James Lin

Vice director, HPC Center of Shanghai Jiao Tong University; Chair, Technical Committee HPC of Shanghai Computer Society (SCS TCHPC)

[Torsten Hoefler 🇨🇭 @thoefler]

Mateo Valero discusses history of vector architectures: Back to the Future in #HPC and #AI.

He also mentions the three quantum computers connected to an NVIDIA Blackwell system in BSC's chapel ⛪️.

Ending his talk looking for a future site 😀.

[Torsten Hoefler 🇨🇭 @thoefler]

Yutong Lu reveals the LineShine exascale architecture for #HPC - #AI convergence: 2 exaflop sustained fp64 at 40 MW with 1.34 million ARM CPUs using SME and 4TB/s HBM connected with 800G.

From "offline to online acceleration with SME".

[Torsten Hoefler 🇨🇭 @thoefler]

Taisuke Boku follows on another interesting note: Japan is "changing lanes" from CPU to GPU and from "Made in Japan" to a collaborative "Made with Japan", partnering Fujitsu with NVIDIA.

"We don't chase fp64 performance but focus on #AI flops".

Onward!

[Torsten Hoefler 🇨🇭 @thoefler]

Jack Dongarra's opening keynote at HACI 2026!

#HPC in transition - the investments in compute infrastructure are unprecedented but focus mostly on #AI workloads. Scientific numerical simulations need to adjust!

Many interesting challenges ahead.

[陈成 @chenchengpro]

Claude Code 泄露了全部源码——不是被黑客攻破，是 Anthropic 自己把 source map 打包进了 npm 发布物。

一个 57MB 的 https://t.co/rrsLUlL2Xp 文件，里面藏着 4756 个源文件的完整内容。其中 1906 个是 Claude Code 自身的 TypeScript/TSX 源码，剩下 2850 个是 node_modules 依赖。

提取方法极其简单：https://t.co/rrsLUlL2Xp 本质就是一个 JSON，里面有两个关键数组——sources（文件路径）和 sourcesContent（对应的完整源码）。两者索引一一对应。不需要反编译，不需要反混淆，sourcesContent 里存的就是一字不差的原始代码。提取脚本见文末。

从还原的源码可以看到：Claude Code 用 React + Ink 构建 CLI 界面，核心是一个 REPL 循环，支持自然语言输入和 slash 命令，底层通过工具系统与 LLM API 交互。架构设计、系统提示词、工具调用逻辑，全部一览无余。

这件事的本质是一个经典的安全疏忽：source map 是开发调试用的，包含从变量名到注释的所有信息，不应该出现在生产发布物中。Anthropic 后来意识到了这个问题，移除了 source map，GitHub 上提取源码的仓库也被 DMCA 了。但早期版本的 npm 包已经被存档，源码早就在社区流传。

给所有发布 npm 包的开发者提个醒：发布前检查你的 .map 文件。一行 sourcesContent 就能让你的所有代码公之于众。

https://t.co/pR6792kVsN

[WiseInvest @WiseInvest513]

有点强啊，已经研究了两小时了，补充了很多缺失的知识点。

可以说是我目前看到最全的 OpenClaw 学习教程！

中推的博主有一个非常好的点，那就是大家做的内容真的非常优质，因为只有真正优质的内容才可以跑出来。

相比于国内一些标题党，如果遇到要学习的内容，问一下 Grok，给你推荐一些内容，是真的有学习的价值！

推荐收藏！

[Torsten Hoefler 🇨🇭 @thoefler]

"Shock! Shock!" is how the "Father of the Analysis of Algorithms" Don Knuth described an interaction with Claude, proving a problem he worked on for weeks

#Supercomputing enables the unthinkable and begins to automate the scientific method and progress

https://t.co/8GNrhLqn4o

[HPCwire @HPCwire]

Nvidia faces criticism from the HPC community over flat FP64 gains as GPUs pivot toward AI. While experts like Jack Dongarra note limited 64-bit progress, Nvidia says FP64 isn’t abandoned, and future GPUs will boost it.

