Hedge Fund Day 3
Hedge Fund Day 3: 华毅瀛飞今天看华毅瀛飞。
这家公司做的是高速光通信里的核心器件。简单说，就是数据中心、交换机、光模块之间传输大量数据时，需要把电信号转换成光信号，再通过光纤传出去。华毅瀛飞做的就是这条链路里的关键芯片和器件。
我把它理解成一个早期硬科技项目，不是现在就有稳定现金流的公司。核心问题只有一个：
它能不能把技术、客户需求和产线能力真正变成可量产、可交付、可收款的产品？
It Is Inside The Module一开始最容易把这个东西理解错。
它不是 Type-C、Lightning 那种外部接口。那些是你能看到、能插拔、能感知的东西。
华毅瀛飞做的更像是光模块里面的关键零件。外面看到的是接口和线，里面真正决定速度、功耗、距离和成本的是这些光电器件。
大概可以这样理解：
Product
How I understand it
Main use
EML
激光器和调制器集成在一起的高速光源
数据中心高速光模块
DML
更简单、成本更低的激光器
运营商接入网和部分中短距场景
CW DFB
稳定持续发光的光源
硅光模块里的外置光源
T ...
Hedge Fund Day 2
Hedge Fund Day 2: Silicon Phased Array ChipsToday I looked at silicon-based millimeter-wave phased array chips.
Very long name.
Basically:
Use silicon process to make RF chips smaller, cheaper, and more integrated, so phased array systems can be used in satellite terminals, radar, 5G/6G, cars, and industrial sensing.
At first I thought this was just a chip story.
But the more I read, the more it felt like a supply chain story.
The important part is not only:
Can the company make a good c ...
Hedge Fund Day 1
Hedge Fund Day 1: 算力调度不是 GPU 租赁今天准备看共绩科技。
一开始我以为这就是一家 GPU rental company。
后来发现这个判断有点偷懒。说它是 GPU 租赁，不能算错，但也像把滴滴说成“汽车租赁公司”。表面上好像对了，实际上把最重要的东西漏掉了。
共绩真正想做的，应该更接近一个 AI 时代的算力滴滴：把分散的、闲置的、波动的 GPU 资源，调度给同样分散、同样波动的 AI 需求。它们官网写的是弹性算力服务平台，融资新闻里更直接，把这套东西叫成“电网式算力调度网络”。
这个比喻也不能乱用。因为只要用了“滴滴”两个字，听起来就很容易变成创业 BP：
我们是 XX 领域的 Uber。
然后投资人开始微笑，创业者开始冒汗。
所以今天我不想先下判断。我更想记录一下：如果我要认真看这家公司，哪些问题是真的问题。
1. 不能只问有多少卡看算力公司，最容易问的问题是：
你们有多少 GPU？
这个问题当然重要，但它太粗了。
如果一家公司的故事是“我们有很多卡”，那它很快就会进入一个很朴素的竞争世界：谁卡多，谁便宜，谁交付快。这个世界也不是不好，只是听起来不太 ...
Post-CV: When the Pig Gets Overconfident (AKA Bloated)
I Am BloatedAfter finishing the Jenga stacking project, I became bloated.
Not physically. Mentally. The dangerous kind.
Jenga stacking semi-automated with OpenCV, returning best results with human authorization for grabbing.
.video-card {
width: min(100%, 360px);
margin: 24px auto;
}
.video-card__frame {
aspect-ratio: 9 / 16;
overflow: hidden;
border-radius: 18px;
background: #111;
box-shadow: 0 10px 28px rgb(0 0 0 / 16%);
}
.vide ...
The Origins of an Entrepreneurship Column and the Shifting Tides of the Semiconductor Industry
DaTong ChenA few years back (2012), I was invited to write a yearlong entrepreneurship column for the Tsinghua alumni magazine Shuimu Tsinghua. Several installments recounted my experiences founding OmniVision Technologies in Silicon Valley and SpreadTrum Communications in Shanghai — a sort of summary of the first half of my life. But who could have guessed that in recent years, both companies would be acquired, delisted from the U.S. markets, and brought back to the motherland to embark on enti ...
When a Pig Started to Learn OpenCV...
BackgroundI have a goal of stacking Jenga blocks. But before the robot arm can do anything fancy, I need to stop being scared of images.
Short GoalSo I started with this static photo of Jenga blocks from my desktop camera. The small mission: draw accurate rotated boxes, center dots, and labels around the blocks by the end of this post.
The starting photo: five blocks, 640 x 480 pixels, shot with an Intel RealSense D435. The lighting is not perfect, the shadow is dramatic, and OpenCV i ...
Edelweiss with Frequencies
Click to see powerpointSin Frequencies Precal
Open notebook in Google Colab
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172import numpy as npimport matplotlib.pyplot as pltfrom IPython.display import AudioSAMPLE_RATE = 44100def make_note(frequency, duration, volume=1.0): t = np.linspace(0, duration, int(SAMPLE_RATE * duration), endpoint=False) wave = volume * np.sin(2 * np.pi * frequency * t) fa ...
Robotic Arm Dos
One year later…Last year was my first encounter with the robotic arm. I managed to understand what it is and how to set it up. Sounds superficial? Indeed!
This spring semester I decided to face the robotic arm again, with a bit more patience and a bit more courage. I felt confident that at least I could learn a bit more on the software side.
Goals:The GRAND goal was to stack Jenga blocks with closed-loop visual control rather than scripted coordinates.
To achieve this, I gave myself some Sub-Goa ...
What should I do?
My Exploration of Tech:I started exploring tech at the beginning of the summer before freshman year. Before that, most of what I cared about was economics, politics, food, and tennis. So honestly, it was a pretty late start.
My real, almost logarithmic boost into technology came in the fall of my sophomore year. Through Tech and Innovation I, I found a class that gave me much more freedom to access and experiment with different areas of technology.
A lot of times, I would watch YouTube videos an ...
Post-The Great Gatsby
Open full presentation