Tangos Chan

Partner at China Growth Capital, ex-blogger in Beijing, high-tech, fintech, internet.

[Tangos Chan @Tangos]

非常认同 Jim 的观点，非常期待基于世界模型的 Groot，也希望业界和学界能够一起协作加快推动benchmarking 的问题

[Jim Fan @DrJimFan]

Everyone's freaking out about vibe coding. In the holiday spirit, allow me to share my anxiety on the wild west of robotics. 3 lessons I learned in 2025.

1. Hardware is ahead of software, but hardware reliability severely limits software iteration speed.

We've seen exquisite engineering arts like Optimus, e-Atlas, Figure, Neo, G1, etc. Our best AI has not squeezed all the juice out of these frontier hardware. The body is more capable than what the brain can command. Yet babysitting these robots demands an entire operation team. Unlike humans, robots don't heal from bruises. Overheating, broken motors, bizarre firmware issues haunt us daily. Mistakes are irreversible and unforgiving.

My patience was the only thing that scaled.

2. Benchmarking is still an epic disaster in robotics.

LLM normies thought MMLU & SWE-Bench are common sense. Hold your 🍺 for robotics. No one agrees on anything: hardware platform, task definition, scoring rubrics, simulator, or real world setups. Everyone is SOTA, by definition, on the benchmark they define on the fly for each news announcement. Everyone cherry-picks the nicest looking demo out of 100 retries.

We gotta do better as a field in 2026 and stop treating reproducibility and scientific discipline as second-class citizens.

3. VLM-based VLA feels wrong.

VLA stands for "vision-language-action" model and has been the dominant approach for robot brains. Recipe is simple: take a pretrained VLM checkpoint and graft an action module on top. But if you think about it, VLMs are hyper-optimized to hill-climb benchmarks like visual question answering. This implies two problems: (1) most parameters in VLMs are for language & knowledge, not for physics; (2) visual encoders are actively tuned to *discard* low-level details, because Q&A only requires high-level understanding. But minute details matter a lot for dexterity.

There's no reason for VLA's performance to scale as VLM parameters scale. Pretraining is misaligned. Video world model seems to be a much better pretraining objective for robot policy. I'm betting big on it.

[Tangos Chan @Tangos]

中国的顺差资金花到哪儿了？

[Tangos Chan @Tangos]

如果AI找到的Ripasudil对于治疗干性黄斑病变确实有效的话，那对很多老年人真的是福音了 https://t.co/UTSTdZ726r

[Tangos Chan @Tangos]

https://t.co/matYBAK1QX 这个一水肌酸对阿尔兹海默症的效果研究初步看起来还挺令人振奋的。如果后续研究能证明一水肌酸的效果的话，那既可以用来促进肌肉生长，提高运动表现，还能改善脑活力，健脑又健身，一举两得了

[Tangos Chan @Tangos]

华贸的这家费大厨也是奇怪，开了几年了，周围的商家都换过一两拨了，他们家还是这么多人等位吃饭，其他城市各种类似的“小炒黄牛肉”这些早就已经遍地可见，也没见到有同类商家进驻华贸，是因为业主有意筛选控制了？

[Tangos Chan @Tangos]

Just tried https://t.co/uan4LJQ6jV, it is truly powerful, a great example that AI enhance efficiency and simplify our life.

[Tangos Chan @Tangos]

“treat it like an OS”怎么理解？

[vitrupo @vitrupo]

Sam Altman says ChatGPT is splitting by generation.

Older users treat it like Google. Millennials use it as a life advisor.

“College students treat it like an operating system.”

They've built workflows, memorized prompts, connected files. Now, with memory, it has full context on everyone in their life and everything they've talked about.

[Tangos Chan @Tangos]

测试了一下字节的这个 Deep Research 工具，自己试验的效果感觉一般，不知道是不是有些配置没搞对。不过确实也还有很多需要完善的。 https://t.co/yGO8D31BEd

[Tangos Chan @Tangos]

I'm using the Fellou AI browser to publish this post.

[Tangos Chan @Tangos]

@ayuan1000 领导指令那里我听到也觉得尬了

[Tangos Chan @Tangos]

Exactly, VC should hedge their LLM startups investments by buying NVDA.

[Elad Gil @eladgil]

At this point LPs in venture funds should just wire money direct to NVIDIA & skip the LP->VC->Startup->NVIDIA steps

[Tangos Chan @Tangos]

今年以来因为疫情的确出差少了很多，但以有限的出差经历来看，这样大面积的取消并非疫情下的常态。

[硅谷王川 Chuan @Svwang1]

这个角度有道��

[Tangos Chan @Tangos]

Met an quantum-resistant algorithms startup months ago, an important cybersecurity sector in coming future.

[MIT Technology Review @techreview]

Cryptographic algorithms are what keep us safe online, protecting our privacy and securing the transfer of information. But when quantum computers become powerful enough, they could theoretically crack these algorithms. The race is on to find new ones. https://t.co/1lJvRAKKnY

[Tangos Chan @Tangos]

还有大纲功能，安装试用一下

[Orange AI @oran_ge]

昨天发现一个 Chrome 的插件 Clearly Reader，可以让网页变为阅读模式的同时，生成非常舒服的大纲，比如这篇5000词的文章，有了大纲直接重点都出来了。（当然一些排版垃圾结构混乱的公众号文章肯定生成不了。）

[Tangos Chan @Tangos]

An interesting read.

[Bloomberg @business]

NEW: A Chinese spy tried to steal GE’s secrets. Instead he got caught by the US—and gave the world an unprecedented glimpse into China’s espionage apparatus.

Read The Big Take 🔽
https://t.co/Dmzx4S3byZ

[Peter Yang @petergyang]

The user retention funnel for most products

[Tangos Chan @Tangos]

市场都寄希望二十大，明年春还有十四届人大呢。清零的口径不觉得会改变，只是看在这个“大原则”下给多大的操作空间了。

[外汇交易员 @fxtrader]

高盛首席中国经济学家闪辉等分析师预计，二十大将主要侧重于党内领导层调整，明年上半年人大召开后清零防疫政策才有可能转变。

[Tangos Chan @Tangos]

What will happen when winter comes?

[Alf @MacroAlf]

Almost the entire European continent is now operating at electricity prices above EUR 600/MWh

This is roughly equivalent to $1000 (!) per barrel of oil

The last decade average cost of electricity was in the EUR 20-30/MWh range

Not sure a few hikes are going to fix the EUR here

[Tangos Chan @Tangos]

Creative Commons is turning 20 years old.

[Lessig 🇺🇦 @lessig]

@CreativeCommons was born 20 years ago. "Come" to the global summit next week to celebrate.