Hello World — Welcome to Munglabs
Our first blog post. A humble lab, built entirely with a local AI stack.
Hello World
Welcome to the Munglabs blog. If you’re reading this, the whole stack is working — from the local LLM that wrote this post to the Giscus widget below powered by GitHub Discussions.
What is Munglabs?
Mung means “just” in Javanese. We’re just a lab — but what we do inside is anything but casual. This is a space for:
- AI Research — Exploring machine learning, LLMs, and practical implementations
- Electronics & IoT — Microcontrollers, sensors, connected devices
- Software & Tools — Developer tools, automation, open-source utilities
- DIY & Tinkering — Prototyping, 3D printing, experiments
- Content Creation — Sharing what we build
Humble in spirit, rigorous in practice.
This landing page was 100% vibecoded
Yes, really. Every line of this website — the Astro config, the React components, the Tailwind styles, the MDX blog, this very post — was written by a local LLM. No cloud APIs. No external services.
Here’s the stack:
| Component | Detail |
|---|---|
| Model | Qwen3.6 27B FP8 |
| Inference | SGLang |
| GPU | 4× RTX 5060 Ti |
| CPU | Xeon (Broadwell-EP) |
| Orchestrator | Opencode |
| Framework | Astro + React + Tailwind CSS |
The entire conversation happened locally. The LLM ran on-prem, served via SGLang, orchestrated by Opencode. Zero tokens sent to the cloud.
See the code
This landing page is open source. You can browse the full codebase, fork it, or just look under the hood to see what vibecoding looks like in practice:
👉 https://github.com/munglabs/fe_landing_page
What’s next?
More posts to come. Project write-ups, technical deep-dives, lab notes, and whatever else we build. Stay tuned.
Got thoughts? Drop a comment below.