Beicheng Lou

research @flappyairplanes. prev @AnthropicAI @jumptrading. @Stanford alum.

[Beicheng Lou @lou_beicheng]

@PandaAshwinee 🫶 bamboo would work too

[Beicheng Lou @lou_beicheng]

@dannytse_ @AnthropicAI @flappyairplanes 🙇🙇

[Beicheng Lou @lou_beicheng]

I left @AnthropicAI and joined @flappyairplanes

[Flapping Airplanes @flappyairplanes]

We’ve spent the past year developing new primitives for GPUs. We believe this enables new algorithms to change how machines learn, open new markets, and dramatically reshape AI’s cost structure.

We have raised a $1 billion Series A, co-led by Index and Lux.

[Beicheng Lou @lou_beicheng]

@SallyHZhu @flappyairplanes Impossible without you!

[Beicheng Lou @lou_beicheng]

@kushal1t So much fun!

[Beicheng Lou @lou_beicheng]

@jenniferyxlin Impossible without you!

[Beicheng Lou @lou_beicheng]

@cerise_ima 🧙‍♂️

[Beicheng Lou @lou_beicheng]

@HominLuo 🫶

[Beicheng Lou @lou_beicheng]

@aidanmantine @AnthropicAI @flappyairplanes ❗️❗️

[Flapping Airplanes @flappyairplanes]

We’ve developed the world’s most complete megakernel framework. It can perform entire training runs, serve inference, play video games, and even do all three simultaneously in a single kernel.

[Beicheng Lou @lou_beicheng]

@spectorb 👋

[Beicheng Lou @lou_beicheng]

@christinexye 🫶

[Beicheng Lou @lou_beicheng]

@flappyairplanes Fasten seatbelt. Expect turbulence.

[Beicheng Lou @lou_beicheng]

Data efficiency is the key! We have what it takes

[Beicheng Lou @lou_beicheng]

This is my last week at @AnthropicAI.
It's been a true privilege, and an exciting journey full of wonder, inspiration, and cherished friendships.
Moving forward, I'm excited to explore new paradigms in a leaner setting.
What is the nature of intelligence? What is the missing piece behind continual learning, automated research, and recursive self-improvement?
Always happy to grab coffee with another curious mind!

[Beicheng Lou @lou_beicheng]

@JoeJBenton 🫡

[Beicheng Lou @lou_beicheng]

Scientific discovery is about efficient probing of the nature oracle for generalizable knowledge. We engineer better tools to be more efficient, and build higher abstractions to be more generalizable. It requires residualizing against what we already knew, and is therefore naturally a recursively self-improving process. Exciting time to see how similar ideas evolve and diverge!
P.S. isn’t Möbius strip a better icon for the “loop” 🤯 @DiscoLoopAI

[Discovery Loop @DiscoLoopAI]

Introducing Discovery Loop ♾. We’re building AI to run experiments at unprecedented scale and to solve the biggest bottlenecks in science and engineering.

Our founders @JeffDean, @Sanjay_Ghemawat, @quocleix, @OriolVinyalsML helped build the AI infrastructure and models that scaled the modern world. Now, we are scaling discovery itself.

The future can't wait.

Learn more at: https://t.co/6Fr8dX5dDB

[Beicheng Lou @lou_beicheng]

This is my last week at @AnthropicAI.
It's been a true privilege, and an exciting journey full of wonder, inspiration, and cherished friendships.
Moving forward, I'm excited to explore new paradigms in a leaner setting.
What is the nature of intelligence? What is the missing piece behind continual learning, automated research, and recursive self-improvement?
Always happy to grab coffee with another curious mind!

[Beicheng Lou @lou_beicheng]

Beautiful and unique!

[Claude @claudeai]

On several coding and knowledge work evaluations, Opus 5 is the new state-of-the-art: