Check out Differentiated Understanding, the podcast by AI Proem — a 360 view of China's internet and tech space. @BrianWongOPR @ruima @freefader @dianawudavid @jasminewsun @jingyanghk

Dan Wang (@danwwang) is one of the sharpest thinkers on China and its relationship with the US. Dan spent six years living in China studying its manufacturing and technology ecosystem, and his new book Breakneck is the best framework I’ve seen for understanding US-China

This video makes it much clearer that the Neo humanoid robot is not autonomous but almost 100% teleoperated and is basically collecting training data from you. No person who actually does chores saw the marketing video and believed it was useful anyway. It was so slow and did

Yann LeCun: "The big secret of the [humanoid robot] industry is that none of those companies has any idea how to make those robots smart enough to be useful." He says that while they're fine for specific tasks, "a bunch of breakthroughs" are needed for domestic use.

Why browsers are so important. - We created an AI god and trapped it in a text box - The AI god is used mostly on desktop v mobile - Desktop (not Macbook) = browser where work is done - Browser becomes the main battleground where normal people will access AI

Highly highly recommend this profile of @JoshuaKushner by @joincolossus It is written really well and the lessons are timeless. From Rae to Rae there are true generational lessons of resilience and being a mensch. https://t.co/o17aaJ3GNf

Did you know China’s latest overcapacity isn’t in EVs or steel — it’s in AI PhDs? I'm only kind of joking ... @jjding99 shared a story in his ChinAI newsletter about this phenomenon. It follows PhD graduates from pretty-good-but-not-elite Chinese universities who are

Sharing my piece in @ForeignAffairs where I argue that, unlike the common perception, China’s overcapacity problem is NOT primarily about government subsidies (many of which have been rolled back) or lack of domestic demand (true in general, but not the leading cause in green

I used to think model APIs are pure commodities. In reality, stickiness creeps in at the workflow level. Sure, you can swap the model for a single step, but building evals and re-verifying outcomes across many steps takes time and money.

OpenAI is trying the same pattern in an AI context: ChatGPT as the front door, verified identity and permissions as the trust layer, an agent runtime for real actions (not just chat), and distribution that lives where users already are.

WeChat became an operating surface by pairing identity and payments with an in-app runtime; Mini Programs turned social + discovery + checkout into one flow and created a #flywheel for developers and brands.

TL;DR OpenAI’s new “apps within ChatGPT” clicked for me because it mirrors how Tencent WeChat made Mini Programs the easiest way to get things done: one trusted surface captures intent, then routes you to lightweight capabilities that live inside it. @oliviaplotnick

Hey @benthompson heard that you also made this analogy. https://t.co/9kOTDXYS5G

Apple:

Spotify:

Substack:

YouTube:

Goal: piecing together clear context, nuanced understandings, rigorous ideas, and real-world impact.

This intro gives you a quick sense of who I am and why I’m bringing together academics, investors, policymakers, entrepreneurs, and journalists to discuss the most significant shifts in tech and AI across APAC.

Had a great time discussing China's AI+ initiative and the big-picture context of economic sea change with @Gracemzshao! Check it at her must-read substack AI Proem, YouTube, or Spotify https://t.co/Yd0AfXESE7 More links....