George

Living in Brisbane

[George @PCoal]

@Snixtp Thanks mate

[George @PCoal]

@Snixtp Could you please share the link for the splitter? Plan to get three for adding more cards? Thanks

[stevibe @stevibe]

I explored a further possibility with local models: Qwen3.6 35B A3B + NVIDIA LocateAnything-3B as a local Computer Use agent (proof of concept).

In the demo, I asked it to switch my Mac to light mode. It did. Then back to dark. Did that too — finding the right toggle in System Settings, clicking it, and verifying the change itself.

It's fully screenshot-based, so no Accessibility API needed. If it's on screen, the agent can see it and act on it. This runs entirely on your own hardware — private, local, built from two small open models.

[Ken W @kenw_2]

这是哪位大师的vibe coding神作么？

[George @PCoal]

@Snixtp Thanks. But will the split reduce the inference speed or it just minor? Interesting to see how this works

[George @PCoal]

@joaosump @0xSero I have the same dual 3090 and got much better results with vllm even running with wsl&docker on windows

[George @PCoal]

It would be cooler if we could run one hermes on each of machine and they are aware of each other and can share skills and transfers data within the same home network

[Teknium 🪽 @Teknium]

Hermes Agent tip of the day:

You can backup and transfer your agent cleanly and simply. Want to move it to a new VPS that is bigger? Take it to your new Mac Mini?

`hermes backup` - > creates a zip

install hermes on a new machine fresh

transfer the zip to that machine

then run `hermes import`

done :)

Full docs:

https://t.co/gJlheqB7it

[Teknium 🪽 @Teknium]

Hermes Agent tip of the day:

You can backup and transfer your agent cleanly and simply. Want to move it to a new VPS that is bigger? Take it to your new Mac Mini?

`hermes backup` - > creates a zip

install hermes on a new machine fresh

transfer the zip to that machine

then run `hermes import`

done :)

Full docs:

https://t.co/gJlheqB7it

[George @PCoal]

@pupposandro Thanks. I will try tomorrow. Also I noted z-lab updated Qwen3.6-27B-DFlash 8 hrs ago

[noname @malikwas1f]

🧵 Qwen3.6-27B on consumer RTX 3090 — honest update.

#vLLM #LLM #Qwen3 #RTX3090 @vllm_project @Alibaba_Qwen

That "85 TPS sustained on 1× 3090" post a few days ago? The TPS was real. **The tool calls were silently broken.** A cudagraph × spec-decode bug class we didn't catch until running a full functional verify.

Filed upstream → root-caused → fixed. Now serving correctly with measured numbers across 5 configs on 1 and 2 cards. Thread 👇

(202 chars body, plenty of headroom)

[Qwen @Alibaba_Qwen]

🚀 Meet Qwen3.6-27B, our latest dense, open-source model, packing flagship-level coding power!

Yes, 27B, and Qwen3.6-27B punches way above its weight. 👇

What's new:
🧠 Outstanding agentic coding — surpasses Qwen3.5-397B-A17B across all major coding benchmarks
💡 Strong reasoning across text & multimodal tasks
🔄 Supports thinking & non-thinking modes
✅ Apache 2.0 — fully open, fully yours

Smaller model. Bigger results. Community's favorite. ❤️
We can't wait to see what you build with Qwen3.6-27B! 👀

🔗👇
Blog: https://t.co/P2Zx7FwMxB
Qwen Studio: https://t.co/c4vm4LuZrU
Github: https://t.co/zKDEbv0R4U
Hugging Face:
https://t.co/N67hyzxvfr
https://t.co/SSdtbWRDap
ModelScope:
https://t.co/xODf1pj9kw
https://t.co/xXhoqlJ2AB

[Raymond Weitekamp @raw_works]

crazy preliminary results from qwen 3.5 last night:

Preliminary — Qwen3.5 + dspy.RLM on LongCoT-Mini:

27B lands at #2 (33%) — behind only GPT 5.2, +11pp ahead of Gemini 3 Pro.

9B lands at #4 (17%) — still beats Sonnet 4.5.

RLMs unambiguously SOTA on this, more soon!

[Qwen @Alibaba_Qwen]

⚡ Meet Qwen3.6-35B-A3B：Now Open-Source！🚀🚀

A sparse MoE model, 35B total params, 3B active. Apache 2.0 license.

🔥 Agentic coding on par with models 10x its active size
📷 Strong multimodal perception and reasoning ability
🧠 Multimodal thinking + non-thinking modes

Efficient. Powerful. Versatile. Try it now👇

Blog：https://t.co/EXx5y466su
Qwen Studio：https://t.co/bg4tAU1p74
HuggingFace：https://t.co/w4pDX14DZS
ModelScope：https://t.co/SuRyLzdQiO
API（‘Qwen3.6-Flash’ on Model Studio）：Coming soon～ Stay tuned

[George @PCoal]

@loktar00 @edgaraveloso Which motherboard and cpu you used for the 6 gpus?

[郭宇 guoyu.eth @turingou]

今天正式发布了我的第 13 个 vibe 产品 https://t.co/L09qUdQhGA

这款产品比较特殊，它是一个电话服务。准确来说，tuwa 是一个 AI 电话网络，连接着世界上超过 100 种不同语言的人们和互联网上的 agent。

任何人都可以不下载 tuwa 而使用它，你只需要拨打免费的转接热线电话 [phone]，告诉它你需要打给哪个号码，tuwa 就会帮助你拨打对应的号码，你说自己的母语，对方听到的却是TA的母语，反过来也是一样。兼容任何电话，对方不需要安装应用。固话、手机，世界上任何一个角落，都可以。

tuwa 支持 100 多种语言的实时翻译，你甚至可以在打电话时随便切换语言和对方对话。除此之外，tuwa ��支持语音克隆，每一通电话，都会让你的 AI 语音听起来更像你。

当然，我也为它设计了方便的 web app，如果你想，可以不通过转接电话而使用 web app 拨打，并设置自己喜欢的声音，使用外呼 agent 拨打电话，连接自己的 agent（例如 openclaw 或者 codex/claude code）并让他们自由的呼入与呼出。

外呼电话 agent 是我最喜欢用的 tuwa 功能，只需要交待清楚事情，比如完成餐厅预订，它就会在你希望的时间主动拨打对方的电话，说明来意，达到目的，并记录和翻译所有对话内容。

tuwa 的使用和收费都很简单，每月免费额度，固定套餐，按需付费。

这个产品的命名灵感来自于日语的「通話 tsuwa」最初，我只是想设计一个能帮我预订餐厅的电话服务，但后来，我在 vibe 的过程中慢慢意识到，世界上仍然有很多人无法体验 AI 带来的变化与便利，而电话，是连接他们最简单与自然的方式。我希望 tuwa 能帮助外语普及率低，偏远地区和第三世界国家的人们体会到这一点。

[George @PCoal]

@thmsenglsh When I moved to Perth in 2014 from China and found that almost all shops were closed after 6pm.

[Jessi 🤘🏻 @jessinicoleb]

because you can’t open windows in space

[George @PCoal]

@simon_ree We should have Claude to run the country