Hugging Face Journal Club: GLM-5: from Vibe Coding to Agentic Engineering
Recording of 19/02/2026, where the Hugging Face H4 team discusses "GLM-5: from Vibe Coding to Agentic Engineering".
This is the technical report of https://huggingface.co/zai-org/GLM-5. Paper: https://huggingface.co/papers/2602.15763
How to create your own custom conversation app on Reachy Mini 🤖💬
Get a Reachy Mini : https://pollen-robotics-reachy-mini.hf.space
The Apps : https://pollen-robotics-reachy-mini.hf.space/#/apps
Reachy Mini Astronomer : https://huggingface.co/spaces/dlouapre/reachy_mini_astronomer
Reachy Mini SDK : https://github.com/pollen-robotics/reachy_mini
---
00:00 Introduction To Reachy Mini
01:26 The Conversation App
03:30 Designing the Astronomer App
05:39 Forking the Conversation App
08:03 Test and publish your App
Kimi K2.5 vs Claude Code (REAL Use Cases): New KING of Coding??
Kimi K2.5 vs Claude Code on REAL software engineering tasks! Testing open source vs proprietary AI coding agents on 3 practical challenges developers face daily.
Tasks tested:
1. Frontend: Convert wireframe to responsive Tailwind landing page
2. Bug Fix: Solve real HuggingFace Hub GitHub issue
3. Migration: Port complete Angular app to React
Both agents tackle identical prompts. See which one handles frontend design, debugging Python internals, and full-stack framework migration better.
Final Score: Kimi 3 - Claude 2
---
Links:
- Kimi K2: https://github.com/MoonshotAI/Kimi-K2
- Claude Code: https://github.com/anthropics/claude-code
- HuggingFace Issue: https://github.com/huggingface/huggingface_hub/issues/3576
- Task code: https://github.com/alejandro-ao/kimi-k2-real-world
---
⏰ Timestamps
00:00 Introduction
01:00 Task 1: Frontend Design
04:06 Task 2: GitHub Issue Fix
06:16 Task 3: Angular to React Migration
12:22 Results & Verdict
MoE Token Routing Explained: How Mixture of Experts Works (with Code)
This video dives deep into Token Routing, the core algorithm of Mixture of Experts (MoE) models.
Slides: https://huggingface.co/ariG23498/moe-routing-algorithm/blob/main/MoE_routing_algorithm.pdf
Colab Notebook: https://huggingface.co/ariG23498/moe-routing-algorithm/blob/main/routing_mechanism.ipynb
Chapter Timestamp:
Introduction: 00:00
Laying the Foundation for Mixture of Experts (MoE): 00:09
Focus on Token Routing: 00:50
What is a Mixture of Experts Layer?: 02:36
Problem Statement and Configurations: 04:48
Compute Router Logits: 08:31
Sparsity and Selecting Top K Experts: 10:54
Normalizing Logits to Router Probabilities: 12:43
Slot Selection: 14:39
Dropping Oversubscribed Tokens: 16:51
Updated Normalized Token Weights: 20:36
Updated Slot Selection and Token Slots: 21:34
Final Weight Matrix Construction: 24:35
Conclusion: 32:41
Training Dashboards with Trackio + Hugging Face
Integrate Trackio directly with Hugging Face Hub and automatically log, push, and visualize your training metrics on the model page.
---
🔗 *Links*
- 📈 Trackio Blog Post: https://huggingface.co/blog/trackio
- 🧠 Hugging Face Hub: https://huggingface.co
- 🛠️ Transformers: https://huggingface.co/docs/transformers
---
⏰ Timestamps
00:00:00 - Intro
00:00:10 - Set up fine-tuning with Transformers
00:00:27 - Add Trackio reporting to the training script
00:00:41 - Train the model and push to Hugging Face Hub
00:01:38 - Open the trained model page
00:01:49 - Explore training metrics tab and dashboard
00:02:19 - Access metrics directly from the model card
OpenAI Agents SDK Crash Course (with Hugging Face Models)
Build production-ready LLM agents with the OpenAI Agents SDK.
---
🔗 Links
- Agents Series Playlist: https://www.youtube.com/playlist?list=PLo2EIpI_JMQsw73Q7febEuUFBFcqIg2Mq
- Code from the video: https://github.com/huggingface/hub-tutorials/blob/main/notebooks/03-openai-agents-sdk.ipynb
---
⏰ *Timestamps*
0:00:00 - Intro
0:01:42 - Quickstart
0:05:33 - Running open models from Hugging Face
0:08:54 - Tracing runs in the OpenAI platform
0:12:13 - Structured Output
0:15:40 - Multi-agent orchestration: handoff vs manager
0:21:33 - Building and using tools
0:23:39 - Streaming tokens and agent events
Reachy Mini at Nvidia's Jensen CES keynote
Building Agents with Smolagents
Learn how to create minimalist agents with open models using Smolagents by Hugging Face.
---
🔗 *Links*
- 🧑💻 Code from the video: https://github.com/huggingface/hub-tutorials/blob/main/notebooks/02-intro-to-smolagents.ipynb
- 🤗 smolagents (Docs): https://huggingface.co/docs/smolagents
- 🧰 Hugging Face Hub & Spaces: https://huggingface.co/spaces
---
⏰ *Timestamps*
0:00:00 - Intro
0:01:34 - Environment setup
0:02:52 - Creating agents
0:08:58 - Create an Image-Generation Tool
0:14:13 - Hugging Face Hub & Spaces
0:17:07 - Chatbot UI with Gradio
How to contribute to Open Source - 7 EASY steps 🤗
This video walks through the steps of making an open-source contribution, with a practical example of fixing a bug in Trackio with Cursor.
Error 400: API key not valid. Please pass a valid API key..
Domain code: global
Reason code: badRequest
Hugging Face Journal Club: GLM-5: from Vibe Coding to Agentic Engineering