🎙️ Episode Introduction
What happens when "creating a world" becomes as simple as shooting a video?
In this episode of Signal and Noise, we sit down with Yachen Song, founder of VAST. From SenseTime's AI animation and gaming business, to his role as an early member in MiniMax's founding, to launching VAST — he has witnessed the rapid evolution of AI-powered content creation, and gradually spotted a gap that remains unfilled: while the barriers to creating text, images, and video keep dropping, 3D content production remains stuck in the "industrial era."
In Song's view, the "3D camera" of our era has yet to be invented. What VAST aims to do is first use foundation models to drive production costs close to zero, then enable ordinary people to create and share their own worlds without mastering complex tools.
In this episode, we discuss the real inflection point for AI 3D, and content forms beyond gaming; how a company without OKRs, one that encourages open debate, confronts something that cannot be validated in the short term; and how a long-term-oriented founder maintains patience and judgment amid uncertainty.
Toward the end, he brings up The Man Who Planted Trees: no matter what happens outside, the planter simply goes day after day, burying seeds in the soil, accomplishing what seems impossible.
Today's VAST may still be just a shrub, but he is willing to keep planting until it becomes a forest.
👤 Guest Bio
Yachen Song (Simon), Founder and CEO of VAST.
He studied International Relations and Economics at Johns Hopkins University for his undergraduate degree. He then joined SenseTime, participating in business strategy planning and incubating multiple AI projects from zero to one, pushing AIGC technology into animation and gaming. In 2021, he was an early member in founding MiniMax; in 2023, he founded VAST, setting his entrepreneurial direction on generative 3D.
Currently, he leads his team from 3D foundation models toward world models and interactive content platforms, with the goal of transforming 3D creation from specialized production into a tool accessible to more people.
🕒 Selected Timestamps
03:24 Why does making a AAA game today still feel like producing a high-budget film?
09:16 UGC has almost no expected returns — how do you drive production investment close to zero?
15:03 When everyone can create worlds, what happens to the basic unit of content?
26:05 "Pure people hold the signal": creating a low-noise environment for top talent
29:43 Sometimes not doing is harder than doing: why do big companies' UGC platforms tend to go off track?
33:59 Before consensus arrives, patiently telling a story that cannot yet be proven
39:06 "Not that I accomplish it, but that it is accomplished through me": how he understands the relationship between person and vision
48:40 First see a shape, then use thinking to build the path toward it
57:18 Someone who never wanted to start a company — why did he ultimately choose to become CEO?
01:07:09 The Man Who Planted Trees: the outside world is all noise; the seed in hand is the direction
📚 References
Tripo AI: VAST's core 3D creation product, capable of generating 3D models from text or images. It corresponds to the first step discussed in the episode: transforming expensive, professional 3D production into a capability accessible to more people.
Project Eden: A world model research preview released by VAST AI Research. It is oriented toward multiple people and AI agents sharing a single continuously running world, emphasizing that world states can be remembered and remain consistent across different perspectives.
3D Foundation Model: Through large-scale data training, models gain general capabilities to generate, understand, or process 3D assets. In the episode, 3D foundation models play the role of "solving production efficiency first," laying groundwork for larger-scale interactive content supply.
World Model: The emphasis here is not on continuously generating frames, but on giving a world relatively stable space, states, and rules of change. When a person or agent leaves the field of view, changes that have already occurred still persist — this comes closer to an enterable, interactive world.
Ding Liang: CTO of VAST. B.S., M.S., and Ph.D. from Tsinghua University. Formerly head of general models at SenseTime, with long-term research in artificial intelligence and computer vision.
Yanpei Cao: Chief Scientist of VAST. B.S. and Ph.D. from Tsinghua University's Department of Computer Science. Previously conducted research at Kuaishou Y-tech, Tencent AI Lab, and PCG ARC Lab, with research directions including 3D vision and generative 3D.
Xinyu Mao: CFO of VAST. Previously Executive Director at CICC's Investment Banking Division, with long-term experience in corporate listings and capital markets.
UGC / PGC: UGC refers to user-generated content; PGC refers to professionally generated content. Song argues in the episode that ordinary users creating UGC may not expect returns, so tools must compress time, cost, and professional barriers sufficiently low.
AAA Games: Typically refers to large-scale games with high budgets, production scale, and quality requirements. The episode uses this to illustrate the industrial threshold of traditional 3D content: a single work often requires massive teams and long cycles, making supply naturally limited.
"To advance civilization for the world, to create happiness for humanity": From Li Dazhao's 1916 The Youth, published in New Youth. Written during the New Culture Movement, it connects individual youth to the renewal of family, nation, ethnicity, humanity, and the world.
"Doubting Thomas": From John 20. Thomas, upon hearing of the Resurrection, declared he could not believe unless he saw and touched the wounds himself; "Doubting Thomas" later came to refer to someone who requires evidence before believing.
Consciousness-Only School (Yogācāra): A major school of Mahāyāna Buddhism, centered on analysis of "consciousness." The main Chinese Yogācāra system was established by Xuanzang and his disciple Kuiji during the Tang Dynasty, also commonly known as the Dharma Characteristics School.
Pure Land School: A Mahāyāna Buddhist tradition focused on faith, vow, nianfo (Buddha recitation), and rebirth in Amitābha Buddha's Pure Land. In Chinese Buddhist history, the Pure Land School developed an independent and widespread tradition of belief and practice.
Eight Consciousnesses: The Yogācāra classification of cognitive activities, comprising eye consciousness, ear consciousness, nose consciousness, tongue consciousness, body consciousness, mind consciousness, manas, and ālaya consciousness.
Transformation of Consciousness into Wisdom: A Yogācāra term referring to the conversion of the eight consciousnesses into four wisdoms: the great mirror wisdom, the equality wisdom, the observational wisdom, and the accomplishment wisdom.
MMORPG and SLG Games: MMORPG refers to massively multiplayer online role-playing games; SLG typically refers to strategy games.
The Man Who Planted Trees: A short story by French writer Jean Giono, about a shepherd who planted trees for years, returning a wasteland to forest. Song uses it to describe VAST: not treating every outside change as direction, but simply planting this one tree in front of you.
🎵 Music
Jordan Critz - Beau Et Rapide (Piano)
🎤 Production Team
Host｜Jinjian Zhang
Produced by｜Oasis Capital
Editing & Production｜Shengdu Studio Podcast Workshop
💬 Moment of Engagement
If one day, creating an enterable, interactive 3D world becomes as simple as editing a short video, what would you want to build first? What rules would it follow?
Disclaimer
All investment-related content in this podcast is for exchange and sharing purposes only, for reference, and does not constitute any market prediction, judgment, or investment or advisory recommendation. Thank you for your interest in this podcast's original content! If reposting or quoting content from this podcast, please indicate the source. Please contact Oasis Capital and obtain consent before reposting.
[View episode transcript on Xiaoyuzhou]