Tim Wu (@changtimwu) profile picture

Fight for the second half.

Taipei
Joined December 2007
Photo by Tim Wu (@changtimwu)Photo by Tim Wu (@changtimwu)Photo by Tim Wu (@changtimwu)Photo by Tim Wu (@changtimwu)Photo by Tim Wu (@changtimwu)Photo by Tim Wu (@changtimwu)Photo by Tim Wu (@changtimwu)Photo by Tim Wu (@changtimwu)Photo by Tim Wu (@changtimwu)Photo by Tim Wu (@changtimwu)
While its ruling govt remains silent, many among Taiwan public and vast majority of TW biz & tech sectors are heavily supportive of @JensenHuang. Tonight Taipei 101 lights up "Taiwan ❤️ Jensen", organized by Taipei Computer Association, other tech companies and Taipei 101 itself.
54
193
19
1,974
60,536
The government shouldn’t pick winners and losers in markets nor CEOs of companies. America is a country where everyone has a chance to start a business, to have a fair chance to compete, and succeed or fail based on their own merits.
Nvidia will be one of more than two dozen companies designated artificial intelligence “national champions” as part of an effort by the Trump administration to help AI firms win business overseas on.wsj.com/4iU222d
139
182
16
2,850
89,137
Wir NVIDIA-Aktionäre sollten uns verbünden und den NVIDIA-Verwaltungsrat neu besetzen. Dank Jensen Huang sind rund 27.000 NVIDIA-Mitarbeiter durch Mitarbeiterbeteiligungsprogramme zu Millionären geworden. Jensen Huang hat den Wert der NVIDIA-Aktie um 70.000 % gesteigert. Der NVIDIA-Verwaltungsrat handelt kleingeistig und weder im Interesse der Eigentümer noch im Interesse der Mitarbeiter. Moralische Schuld hat Jensen Huang gegenüber seiner Ehefrau. Wenn sie ihn vor die Tür setzt und mit 70 Mrd. USD aus der Scheidung hervorgeht, ist das eine ausreichende Strafe für seine Fehltritte. Als NVIDIA-Aktionär fordere ich, den NVIDIA-Verwaltungsrat zu feuern und jeden Huang zurückzuholen.
962
4,386
191
37,611
1,046,440
Não entendo os EUA. Ter uma amante derruba um CEO. Mas pedofilia não derruba o presidente?
6,083
68,472
1,274
578,124
12,679,242
For my first post, I’m sharing a letter @NVIDIA signed on why open models matter. AI will transform every industry, power every company, and be built by every country. Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty. The world needs both frontier closed models and frontier open models. images.nvidia.com/pdf/Open-W…
16,561
34,451
12,073
193,804
69,421,328
Tim Wu retweeted
百亿身家的 Palantir: 很多公司走到 FDE 2.0了,而太多公司连 FDE 1.0 还没开始... 视频中举了沃尔玛生鲜配送案例: 1/ 大早上雪天中,沃尔玛生鲜配送车故障了 2/ 办公室里,运营,物资保障,财务,IT盯着花了千万百万开发的 ERP, BI 系统开会 3/ 烟味,咖啡味,披萨味,满屋。财务说要赔10万美刀,法务说准备好官司,没人关心那一车子生鲜 4/ FDE 2.0 进场,打开 Foundry 定位到出事地还有备用生鲜车,物资保障部这才电话过去顶上 事后复盘,为什么 ERP/BI 通通失效,FDE 却如此神通? FDE 过来人老黄说,去年我就站你们运营身后看着他把车辆数据从 ERP 导出 CSV,又从 CSV 筛选出半年没做保养的车,去 BI 系统看 WIP (working in progress) 的出车记录,一天的工作就是盯车,我猜测他一定是在监控车况 “我承认,我是有私心的,但这次我赌赢了...”
8
21
130
14,739
都能直接讀寫 native machine code, 編輯 native weight 也不稀奇了, 以後也不用什麼蒸餾了, 調幾個參數某項能力就大躍進 7.9% -> 48.1% , 這根本是 AI model 界極速種群能力提升的基因編輯技術, 恐怖喔~~🤣
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math. It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training. We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵 Blog: rikiyat.github.io/blog/weigh…
Readers added context they thought people might want to know
LLM was given weights explicitly as text in prompt and used code/forward passes (per authors' logs) to ablate Qwen looping direction, a known thinking-mode issue per model card. Primarily fixed repetition in that mode, not general 'natively' improving weights/math. rikiyat.github.io/blog/weight-sp… github.com/RikiyaT/lm_und… huggingface.co/Qwen/Qwen3.5-0… huggingface.co/RikiyaT/Qwen3.…
1
170
又不能喝咖啡了 又可以喝咖啡了 msn.com/zh-tw/%E9%A3%9F%E7%8…
1
70
這台終於有值得一看配置😆。在單台 GB10 上跑 Qwen3.8-Flash-Next * Local SSD mmap 穩態能跑到 68–70 tok/s. * 但如果系統有其他 process 搶 RAM,page cache 被清, 引發 cold faults,速度會直接斷崖式下跌。 * 想省記憶體別動本體(已經是 W4A16),去把 49GB 的 PLE table 轉 INT4,現賺 25GB RAM 還不傷 eval。完整 stack 與設定細節放留言。 github.com/Saren-Arterius/qw…
2
81
這套是直接善用 KIND, 然後又跑了 vm, 想法大概是這一層只是要 relay GPU resource, heavy 一點沒關係 saiyam1814.github.io/kiac/do…
10
Charles Kindleberger "There is nothing so disturbing to one's well-being and judgment as to see a friend get rich." "這世界上最擾亂人心的事,莫過於看著身邊的朋友賺大錢" 說真的, 或許每一輪泡沫都是靠這種嫉妒心態堆積出來的, 整天聊投資也是種 party killer
3
66