[Dev Catch Up # 94] - GLM 4.7, Claude in Chrome, Manus joins Meta, Nvidia Groq Deal, IQuest-Coder-V1, Gitmal, Qwen3-TTS, Deepseek's mHC, Alibaba's GUI Agents, Meta's Self Play SWE-RL and much more!
Bringing devs up to speed on the latest dev news from the trends including, a bunch of exciting developments and articles
Welcome to the 94th edition of DevShorts, Dev Catch Up.
Happy New Year 2026!!!
For those who joined recently or are reading Dev Catch Up for the first time, I write about developer stories and open source, partly based on my work and experience interacting with people all over the globe.
Thanks for reading Dev Shorts! Subscribe for free to receive new posts and support my work.
Some recent issues from Dev Catch up:
Join 8500+ developers to hear stories from Open source and technology.
Must Read
Claude finally comes to the browser. It can navigate pages, click buttons, fill forms, pull analytics, organize Drive files, clean your inbox, etc. It is in beta for paid users. Anthropic highlights safety risks, so start with trusted sites only. Read about Claude in Chrome to learn more details.
Z.ai just released GLM 4.7. It is an open-source coding model. It crosses 70% on SWE benchmarks. It focuses on coding, reasoning, and tool usage. It is priced much lower than Claude and targets developer agents. Read about GLM 4.7 Coding Model to learn more.
A new generation of code models called IQuest Coder and LoopCoder has been released. They are built for software engineering and competitive programming. The Loop architecture improves throughput and reduces memory usage. Check IQuest Coder to know more details.
Alibaba’s Qwen released two new Flash TTS models. They focus on voice design and voice cloning. One model creates a reusable voice from text descriptions. The other clones a voice from audio and supports around 10 languages. Read about Qwen3 TTS to learn more details.
OSS Highlight of the Week
This week’s OSS highlight is Gitmal. It is a static page generator for Git repos. It turns your repo into browsable HTML pages with files, commits, syntax highlighting, and markdown rendering. You can run it locally, with Docker, or self-host it for sharing code without GitHub. Check Gitmal GitHub Repo to explore and try it out.
Good to know
Manus is joining Meta. Manus builds a general-purpose autonomous AI agent. It can handle tasks like research, coding, and data analysis. The service will continue to run and also integrate into Meta AI. Read about the Manus Meta partnership to know more details.
NVIDIA is spending about $20B to acquire Groq’s ultra-fast inference technology and its engineering talents, including the founder, Jonathan Ross. This is NVIDIA’s biggest move so far. It is a licensing deal rather than a full acquisition. Check NVIDIA Groq deal to know more details.
If you are using plugins with Claude Code, check this GitHub repo. It contains official plugins for Claude Code. One of them is the Ralph Wiggum plugin for running iterative agent loops. Check the Claude Code plugins repo to learn more details.
Most developers have started using AI for coding. This Hacker News thread shows a discussion on how to get better at AI programming. A Claude Code team member shares tips like using CLAUDE.md, Plan mode, and letting the model check its own work. The discussion also highlights context loss and why starting fresh chats often works better. Check the discussion to know more.
Notable FYIs
DeepSeek just published new research on model architecture. The paper introduces mHC, a method that improves training stability and scalability with little extra compute cost. Tests on 3B, 9B, and 27B models show better benchmark results. Read the paper to understand where DeepSeek may be heading next.
Google is expanding NotebookLM in two ways. It adds Lecture Mode in NotebookLM. It can convert multiple sources into an AI narrated lecture for passive learning. You can also upload NotebookLM notebooks directly into the Gemini app. Read both updates to know more details.
If you are working with coding agents, check Conductor. It lets you run multiple coding agents at the same time. You can spin up parallel Claude Code and Codex agents in isolated workspaces with a clean UI. It handles git worktrees, reviews, and merges for you. Check Conductor to learn more.
If you are working on image generation, check FLUX.2 dev Turbo. It is a LoRA adapter for the FLUX.2 dev base model. It generates high quality images in just 8 steps. It supports text to image and image editing. Check FLUX.2 dev turbo for more details.
Meta rolled out Self play SWE RL. In this setup, the model writes buggy code and learns by fixing its own mistakes. The work shows agent speed is a system problem, not a model problem. It introduces Agent Infer to reduce token waste and speed up tasks. Read this to learn how Meta improves AI agents.
Tencent released HY Motion 1.0, a text to 3D human motion generation model. It turns simple text prompts into skeleton-based 3D animations. The model scales with better motion quality and instruction following. Check HY Motion 1.0 to learn more details.
Alibaba introduced MAI UI, a foundation GUI agent for mobile assistance. It turns text instructions into real actions across apps like office tools, shopping, travel, and daily tasks. MAI UI shows strong results in GUI grounding and mobile navigation. Read about MAI UI to learn how mobile GUI agents are evolving.
That’s it from us with this edition. We hope you are going away with a ton of new information. Lastly, share this newsletter with your colleagues and pals if you find it valuable. A subscription to the newsletter will be awesome if you are reading it for the first time.