over. In ByteDance's context, changes in the legal representative and core management are never a mere industrial and commercial filing process; they correspond to the reconstruction of battle lines and the transfer of resource allocation power.
This entity, wholly controlled by Beijing Douyin Information Service Co., Ltd., holds the full set of core assets including "Doubao Large Model" and "Doubao Teacher," and the handover of personnel power comes precisely at a sensitive turning point in the application and implementation of large models.
The outside world is keen to argue about exactly how many people Doubao laid off, but that is missing the point. What truly deserves to be asked is: why was it precisely the "trading team" that was cut?
Over the past year, the domestic large model track, after a frenzied technological breakthrough, quickly hit the hard ice of commercialization. As the C-end unicorn produced by ByteDance, Doubao surged forward in DAU thanks to Massive Engine's traffic-buying machine. But when the DAU ceiling became clearly visible, large model applications had to face the cruel test of monetization.
Previously, ByteDance tried to directly graft its familiar e-commerce livestreaming and local life logic onto AI dialogue scenarios. Forming a so-called trading team was nothing more than an attempt to replicate Douyin's myth that "traffic equals GMV."
Facts proved that the technological turning point has not yet supported a complex transaction loop. Large models are stunning enough at understanding intent and generating content, but once connected to a transaction chain involving fund payment, fulfillment and delivery, and after-sales disputes, the extremely high cost of hallucination tolerance becomes a fatal weakness. Users are willing to chat idly with a humorous AI and have it write weekly reports, but very few are willing to place an order for a piece of clothing in a chat box.
The marginalization and contraction of the trading team essentially means ByteDance encountered real resistance on its commercialization path.
Yan Lin's exit and Guo Chang's entry mark Doubao's farewell to the first half of blind trial and error and wild growth across all lines, and its entry into a tightening cycle in which group executives strictly scrutinize ROI. The "gravity of escaping mediocrity" that Liang Rutian has repeatedly emphasized, directly projected onto the large model department, means quickly stanching the bleeding from redundant exploration that cannot close the loop in the short term.
This organizational adjustment that was denied is by no means calm and uneventful. It strips away the false heat of the AI concept and reveals the real accounts on the major tech company's abacus. ByteDance is abandoning certain overly impatient commercialization fantasies and gathering its five fingers back together, and this pain has only just begun.