Read more: https://t.co/yXFp9Rsv6g

[OGAWA, Tadashi @ogawa_tter]

"Government, University, & Industry Cooperation: The NVIDIA Story", Bill Dally, NSB Meeting, Jul 23, 2025 https://t.co/ofFQg9EvSo
https://t.co/vrsOFb6UHa

D. Kirk, SC19 https://t.co/kXot3oK9w2
IAA WS 2008 https://t.co/6Qx9iW9ve2

<= Benjamin Franklin Medal https://t.co/ktRP6wBuyF

[James Lin @JamesLinHPC]

Surprise but happy to meet @thoefler in Shenzhen during the #CGC25. Next time will be in SCA/HPC Asia chaired by @ProfMatsuoka in Oasaka next January.

[Satoshi Matsuoka @ProfMatsuoka]

Registraion for the SC/HPC Asia 2026 @ Osaka is open. Register now! https://t.co/DVBLZg577p It will become the biggest HPC/AI/Quantum research&industry event in Asia, at the scale of EU ISC, with over 100 global exhibitors, 36 papers/101 submitted, 28 workshops, 20 tutorials, over 20 BoFs, posters, multiple invited sessions, and more.

[Satoshi Matsuoka @ProfMatsuoka]

I will be hosting the last panel of the “Global Commons Forum” https://t.co/kWuWAOKYPq along with my friends Hiroaki Kitano @ SONY-OIST and Thomas Schulthess @ ETH-CSCS etc. to be held at U-Tokyo. We will discuss the plus&minuses of modern HPC & AI towards global sustainability.

[OGAWA, Tadashi @ogawa_tter]

=>
HUAWEI CONNECT 2025, Sep 18
徐直军 https://t.co/DHe2oijNcd
Kunpeng 920, 2024 Q1
64C, 80C/160T
Kunpeng 950, 2026 Q4
96C/192T, 192C/384T
Kunpeng 960, 2028 Q1
High-performance model: 96C/192T
High-density model: ≥256C/512T

HBM?
Apr 25 https://t.co/5cRl1Xe8pd

[OGAWA, Tadashi @ogawa_tter]

"Architectural Implications of Advanced Computing Sanctions", August Ning, Princeton, UCF Seminar, Feb 6, 2025 https://t.co/SXWHnVUXh8
=>
"Chip Architectures Under Advanced Computing Sanctions", August Ning & D. Wentzlaff, ISCA 2025 https://t.co/c7eKSRB8kh
https://t.co/zXm8KGMfKO

[OGAWA, Tadashi @ogawa_tter]

@ProfMatsuoka =>
"RIKEN, Japan’s Leading Science Institute, Taps Fujitsu and NVIDIA for Next Flagship Supercomputer (FugakuNEXT)", Aug 22, 2025 https://t.co/Uy4gYGyA7g

"FugakuNEXT" System Architecture Overview, @ProfMatsuoka https://t.co/QDGAfPNy23
Ozaki Scheme
NVIDIA, Jul 1, K. Ozaki, Jul 2

[OGAWA, Tadashi @ogawa_tter]

=>
"Of Oxen and Chickens: Seymour Cray's Legacy in the Convergent Future of HPC and AI", @ProfMatsuoka, TPC Seminar Series, Jul 23, 2025 https://t.co/qqrzVuD7nQ

AI for Science Supercomputer, TRIP-AGIS, Jul 28 https://t.co/7d8S9BEefx
Grace Blackwell
15.539EF (FP8)
64.16PF (FP64)

[James Lin @JamesLinHPC]

@ProfMatsuoka Congratulations, Sensei!

[James Lin @JamesLinHPC]

The Deepseek team seems to continue keeping a low profile in China. I didn't see any booth or talk from them during #WAIC25. AI is hot, while @deepseek_ai keeps cool. Looking forward to their V4 and R2 soon!

[James Lin @JamesLinHPC]

Who will take care of the old people? i see the future. #WACI25 in Shanghai.

[James Lin @JamesLinHPC]

I saw the HUAWEI CloudMatrix 384 at #WAIC25 in Shanghai.