Yupeng Cao
871 posts

Yupeng Cao
@cyp0630
Multimodal, NLP ; Crypto


This is the kind of market where you simply buy what the president tells you to: Chips $INTC Intel $NVDA NVIDIA $AMD AMD $TSM TSMC AI Infrastructure $DELL Dell $VRT Vertiv $NBIS Nebius $CRWV Coreweave $IREN IREN $HUT Hut 8 Rare Earths $USAR USA Rare Earth $CRML Critical Metals $TMC The Metals Company Critical Minerals $TMQ Trilogy Metals $UAMY United States Antimony $MP MP Materials $LAC Lithium Americas Quantum $IONQ IonQ $QBTS D-Wave Quantum $RGTI Rigetti Energy $BE Bloom Energy $GEV GE Vernova $FCEL Fuelcell $PLUG PlugPower $VICR Vicor Memory $SNDK Sandisk $MU Micron $STX Seagate $WDC Western Digital $P Everpure $MRAM Everspin Space $RKLB Rocket Lab $ASTS AST SpaceMobile $PL Planet Labs $FLY Firefly Aerospace $GILT Gilat Drones $AVEX Aevex $ONDS Ondas $UMAC Unusual Machines Nuclear $XE X-energy $LEU Centrus Energy $CCJ Cameco $OKLO Oklo $UUUU Energy Fuels AVs & Robotics $TSLA Tesla $AMZN Amazon Batteries $TE T1 Energy $ELVA Electrovaya $FLNC Fluence Energy Never miss the leading stocks again: fullstackinvestor.co/portfolio

















Anthropic pays $750,000+ a year for engineers who can build LLM architectures from scratch. Stanford taught the entire thing in 1 hour lecture & released it for free. Bookmark & watch this today before someone takes it down.


做教育App的注意了,这个东西可能要颠覆你的技术选型 一直以来移动端渲染数学公式这件事,WebView就是那根永远绕不过去的刺。你加载一条公式,背后跟着的是整套浏览器内核,内存哗哗地吃,启动龟速,用户体验烂得你自己都不好意思说。 然后我看到了 RaTeX。 这玩意儿用纯 Rust 从零写的公式渲染引擎,直接把 JavaScript 和 WebView 整个从链路里踢飞,不是优化,是直接干掉。 一套核心代码能覆盖七个平台: ① iOS ② Android ③ Flutter ④ React Native ⑤ Web ⑥ PNG ⑦ SVG 全原生渲染,零 JS 依赖,完全离线跑,不需要联网加载任何东西。 功能上也没偷懒: ① 语法对 KaTeX 兼容度约 99%,你原来的公式基本照搬 ② 分数、根号、积分、矩阵这些常用表达都稳 ③ 顺带支持化学方程式和物理单位书写 ④ 理工科场景基本全覆盖 说白了就是:又快、又轻、又能跨端、还不依赖任何 Web 技术栈。 你做教育、学术类 App,现在还在用 WebView 那套又重又慢的方案,纯粹是在给自己挖坑。 这个值得认真看一眼,不是那种PPT项目。 🔗 github.com/erweixin/RaTeX


JP Morgan's investment research team just shared exactly how they built their multi-agent system "Ask David", and it's the same architecture pattern showing up everywhere: - supervisor agent orchestrates - specialized subagents handle retrieval, structured data, analytics - LLM-as-judge reflection node before the answer ships - human-in-the-loop for the last accuracy gap worth watching for anyone building:










