On October 26, a rumor that had circulated for days hit a reversal. Reports had claimed that Zhou Chang, one of the core research leads at ByteDance's Seed AI, had resigned and was preparing to launch an embodied-intelligence startup, with approaches from major funds and leading model companies. That same day, a person close to ByteDance gave a flat denial: Zhou has not left.
Why One Researcher Matters This Much
The market's sensitivity comes down to a résumé with few domestic equivalents.
Zhou earned his PhD in computer software and theory from Peking University in 2017, served as an algorithm expert at Alibaba's DAMO Academy, and went on to lead the Tongyi large-model effort. In June 2024, just before he left Alibaba, Qwen2 topped the open-source leaderboards, outperforming the then-benchmark Llama 3-70B. At ByteDance he initially worked on multimodal interaction and world models, taking over the visual generation business in mid-2025. Seedance 2.0, the video generation model released in February 2026, drew global attention and is widely credited to his team. By July 2026 his remit covered multimodal models, visual generation, and world models, extending into robotics.
Between Rumor and Denial
The full chain of events is worth noting: first came word of a resignation aimed at an embodied-AI startup, followed by an explicit denial from the person's own employer. Around key personnel at top model companies, this rumor-then-denial pattern has recurred for years, and it mostly reflects how the market prices scarce talent — embodied intelligence being one of the hottest funding themes right now.
For ByteDance, the Seedance line and visual generation are in a head-to-head race with global rivals, and continuity at the top directly affects the pace of iteration. The denial removes short-term uncertainty, but the contest for elite AI researchers is clearly not slowing down.