Forward deployed investor. What I learned shipping at Google, Snap, and my own company — about the gap between tech and business.
Palo Alto, CA
Joined May 2009
- Tweets2K
- Following363
- Followers375
- Likes1.4K
Congxing Cai retweeted
1/ Today, we’re introducing Geodesic Intelligence, and launching NovaDDE and NovaAtom-Lite-Preview.
Geodesic is building an AI-native platform for protein therapeutics, building AGI for drug discovery to find the shortest path from biology to medicines.
geodesiclab.com
Congxing Cai retweeted
Sound Effects 1.0 from @Sonilo_music is now available on fal.
Generate realistic sound effects from video or text.
Upload a video to get a fully synced audio track, or describe the exact sound you need with a text prompt. Generated in seconds.
RT @luoj: Most friends and colleagues who've known me for less than 20 years don't know this: I studied Electrical Engineering in undergrad…
This quoted post is unavailable.
Congxing Cai retweeted
We tend to systematically ignore the role of violence, the absolutely central role of war and slavery in creating and shaping the basic institutions of what we now call “the economy”.
- David Graeber
Debt: The first five thousand years
Congxing Cai retweeted
AI job displacement is real. The owners cutting headcount aren't bluffing — they genuinely can do more with fewer people now.
But "just learn AI tools" as career advice has the same problem as "just invest wisely" as financial advice. Technically true. Practically useless for the people who need it most.
The interesting question isn't whether new jobs appear. They will. It's the transition function — how much pain happens between the old equilibrium and the new one, and who bears it. Historically, the answer is: the people with the least power bear the most cost, and the transition takes longer than the optimists predict.
Congxing Cai retweeted
“We were going to run out of money.”
Groq was 3 weeks away from death and @JonathanRoss321’s leadership team put together a list of layoffs that also would have killed the company.
He realized that the only way to survive was to convince their employees to take a salary cut. The wild story of Groq Bonds:
“We had an all-hands and put up World War II-looking pictures of war bonds. We called it ‘Groq Bonds.’”
80% of Groq’s employees participated in exchanging their salary for equity.
Half of them went to the statutory minimum salary by law.
“Engineers get paid hundreds of thousands of dollars. These folks cut their salary down to $50,000 to $60,000.”
“By doing this, we put everyone's hand on the steering wheel. They were participating in saving our runway and we had less than 10% attrition. It might've been closer to 5%, which was probably better than our attrition rate before then.”
A conversation with @JonathanRoss321, founder of @GroqInc. This is the first podcast that Jonathan talks about his $20 billion partnership with NVIDIA.
In this conversation, we talk about his decade building Groq: why West Coast VCs kept passing on the company, how fast Jensen works, overcoming near-death experiences, and why he thinks the entire AI age turns on a single skill — knowing what to ask.
Hope you enjoy this conversation:
0:00 The $20 Billion NVIDIA Deal Closed In 3 Weeks
0:25 Why GPUs And LPUs Are Better Together
1:46 When AI Talks To AI, Speed Wins
3:30 Always Start With A Hobby Project
5:55 Ask The Right Questions, Not Answer Them
8:23 There Are Infinite Ways To Be A Leader
13:00 I Was One Of The World's Worst Leaders
14:34 Fewer Constraints, More Room To Surprise You
16:31 At NVIDIA There Is No Politics
19:44 You Have To Learn Confidence
22:23 East Coast VCs Think, West Coast VCs Follow
23:50 The Keynesian Beauty Contest Of Silicon Valley
26:48 The Autonomy That Created The NVIDIA Deal
30:07 Making A Model Smarter By Making It Faster
34:52 Reality Quotient Beats Intelligence Quotient
35:44 Find The Dominant Game And Play It
37:11 A Founder's Job Is Full-Time Change Management
38:34 Return On Luck: Seize It Better Than Anyone
42:54 You Can't Sell Speed, You Have To Let People Try It
46:32 I Intend To: Intentional Leadership
51:07 Groq Bonds: Trading Salary For Survival
54:13 Hire For Negatives, Grow For Positives
58:46 Loss Aversion And Booking The Win Early
1:00:37 How Michael Jordan Weaponized Humiliation
1:03:13 Manufactured Discontent Drives Everything
1:05:02 Every Day Without Compute Has A Real Cost
1:07:07 Code Was Rationed, Now It's Nearly Free
1:10:04 Teach Kids To Ask Questions, Not Answer Them
Includes paid partnerships.
