@CloudNativeFdn Ambassador, @Microsoft MVP, PyCon China & KCD Beijing organizer.🥑Open Source + Cloud Native + AI Infra. Working at @Kong . Opinions are my own
- Tweets 4,197
- Following 808
- Followers 3,190
- Likes 8,935
turbopuffer 这篇文章虽然有点标题党，但在 hacknews 下面的这些讨论还都是蛮有价值的。在搞向量数据库/索引/RAG b的话，推荐看看这里的一些讨论。
RIP, vector database
news.ycombinator.com/item?id…
23
1,693
我们最近刚发布了新产品 Volcano。
大家如果有什么想法想快速部署上线之类的，那么这个平台应该都很合适，也比较适合让 agents 去操作，欢迎试用体验！
volcano.dev/
JUST ANNOUNCED: Volcano 🌋
Volcano is everything your apps and agents run on, in one project.
• Branchable Postgres databases
• Durable functions
• Frontend hosting
• Authentication
• File storage
• Distributed locks
Try it today: volcano.dev
4
2
21
2,950
看来之前网传的那份数据就是 Gemini 4 Argon 了，它也学会了给模型加代号，不再只是 pro/flash 这样了。
另外一点就是价格，旗舰模型，定价和 GPT-6.1 Sol 同样，这就很厉害了。
（但是以上这些都只是它宣传的内容，现在还没实际上手，等有权限了再体验一波
Announcing Gemini 4 Argon, our new frontier model.
Argon is built to sustain deep reasoning across complex, long-horizon workflows and delivers frontier performance in complex workflows across real-world software engineering, enterprise knowledge work like legal and finance, and cybersecurity defense.
We’re also expanding the model’s output token limit to an industry-leading 1M tokens.
Argon is currently rolling out to a set of trusted cyber defenders in the Fairwind Program, with broader availability as soon as possible.
2
3
1,426
刚看到 @gmi_cloud 公布了他们 6.68 亿美元的 B 轮融资，说起来 GMI 的营收曲线可以说很漂亮了， 2024 年才算刚起步， 现在 ARR 6 亿美元，年增 5 倍的样子，这也说明这类 neocloud 的市场需求很大。
而且它还是 NVIDIA 全球 7 家 Reference Platform NCP 之一，但这个事儿吧，也和它的起源有很大关系
Announcing our $668M Series B!
Led by ARCHIV with participation from @nvidia.
This capital expands our GPU capacity across the U.S., Taiwan, and APAC, and scales our inference platform.
Our contracted ARR has reached more than 9x since the end of 2025.
Thanks to everyone who is building with us 🙏
2
9
2,454
这赛道说不卷是不可能的， SGLang 已经推出 Decision API （在 nightly build 中） 另外支持了将 Qwen3.8-27B 和 Qwen3.5-35B-A3B 作为 Decision model 使用，看起来效果还不错
We turned Qwen3.8-27B into a multimodal decision model.
It beat Pokémon FireRed’s elite four and champion with sub-100 ms decisions from live game state.
With SGLang’s native /v1/decisions, you can now turn LLMs and VLMs into classification and scoring models.
We also added /v1/systemone so Jev-like open models can work with the TypeSafe SDK.
6
36
11,185
让我来总结一下 OpenAI 的 devday：
OpenAI 通过和很多厂商/产品合作，将自己打造成真正的 token 工厂，无论你在使用什么产品，使用闭源模型还是开源模型，都可以通过 OpenAI 进行消费和结算。
这才是真正的 Token Factory。
OpenAI's DevDay made one thing crystal clear:
OpenAI is becoming the token factory — the settlement layer for ALL of AI.
Use any product. Run closed or open models. Bill and pay through OpenAI.
They stopped selling "models"
They're selling the right to count and charge tokens
102
8
2,027
OpenAI's DevDay made one thing crystal clear:
OpenAI is becoming the token factory — the settlement layer for ALL of AI.
Use any product. Run closed or open models. Bill and pay through OpenAI.
They stopped selling "models"
They're selling the right to count and charge tokens
1
1
2,389
前几天发的 pi 在它的核心中添加 mcp 功能的事情已完成。
甚至官方还发布了篇文章，我觉得还挺有意思， 大家有兴趣可以看看
earendil.com/posts/you-said-…
嗯？ 当初咋说的来着？
Pi 打算要加上 codemode 和 MCP 的支持了
github.com/earendil-works/pi…
1
23
7,785
I used Claude Opus 5.5 to create an unofficial promo video for @kong's new mascot, Karl.
The workflow was powered by Hy-Image-3.5-preview and @MiniMax_AI, both available on @gmi_cloud (Hy-Image-3.5-preview is currently free to use).
Pretty impressed with how it turned out 🎬✨
4
1
10
1,558
好好好，这么玩是吧？
更强的能力，更低的价格，但是市场的存储，GPU 都在涨价，请问钱从哪儿来，又去了哪里？
openai.com/index/introducing…
1
8
1,555
顺手做了个非官方的宣传视频，没做太多动效，但是这个我觉得足够用了，主要是省事儿（也省钱，整体成本就花了不到 $5 🤣
I used Claude Opus 5.5 to create an unofficial promo video for @kong's new mascot, Karl.
The workflow was powered by Hy-Image-3.5-preview and @MiniMax_AI, both available on @gmi_cloud (Hy-Image-3.5-preview is currently free to use).
Pretty impressed with how it turned out 🎬✨
1
784
