experimental linux kernel modules and userspace drivers for RDMA-over-usb4/thunderbolt on consumer hardware - hellas-ai/thunderbolt-ibverbs

@kele_plus @aiandcloud 现实中sonnet5就是个废物模型，性价比不如opus和fable。已经不是能力不行那么简单了，是既菜又慢还不能省钱。

Previously we covered YSERVER as an X11 server written in the Rust programming language with the help of Claude Code

Kimi K3 在我这里真的是一坨。Thinking 时间慢得要死，token 烧得老快，事情啥都干不好，分支管理也不行。那些说 K3 好用的人，我感觉全是收了钱的。我这几天用下来，感觉就是拉完了。

Google has released Gemini 3.6 Flash and Gemini 3.5 Flash-Lite. Both halve time per task relative to their predecessors and increase token efficiency, Gemini 3.5 Flash-Lite improves by 11 Intelligence Index points while Gemini 3.6 Flash does not improve in…

@trout_kilg31775 For tensor parallelism, the constraint is latency more than throughput. As long as you get RDMA (RoCE), you can get latency lower enough t o enable that.

i want the US to win in AI both in open source and proprietary models, and i am glad to see this

There’s been a lot of speculation about where we stand on open-weights models. We’ve outlined our views in full here: https://t.co/NtXQuWm2g5

个人根本没有硬件跑得起来。

小服务商跑这种东西可能根本没什么利润还不如跑规模低的模型。

大服务商跑这个假设真能赚钱那还是在帮K厂做生态。

别厂研究者也没法吸收或进行有效二次开发因为仅仅拿到权重。

这种“开源”我把它称为伪开源，蹭开源软件打下的名声。

The GCC Steering Committee announced today that they have accepted the recommended policy of the GCC AI Policy Working Group, which means no 'legally significant' code contributions made via AI/LLM agents will be accepted into this open-source compiler.