# Contemplation Over Yuanming Hu's Insight

## 0. Post Details

- Reference:
- [Physics-based fluid simulation in computer graphics: Survey, research trends, and challenges, Computational Visual Media, 2024.](https://link.springer.com/article/10.1007/s41095-023-0368-y)
- 胡渊鸣：Meshy AI，太极，MIT，清华姚班，图形学，物理仿真模拟，开源，商业化，勇气 ，智慧 ｜ WhynotTV Podcast #2, YouTube, 2025.
{%youtube 6EyUOLGsuk4 %}
- Post by: Jedd Yang
- Date: 2025-08-22
- Keywords: Computer Graphics, Computational Mechanics

## 1. Main

Computer graphics spans geometry, rendering, and simulation. Within simulation, I’ve often noticed computational mechanics work appearing in graphics journals, as I observed in this [paper](https://hackmd.io/nLQn-45cSdmxho671nMglA). That makes me wonder about our stance—the computational mechanics workers and enthusiasts—within simulation, especially under the sim-to-real context.

In a recent podcast hosted by Tairan He from WhynotTV, Yuanming Hu, inventor of [Taichi Lang](https://www.taichi-lang.org/), CEO of [MeshyAI](https://www.meshy.ai/), discusses how the advancement in simulation may have stalled. He reasoned that traditional numerical methods like the Galerkin method might not be the most accurate description of the world, due to its oversimplification of boundary conditions from the real-life complexity. This can limit the validity of synthetic data for sim-to-real transfer, even though such methods still dominate in areas like defense or manufacturing.

By contrast, we’re seeing data-driven methods—either through computer vision or neural networks—gradually taking over the simulation domain. And prospectively, the real-data-driven method might overcome the disadvantage of the model-based method he mentioned above.

To us, the traditional mesh-based/mesh-free methods developers, that is a saddening market-driven result, reminiscent of what happened in control theory when some declared “Control is dead” with the rise of VLMs and VLAs.

:::success
Me and my Master’s advisor, Prof. Tsung-Hui Huang, had a talk over this topic, and we came to a conclusion: Many disciplines have been considered dead for years, yet their spirit lives on.
:::

Whether simulation is dead or simply in transition is open for debate. But Hu’s point is clear: we shouldn’t confine ourselves to a prebuilt framework. New pathways are emerging, as the world spins swiftly on.