[Dev Catch Up # 85] - OpenAI's Atlas, What caused the large AWS outage, GPU Glossary, Claude Code on Web, Google Maps Tool in Gemini API, DevPush, DeepSeek-OCR, Firecrawl's Open Agent builder & 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 85th 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 8400+ developers to hear stories from Open source and technology.
Must Read
OpenAI has entered the AI browser war with Atlas. It’s a browser with ChatGPT built into it. You get ChatGPT help right where you’re browsing. In Atlas, you can now ask ChatGPT to take actions like filling forms, booking appointments, and more. Check OpenAI’s post on ATLAS to know more interesting details.
Anthropic has launched Claude Code on the web. Now you can run coding tasks directly from your browser without opening a terminal. You can connect GitHub repos, run tasks across different repositories. It can also create PRs with clear summaries. Check it out if you’re a Pro or Max user.
Last Monday was AWS’s rough day. The us-east-1 region went down for 14 hours. Slack, Zoom, and thousands of websites stopped working. The root cause was DynamoDB’s DNS failure. If you are interested to know what happened in detail, Check Gergely ‘s post on “What caused the large AWS outage”.
We all use Docker containers for deployment, but few understand how they work internally. This post talks about the three core components of containers. Read this Substack post if you want to understand how containers work.
OSS Highlight of the Week
This week we’re featuring DevPush, a self-hosted open-source alternative to Vercel and Netlify. You install it on your server, connect your GitHub repos, and when you push code, it automatically builds and deploys your apps with zero downtime. It works with Python, Node.js, PHP, or anything that runs on Docker. Check the DevPush Github repo for more details.
Good to know
Most Node.js apps use try-catch for error handling, which makes errors invisible in function signatures. This post explains the Result Pattern where functions return success or failure objects instead of throwing exceptions. Worth reading if you want cleaner error handling in Node.js.
Simon Willison talks about running Claude Code with
--dangerously-skip-permissions. It gives maximum freedom but comes with prompt injection risks. The solution is sandboxing with controlled network to prevent data leaks. Check the Simon Willison post on Living Dangerously with Claude.I came across this tool that converts ebooks to audiobooks. It supports voice cloning and over 1,100 languages. Check this ebook2audiobook GitHub Repo to learn more details.
GPU is a popular word in the AI space. If you are still new and want to learn all the GPU-related terms, check this GPU glossary page by Modal.
Last week, we covered Claude’s Agent Skills. You upload a skill zip file to Claude, so it works better with specific domains. This week, I came across a tool that creates these skills automatically. Skill Seekers scrapes any documentation, website and converts it into a ready-to-use zip file for Claude. Check the Skill Seekers GitHub repo for details.
Notable FYIs
DeepSeek released an OCR model. This is not a regular OCR model. It explores whether more text tokens can be converted to fewer visual tokens with high accuracy. This way, LLMs can process large documents with fewer visual tokens. Check the DeepSeek-OCR GitHub repo for more details.
Google has launched Google Maps tool for Gemini API. Developers can now connect Gemini with over 250 million places. Now we can build location-aware apps for travel planning, or local recommendations with accurate and real-time information. Check Google’s blog on grounding with Google Maps for details.
Google’s Notebook LM is useful, but your data goes to Google. Open Notebook is a self-hosted alternative for Notebook LM. It supports PDFs, videos, audio, web pages. It can also generate multi-speaker podcasts. Check Open Notebook Github repo for more details.
Visual agent builders are rising. Firecrawl has released Open Agent Builder. It’s a drag-and-drop tool to build web scraping pipelines with real-time execution. Check the Open Agent Builder GitHub repo for installation and other details.
I came across a podcast where Andrej Karpathy explains why AGI is still a decade away.. He talks about where today’s agents fall short, especially in continual learning and multimodality. Check the podcast for his insights on AGI.
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.
Excellent analysis! The Claude Code web capabilites really stood out. Direct browser coding with GitHub integration is a genuinely thoughtful step.
Great comprehensive roundup of this week's dev highlights! The Google Maps tool for Gemini API is particularly exciting - grounding LLMs with 250 million real places solves a major problem with AI hallucinations around location data. Combining that with Gemini's spatial reasoning could unlock genuinely useful location-aware applications beyond what we've seen from traditional map APIs. The OpenAI Atlas browser announcement is also significant, though I'm curious whether browsers are the right interface or if we'll end up with more specializd AI-first experiences. And that DevPush project looks like a really solid self-hosted alternative to Vercel - thanks for highlighting it!