Tian Keyu, the founder of a startup that has not yet officially revealed its name or launched any product, has recently secured nearly $30 million in funding. This investment has catapulted the company’s post-money valuation to $200 million. Leading venture capital firms, including 5Y Capital, IDG, and Oriental Bell Capital, have participated in this round of funding. Despite having a relatively small team of around 10 members, the company has ambitious plans to release a comprehensive world model by approximately 2027.

Two years ago, Tian Keyu interned at ByteDance but was terminated due to serious policy violations. Interestingly, during this same period, he co-authored a paper that earned the Best Paper Award at NeurIPS 2024. In this paper, he introduced an innovative method for hierarchically generating images, significantly boosting the speed and efficiency of image creation. His entrepreneurial vision builds on this technical breakthrough, aiming to construct a visual symbol system comprising roughly 200,000 symbols. By inputting 100 million hours of video data, the company seeks to slash video generation costs by at least an order of magnitude.

Currently, the world model sector lacks a stable product format or universally accepted evaluation standards. Nevertheless, investors remain optimistic about its potential, viewing it as the future of AI in managing spatial, temporal, and action-related challenges. Tian Keyu’s technical approach and his paper’s achievements are seen as key factors influencing the company’s valuation.

It’s worth noting that during his time at ByteDance, Tian Keyu modified code to disrupt project training out of dissatisfaction with resource allocation, prompting legal action from ByteDance. He later clarified that his intent was to free up computational resources and ultimately settled the dispute by paying 500,000 yuan in compensation and forfeiting his wages.

Globally, competition in the world model sector is heating up, with both international tech giants and domestic companies actively developing and refining related technologies. The industry’s ambition extends beyond creating realistic visuals—it also aims to grasp the relationships between actions and their outcomes.