[Dev Catch Up # 104] - Gemini 3.1 Flash Live, Claude's Computer Use, SORA gets shut down, HF hf-mount, TurboQuant, CC Cheat Sheet, Email.md, Auto-Mode in CC, LiteLLM Security Incident,MiroFish & 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 104th edition of DevShorts, Dev Catch Up.
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 8800+ developers to hear stories from Open source and technology.
Must Read
Google has introduced Gemini 3.1 Flash Live, its latest real time audio model. It is built for more natural voice conversations and better task handling. Google is rolling it out across more of its products. Check Google’s announcement for more details.
Google Research has introduced TurboQuant. Its a new compression method for AI models and vector search systems. It is designed to shrink memory use without hurting model accuracy. Google says TurboQuant can cut KV memory by at least 6 times. Check Google Research’s post for more details.
Hugging Face has introduced hf mount. It allows you attach Hub storage, models, and datasets like a local file system. You can now attach a remote storage that is 100x bigger than your local disk. It may be useful for Agentic Storage. Check Hugging Face’s post for more details.
LiteLLM was recently compromised in a suspected supply chain incident. The team has shared a detailed report and the affected versions. It is worth checking if you use LiteLLM in production. Check LiteLLM’s security update for more details.
OSS Highlight of the Week
This week’s OSS highlight is emailmd. It lets you write emails in Markdown and convert them into email ready HTML. It helps to create responsive emails without writing complex HTML. Check the GitHub repo for more details.
Good to know
OpenAI is reportedly shutting down Sora, its AI video product. They do this to free up compute for its next major model called Spud. This shows OpenAI is shifting focus back to its next big model release. Check the latest reports for more details.
Anthropic has added a new computer use feature to Claude. It can operate your Mac and complete real tasks. It can open apps, browse, and do desktop actions for you. It is now in research preview inside Claude Cowork and Claude Code on macOS. Check Anthropic’s post for more details.
Cloudflare has launched Dynamic Worker Loader. It lets developers run AI code in isolated sandboxes on demand. It makes agent execution much faster and more memory efficient than containers. Check Cloudflare’s post for more details.
Anthropic has introduced auto mode in Claude Code to reduce constant permission prompts. It can decide some file writes on its own. The goal is to reduce approval fatigue while maintaining safety controls. Check Anthropic’s post for more details.
Notable FYIs
Mistral has introduced Voxtral TTS, its first text to speech model. It is built for multilingual voice generation, low latency, and custom voice support. It is available through API, Mistral Studio and Hugging Face. Check Mistral’s announcement for more details.
MiroFish is an AI prediction engine built on multi agent technology. It takes real world signals like breaking news, policy drafts, or financial data and builds a parallel digital world for simulation. So teams can test possible future scenarios in a digital sandbox. Check the GitHub repo for more details.
Luma has introduced UNI 1, a new multimodal model for image generation. It is built for better image understanding, editing, and reference based generation. Luma says it performs well across image quality. Its API access is coming soon. Check Luma’s UNI 1 page for more details.
There is a Claude Code cheat sheet for quick reference. It brings shortcuts, slash commands, MCP options, and recent updates into one place. It is useful if you use Claude Code often and want everything in one view. Check it out for more details.
Mozilla AI has released cq, a new open standard for shared agent learning. It lets agents store, share, and query past knowledge. So they do not keep repeating the same mistakes. It is worth a look if you are exploring agent memory and team level learning. Check the GitHub repo for more details.
Google has introduced Lyria 3 Pro. Its latest music generation model. It can create longer tracks of up to 3 minutes. It gives more control over parts like intros, verses, and choruses. This will be available in all google applications like Vertex AI, Gemini App. Check Google’s announcement for more details.
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.