[Dev Catch Up # 100] - Perplexity Computer, Nano Banana 2, Claude Code Remote Control, ZeroClaw, WebSocket Mode in OpenAI API, Claude Opus 3 retirement, Plano, Memento 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 100th 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 8700+ developers to hear stories from Open source and technology.
Must Read
Google has released Nano Banana 2, a new image model that brings Nano Banana Pro quality at Gemini Flash speed. It improves lighting, textures, and details while keeping Flash level speed. It is rolling out across Google’s Gemini platforms. Check Google’s announcement for details.
Perplexity has introduced Perplexity Computer. It is a general purpose digital worker. It can research, build, deploy, and manage a project from start to finish. It runs in the same chat interface. Each task runs in its own sandbox. It uses 19 different models for tasks. Check Perplexity’s announcement for details.
Anthropic has released Remote Control for Claude Code. It lets you start a terminal task on your computer and continue it from your phone. You link devices by scanning a QR code in the terminal. It is now available for Max plans only. Check Anthropic’s announcement for details.
OpenAI has added WebSocket mode for the Responses API. This helps long agent workflows, especially with many tool calls. It keeps the connection open. Each time, you just send the new input and the id of the last response. Check OpenAI’s docs for more details.
OSS Highlight of the Week
This week we are featuring ZeroClaw. It is a runtime operating system for agentic workflows. It abstracts models, tools, memory, and execution so agents can be built once and run anywhere. It can run on low cost hardware. It is far lighter than many other agent runtimes. Check the ZeroClaw GitHub repo for details.
Good to know
Anthropic is keeping Claude Opus 3 available even after retirement. Paid users can still access it, and API access is available by request. Opus 3 will also continue writing weekly essays in Claude’s Corner for the next three months. The essays will be reviewed before publishing, but not edited. Check Anthropic’s announcement for details.
Google is deprecating Gemini 3 Pro Preview on Gemini API and Google AI Studio on March 9, 2026. If you are using the latest alias, it will switch to Gemini 3.1 Pro Preview. Check your integrations and migrate in time to avoid service disruption.
Perplexity has partnered with Samsung to bring Perplexity into the upcoming Galaxy S26. It will be built in as a system level AI with its own wake word, “Hey Plex.” Check Perplexity’s announcement for details.
We have a new auto memory feature in Claude. Claude can now remember useful context like debugging patterns and preferred approaches across chats. So we do not have to repeat the same details every time. Check the twitter announcement for details.
Notable FYIs
Anthropic has added Scheduled Tasks to Cowork. Claude can now run recurring tasks at specific times, like a morning brief, weekly sheet updates, or Friday presentations. Check Anthropic’s announcement on Scheduled Tasks for details.
If you do web scraping, Scrapling is worth a look. It is a Python framework that adapts to site changes. It helps handle anti bot pages. It can scale from one request to full crawls. Check the Scrapling GitHub repo for details.
If you work with agents, you know skills matter. Antigravity Awesome Skills is a community repo with ready to use skills for Claude Code, Antigravity, Cursor, and more. Review each skill before you install it on your agent.
We build Agentic apps. But shipping them is not easy. Plano is a model native proxy server and a universal data plane for Agentic apps. It handles routing, guardrails, and observability, so you can focus on the agent logic. Check the Plano GitHub repo for details.
It is not always necessary to fine tune Agents to improve performance. Memento is a framework that helps agents learn. It saves what worked and what failed. It then uses that memory on new tasks. It follows a planner and executor loop. Check the Memento GitHub repo for 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.
Congrats on edition 100 - genuinley a milestone worth celebrating. The Perplexity Computer announcement keeps sticking in my head; 19 models for different task types feels like the first time someone actually architected multi-model at the infrastructure layer rather than just API-switching in the background. The ZeroClaw angle on low-cost hardware is underrated tho - the real agentic unlock might be edge compute more than cloud scale. Consistenly strong newsletter.