yi | $XNY
@drtwo101
Followers
3K
Following
3K
Media
257
Statuses
3K
Building @codatta_io which transforms foundational models into vertical AI solutions. prev: ex-(Pinterest, Alipay), PhD in ML.
San Mateo, CA, USA
Joined May 2009
Appreciate @MessariCrypto recognizes @codatta_io positioning with deAI data stacks. fun fact: Codatta actually evolves from OSS Microscope Protocol, which we started with @coinbase @MessariCrypto and @GoPlusSecurity
docs.codatta.io
0
4
18
Claude Code 正在把“写代码”升级成“通用 AI Agent”。 这条入门视频系统讲清楚了：什么是 coding agent、如何用 Claude Skills（Markdown）构建可复用能力、以及把 AI 直接接入真实业务流程（比如自动发 X）。 核心不是语法，而是：如何把 Claude 变成你的数字员工。 https://t.co/9ClN9yNQ7l
0
1
1
Introducing A2UI: Agent-to-User Interface 🛜Protocol for agent-driven interfaces 🤖Enables agents to generate interactive user interfaces 🐙Open source https://t.co/kGs3tOXtzY
44
129
1K
cool paper studying the differences between SFT and DPO.
arxiv.org
Learning dynamics, which describes how the learning of specific training examples influences the model's predictions on other examples, gives us a powerful tool for understanding the behavior of...
7
87
800
This paper from Stanford and Harvard explains why most “agentic AI” systems feel impressive in demos and then completely fall apart in real use. The core argument is simple and uncomfortable: agents don’t fail because they lack intelligence. They fail because they don’t adapt.
186
636
3K
You’ve maybe heard from me on this topic too many times, but this is the last I’ll offer (at least for now). My worry isn’t the code or the tools themselves. The question is how we keep thoughtful design alive even as new tools and technologies emerge. https://t.co/MPPY1KfQFK
linear.app
20
48
724
Performance Hints Over the years, my colleague Sanjay Ghemawat and I have done a fair bit of diving into performance tuning of various pieces of code. We wrote an internal Performance Hints document a couple of years ago as a way of identifying some general principles and we've
99
1K
8K
Open secret: Frontier generalist models sound cool but specialized models are the workhorse of the industry. The product below: - 0.6b for speech-to-text - 0.1b for speakers - 0.6b for custom vocabulary 3 specialized models totaling 1.3b parameters running in real-time on
Introducing Real-time Transcription with Speakers! - Step change in accuracy, surpassing top cloud APIs - Faster than real-time on Mac and iPhone - Still under 3 watts when all features are enabled Available in Argmax SDK 2.0 for early access! Benchmarks and details in comments.
13
38
560
Impressive
Robots in China are doing it all now, even dancing on stage like pros. Here Unitree robots doing Webster flips and are performing at Chinese-American singer Wang Leehom’s concert in Chengdu. https://t.co/2BNWdok0bf
9K
25K
249K
🎨 Qwen-Image-Layered is LIVE — native image decomposition, fully open-sourced! ✨ Why it stands out ✅ Photoshop-grade layering Physically isolated RGBA layers with true native editability ✅ Prompt-controlled structure Explicitly specify 3–10 layers — from coarse layouts to
149
940
6K
Messari Analysts Holdings 👀 Takeaways from "what the kids are buying these days" - Majors (BTC, ETH, SOL) - Hype - long tail defi (ENA, ZEX, among others) - Narratives (ownership coings, privacy, AI) Summarizing their 2026 convictions: - Crypto equities (HOOD, COIN, GLXY,
We are so back! The Messari Theses for 2026 is live and available for free. Jump into the full report now ⬇️
48
20
268
Bill Gates and Sergey Brin among newly released Epstein photos
ft.com
US congressional Democrats publish another batch of images from late sex offender’s estate
449
2K
11K
Vision-Language-Action (VLA) models struggle with physical understanding because they rely on static images. This paper argues robots need temporal priors. Video models naturally encode motion, dynamics, and cause-effect - critical for real control.
6
55
345
Machine learning is accelerating in robotics, but hardware is still the bottleneck most teams don't talk about. In our new series, Robot Learning in Industry, we sit down with the team at @trossenrobotics, the hardware innovators behind the manipulators used in Stanford's viral
6
20
110
We’ve pushed out the Pareto frontier of efficiency vs. intelligence again. With Gemini 3 Flash ⚡️, we are seeing reasoning capabilities previously reserved for our largest models, now running at Flash-level latency. This opens up entirely new categories of near real-time
51
195
2K