The leader of TikTok's primary site recommendation algorithm, Song Yang, has unexpectedly departed to take up a position at Meta.

Simultaneously, TikTok has initiated a phased rollout of a 'Reduce AI Recommendations' toggle, allowing users to diminish the share of algorithm-recommended content to a minimum of 30%. This move represents the most significant traffic concession in the platform's history.

On November 20, several sources verified that Song Yang (also known as Adam Zhang), the former head of TikTok's main site video recommendation algorithm, has recently left his role and swiftly transitioned to Meta. There, he assumes full responsibility for the recommendation business of Instagram Reels.

This development marks the second leadership shift in TikTok's core algorithm division since Chen Zhijie left to pursue entrepreneurial ventures last year. It also represents the inaugural instance of a Silicon Valley titan directly luring away TikTok's top recommendation algorithm expert.

Song Yang's professional journey encompasses leadership roles at prominent platforms in both China and the United States. He graduated from Zhejiang University with a computer science degree in 2003 and went on to earn a Ph.D. from the University of Pennsylvania in 2008. He contributed to the development of CiteSeerX, one of the world's pioneering computer literature search engines. His career includes stints at Microsoft Research Redmond and Google. In 2020, he joined Kuaishou at the vice president level, overseeing algorithms for short videos, live streaming, and e-commerce across various scenarios. Midway through 2024, he joined TikTok, only to depart again after 18 months. Industry insiders characterize his career trajectory as 'moving too swiftly to be constrained by a non-compete clause.'

Sources within the industry disclose that Song Yang's responsibilities at TikTok were limited to overseeing the main site's Feed stream recommendations. Separate teams managed search, advertising, and e-commerce algorithms. Consequently, his exit is not expected to cause widespread technical disruption, but it may temporarily escalate recruitment and integration expenses. Meta CEO Mark Zuckerberg recently personally extended an offer to Song Yang, whose arrival is perceived as another blow to TikTok's technological infrastructure.

ByteDance opted not to provide a comment. Meta's China representative office stated, 'We refrain from commenting on rumors concerning individual appointments.'

In tandem with these developments, TikTok has commenced a phased rollout of a 'Reduce AI Recommendations' toggle. Users can adjust a slider on the 'Content Preferences' page to decrease the proportion of AI-recommended content from the default 90% to as low as 30%. Subsequently, the platform will display content in reverse chronological order, interspersed with semi-acquaintance content, and offer quantitative indicators such as 'AI exposures reduced this week.'

Should the data prove favorable, the feature will undergo a full rollout in the first quarter of the following year and might be introduced earlier in the EU region to adhere to the annual audit mandated by the Digital Services Act.

Under intense regulatory scrutiny, TikTok has opted to voluntarily relinquish traffic distribution rights. This decision underscores that the competition in the short video sector has transitioned from 'who understands you better' to 'who is bold enough to let you avoid content you don't comprehend.'