NVDA完了，这个梁赛预言中国还有个氢弹级别的

[gambitnyc]

https://m.weibo.cn/1725941200/5127474623286551

[Dachun]

小心FBI

[mlforlife]

宁可信其有，不可信其无啊，东大看来为懂王准备了组合拳

[pathdream]

东大不给美弟一个中指

[toledo]

今年股市真是风波诡谲。

[Dachun]

今年股市真是风波诡谲。

超限战

[gambitnyc]

2025年东大科技进入寒武纪爆发

[guagua1220]

宁可信其有，不可信其无啊，东大看来为懂王准备了组合拳

这个可以有

[直接投降]

NVDA完了, QQQ也完了, 美股也完了.

[drifter]

Groq 要发达？有朋友刚从AMD去groq

[gambitnyc]

Groq 要发达？有朋友刚从AMD去groq

Cerebras马上IPO了

AI inference providers. Cerebras is faster than Groq and can process inputs at speeds that are 20 times faster than NVIDIA GPUs.
Features:

Cerebras
The Cerebras WSE-3 has 4 trillion transistors, 900K cores, and 44 GB of on-chip memory. It can accelerate sparse linear algebra, tensor workloads, and spectral algorithms.

Groq
The GroqCard Accelerator has up to 750 TOPs, 188 TFLOPs, and 230 MB of SRAM per chip. It's designed for high-throughput AI inference with low latency.

[biggestballs]

Cerebras马上IPO了

AI inference providers. Cerebras is faster than Groq and can process inputs at speeds that are 20 times faster than NVIDIA GPUs.
Features:

Cerebras
The Cerebras WSE-3 has 4 trillion transistors, 900K cores, and 44 GB of on-chip memory. It can accelerate sparse linear algebra, tensor workloads, and spectral algorithms.

Groq
The GroqCard Accelerator has up to 750 TOPs, 188 TFLOPs, and 230 MB of SRAM per chip. It's designed for high-throughput AI inference with low latency.

拿到过Cerebras的面试，可惜挂求了

[fyou]

碰瓷的
前两天刚刚Intel+amd碰过瓷了
这个发晚了。。

[BBQ]

完了。美股一键清零

[Iamhappy]

氢弹什么时候 爆。 下个星期吗。 有知道的吗

[drifter]

氢弹什么时候 爆。 下个星期吗。 有知道的吗

肯定得回调差不多了养肥了再爆 又捞一笔

[longtian]

微博发个消息，大家一起上，保证战果扩大10倍

https://m.weibo.cn/1725941200/5127474623286551

[Mountainlion]

吹这么玄乎，肯定屁都不是。
同时有两个巨大影响力的可能性微乎其微。

https://m.weibo.cn/1725941200/5127474623286551