TechMD

🔬 MD fueled by a deep passion for Medicine & Tech. 🌐 Exploring the frontiers of VR/AR/MR & BCI. 🤖 AI enthusiast.

[TechMD @TechMDAI]

Built a local agentic stack where Hermes orchestrates, Codex 5.3 designs workflows, and DGX Spark runs the model team.

Telegram is the front door.
Hermes turns messages into tasks.
Codex helps build and refine the pipelines.
Local models execute natively on Spark.

Current local lineup:
• gemma-4-26B-A4B
• qwen3.5-9b
• qwen3.5-4b

So the flow is:
Telegram → gateway channel → Hermes → Codex-assisted workflow logic → local harness → DGX Spark inference → validated response

Not just chatbot plumbing.
More like building an operating system for coordinated intelligence.

[TechMD @TechMDAI]

@loktar00 @Alibaba_Qwen They have to get ahead of the new qwen 3.7s

[TechMD @TechMDAI]

When qwen 3.7 OSS?

[TechMD @TechMDAI]

@franktheglock @Ominousind She is a beauty

[TechMD @TechMDAI]

RTX Pro 6000 15k at DELL

[TechMD @TechMDAI]

@gospaceport @Ominousind The goodies . I feel the passion

[TechMD @TechMDAI]

@DCRando2020 The Chinese models giving us a life line

[TechMD @TechMDAI]

@DCRando2020 The bar is rising fast.

[TechMD @TechMDAI]

@gospaceport @Ominousind Definitely. I watched your Threadripper Pro video today, and your points about the rules of homelabbing particularly around upgrading and replacing components strategically. Definitely there upsides with GPUs

[TechMD @TechMDAI]

@Ominousind Cheers!! to dumb luck.

[TechMD @TechMDAI]

@Ominousind Right timing Bijan. What gave you the foresight?What a time to be alive!!

[TechMD @TechMDAI]

@nicdunz Codex broke today

[TechMD @TechMDAI]

@thsottiaux @AnthonyBerlin I am getting this on windows too

[TechMD @TechMDAI]

@Teknium now for the mobile app. Desktop app rocks

[TechMD @TechMDAI]

@Scobleizer Who is who?

[TechMD @TechMDAI]

@Teknium @DegenApeDev Love it

[TechMD @TechMDAI]

@loktar00 Bandwidth is important

[TechMD @TechMDAI]

Interesting

[stevibe @stevibe]

Qwen3.6 35B A3B can't fill out a paper form on its own. But give it NVIDIA's LocateAnything-3B — the #1 trending model on HuggingFace — as its eyes, and the two small models get it done together.

(The test: place each element at the right pixel position on a blank form image, not type into a field.)

Setup:
> Qwen is the brain (main model), LocateAnything is the eyes (helper model acting as a tool).
> I gave Qwen a new tool: ask "where's the email field?" and LocateAnything returns the exact x, y, width, height.
> The blue boxes on the screen are its detections. Look how tight they are — it nails every field.

Result:
> Qwen3.6 35B A3B + LocateAnything-3B: form completed, all info correct.
> Name, DOB, ID, gender, marital status, nationality, email, phone, address, postal code: all landed in the right field areas.
> Character-box alignment still a touch loose, but every value is where it belongs.
> 9m10s, 224.5k input, 24.3k output, 21 turns.

Why it matters:
> Qwen alone can't finish this test. Bolt on a 3B model that does exactly one thing > locate > and suddenly it can.
> A combination of small models can do the work of a single large one.

[TechMD @TechMDAI]

Im liking the Hermes desktop

[TechMD @TechMDAI]

@DCRando2020 @0xSero I use when doing inference on Mac with DGX.

[TechMD @TechMDAI]

@aijoey @NVIDIAAI @nvidia @PavloMolchanov @Teknium @NousResearch I will fire it up thanks. I been seeing some interesting things