Tim Wu @changtimwu

Fight for the second half.

 Taipei
Joined December 2007
格力今年也加入贊助
終於到週末了 不錯，我還活著
2
不練別的就練這個 好好刷一杯抹茶比裝女人走路實用多了
Leo Lin
 @leo_lin6
Nov 6
Finally got the full footage for the matcha making policy, includes the "risky" water pouring step. 4x speed. Whisking step was very inaccurate because the left gripper camera broke and we had to take it off; so it really affected the grasping of the whisk and i had to trim that part of the footage
重溫一次 NV 之前最不順的時候, 這樣也是能爬到世界巔峰的 greenbot.com/nvidia-ceo-acce…

Nvidia CEO accepts Tegra 4i failure; To shift focus to cars and other mobile segments

Nvidia took the world and the smartphone industry with a storm when it released the first dual-core processor for mobile devices back in 2011 -- the Tegra 2. It managed to main that momentum with the...

greenbot.com
誰有跑過啊 這個所有 model 都可套用嗎? docs.vllm.ai/en/stable/featu…
連 storage 廠家都可以加入賣鏟子行列。各行各業快想想自己要怎麼硬蹭 AI
fin
 @fi56622380
Nov 3
Replying to @chumacn
如果你关注美股，关注最近WDC, Hynix财报，就会发现存储是AI半导体目前最确定的趋势之一。因为LLM推理比例上升，从内存到SSD到HDD全面紧缺，而需求端一直在上升 “过年前”如果指的是春节，这个风险应该是很小的，这次的存储周期目前还在早期
1
如果你一直 challenge 一篇 paper 的論點, notebooklm 不會 "You're absolutely right", 反而是一直幫那篇 paper defend
最 high 的應該是 @ehang 吧? Do you think Elon is serious about selling flying cars? 😆
How Football Saved Humans - Great Book to Read @HowHumans
Nov 2
Elon says Tesla may unveil a flying vehicle within months. Elon: "Let's just put it this way.If you took all the James Bond cars and combined them, it's crazier than that." @elonmusk @grok is already aware of the car @joerogan watch this
Tim Wu retweeted
Dexmal @Dexmal_AI
Oct 31
🤖We've released RT-VLA---our optimized inference code for the Pi0 model by @physical_int, achieving up to 30fps on a single RTX 4090! Check it out on GitHub: github.com/Dexmal/realtime-v…

GitHub - Dexmal/realtime-vla: Running VLA at 30Hz frame rate and 480Hz trajectory frequency

Running VLA at 30Hz frame rate and 480Hz trajectory frequency - Dexmal/realtime-vla

github.com
4
1
5
補充資料 piped.video/watch?v=poD0m4e_…

真正的恐怖永遠都是來自看不見的地方，「以人為食」可不是虛言

-----------------------------------【相关链接】会员福利介绍：https://youtu.be/OyumK7OV5Es加入barry的后盾团的链接https://www.youtube.com/channel/UCHvGGLbrTQnxsx7IDkA5X9w/join大家都喜欢看...

youtube.com
the last of us 可以再拍一系列, 背景寫在第一代 MCU (mushroom computing unit) 發表上市後, 民間瘋狂種植算力, 真 computing farm 相繼開場,便宜的算力讓生成式AI應用速度呈指數型爆炸,人類文明走向一個新世紀, 但漸漸發現菇菇神經網路有個不可控的小誤差...😁 jiqizhixin.com/articles/2025…
1
結果跟他們之前預估的一樣, 總結就是 Number of CUDA cores matters!
JetsonHacks
 @Jetsonhacks
Oct 31
The results are in, so we can now compare Spark vs Thor LLM performance! This first entry simply links the benchmark results, I'll interpret them in the next article. wp.me/p7ZgI9-3TJ
明明就“獵殺代理人”嘛
David Marcus
 @davidmarcus
Oct 30
Wait. Did 1X just invent remote immigration?
這裡也可以 power by deepseek 喔, 你倒是把 reasoning process 秀出來看看
柴古唐斯越野賽, 這次再把轉播技術成就推上一個新的天花板, 第一次看到空拍+跟拍+即時配樂這麼流暢
1
Tim Wu retweeted
meng shao
 @shao__meng
Oct 31
HuggingFace 发布的超长技术博客（200页，2-4天才能读完），完整记录了团队训练 SmolLM3 的全过程，对于想训练小模型的团队，必看！ 从训练指南、训练核心流程、最佳实践、基础设施和资源推荐，强调「通过精心策划的数据、稳定基础设施和优化流程，实现高性能模型，而非依赖巨型计算资源」，非常值得慢慢学习。 huggingface.co/spaces/Huggin…
8
168
4
623
Tim Wu retweeted
bishara
 @bishara
Oct 30
THIS CLANKER IS USELESS
829
1,973
668
23,928
0
柯柯好強啊
既然升級了改叫 BY3 吧😆
“既然內建 CX7 了, 那就可以跑 GPUDirect 囉?", 別傻了, NV 出菜哪有可能這麼香啦 forums.developer.nvidia.com/…

DGX Spark GPUDirect RDMA

So, my dmesg was also clean and modinfo for my peermem module looks the same. The interesting thing about it though is the depends: line. It is empty whereas on a normal system with peermem it looks...

forums.developer.nvidia.com
1
Load more