Congxing Cai retweeted
Ex-Berkeley PhD who leads SGLang at xAI explained how they serve Grok on 100K GPUs in 23 minutes - better than $2000 inference-at-scale courses.
split prefill and decode -> shard experts across GPUs -> route tokens per expert -> overlap comm and compute -> serve at DeepSeek-API-killing prices.
That loop is why xAI runs Grok on SGLang and third parties beat DeepSeek's own API by 5x on cost.
SGLang + prefill-decode disaggregation + expert parallelism + AMD MI300 - that's the stack.
Watch and save it, then read the article below.
Congxing Cai retweeted
Happy 250th birthday, America! We got you a present. 🇺🇸
The red, white, and blue stars of this globular cluster shine like a sparkler waved on a dark night in this image from @NASAHubble, released in celebration of the United States' 250th anniversary.
ALT A collection of red, white, and blue stars sparkling in the darkness of space. The stars are most concentrated in the center of the image. There is more space between stars as you move to the edges of the picture. Credit: NASA, ESA, and A. Dotter (Dartmouth College); Processing: Gladys Kober (NASA/Catholic University of America)
Congxing Cai retweeted
GLM 5.2 just became the first open-source model to lead a category on APEX-SWE.
It scored a 55.3% Pass@1 on Integration, the top score we've recorded for any model, open or closed source.
On the overall leaderboard, GLM 5.2 scored 37.3% Pass@1, ranking 6th place. That makes it the best open-source model we've tested on APEX-SWE to date.
Right behind it is Kimi K2.7 from Moonshot AI, now the second-best open-source model on the APEX-SWE leaderboard.
Congrats to @Zai_org and @Kimi_Moonshot on two strong model releases.
ALT APEX-SWE | GLM-5.2
Congxing Cai retweeted
Dylan grew up working in motels
Got a screen of death on his XBOX as a kid
Fixing it got him into semis
Somehow this led him to starting SemiAnalysis, the premier semis research co
@dylan522p is fascinating and on an all time run
👀 @sonyatweetybird
Congxing Cai retweeted
If I could pick "grown up" concepts I'd like kids to learn about earlier, counter-intuitiveness would be high on the list. One of the problems with teaching things by rote is that you also implicitly teach that the truth is obvious, when in fact it's often surprising.
Congxing Cai retweeted
As engineering, product, design, DS, etc. melt into a new kind of role, I was reflecting on what roles might look like in the future. For example, when I look at the Claude Code team I see what I think is five archetypes:
1. Prototyper: comes up with brand new ideas; churns out many ideas, most of which don't ship
2. Builder: quickly turns a prototype/idea into production-grade product/infra
3. Sweeper: cleans up the UI, simplifies the code and system, unships, optimizes performance
4. Grower: takes a product that has been built and iterates on it to improve Product-Market Fit
5. Maintainer: owns a mature system to make it secure, reliable, fast, and efficient as it scales
Many people span across 2 roles, and sometimes 3 roles. I also notice that these roles are not really tied to job function -- eg. across Anthropic, some designers match category 1, some 2, some 3; same for engineers, PM, DS.
A healthy team needs a mix of these, depending on the product:
- A product that is new and pre-PMF needs people that are strong at 1+2+3
- A product that is growing and has found PMF needs 2+3+4 and some 5
- A product that has strong PMF needs 3+4+5 and some 2
Maybe product roles of the future will look more like this, and less like the domain-specific roles of today?
Congxing Cai retweeted
Moving bits to and from memory.
Because of parasitic capacitance+resistance of the wires.
The bigger the memory, the longer the wires.
The main trick is to organize the memory hierarchically: registers, small on-chip SRAM, caches of various types, and external RAM.
It's all because we have to use hardware multiplexing: reusing the same multiply-accumulate unit for multiple parts of the network.
