[Dev Catch Up # 73] - Qwen3-Coder, Deployment Strategies, Task Master AI, Better Error Handling, s3grep, LLM Evals, How Anthropic teams use Claude Code, Refactoring methods, ACI. Dev, 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 73rd 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 8000+ developers to hear stories from Open source and technology.
Must Read
Alibaba’s new Qwen3-Coder is released. It supports massive context and feels comparable with Claude Sonnet 4. You can use this with Claude Code and Cline. Check Qwen 3 Coder release notes for more details.
Claude Code now lets you create teams of custom agents. The
/agents
command spins up agents that can collaborate to complete tasks for the user. Check the Twitter thread to see how it is being used.Deployment is one of the important stages in software lifecycle. A well-chosen deployment strategy ensures a great user experience. This post covers about six deployment strategies like Blue-Green deployment, Canary, Shadow deployment and more. Read this substack post for more details,
If you’re exploring evals in LLMs, this page could help you with more clarity. It answers what evals actually measure, how to design them, and why most setups fail. Helpful if you're working on LLM Evals. Read this page to know more about LLM Evals.
OSS Highlight of the Week
This week we are featuring Task Master AI. Task Master is like a project manager for your AI. It helps you organize, track, and ship tasks with simple commands. It’s open source, API friendly, and works inside your editor. Install it as a package or add as MCP config to your IDE to get started. Check the GitHub repo for IDE setup instructions.
Good to know
A study found that developers using AI tools like Cursor were 19 percent slower overall. Even though coding felt faster, delays came from waiting on AI, reviewing output, and context switching. Read this post to know where AI tools help and where they slow you down.
Everyone is building AI Agents, but this post argues for a better alternative. Instead of creating agents that do everything for you, the author recommends building systems that help you think clearly and make better decisions yourself. A good read on a different perspective of utilizing AI in real time.
API retries help you handle failures, but they can also create duplicate events. This post explains why duplicates happen and shares four simple ways to avoid them in your application. Check this substack post to know more about API retries.
Using exceptions to control the flow of your code is a common mistake. This post explains why it makes your code confusing and suggests clearer ways to handle errors. Read this post to learn how to avoid this problem and improve your error handling.
If you have ever wondered how S3 keeps data consistent at large scale, this post explains it clearly. Achieving consistency usually means sacrificing speed or availability, but S3 handles it smartly. Read this post if you want to understand more.
Notable FYIs
We are all figuring out what kind of workflow actually works with AI. This post shares a developer’s real experience building with Claude. With no big frameworks, just four simple docs that helped retain memory, structure, and progress. Check this post if you want to know more details
Anthropic shared how their teams actually use Claude Code internally. They use it every day for debugging, writing tests, documenting systems, and building full-stack projects. Check Anthropic’s post to see how they use Claude Code in real tasks.
Long parameter list is a sign your method is doing too much. This post explains why you should refactor such methods using objects or value classes. You can read this post if you're trying to write cleaner, more maintainable code.
I found a useful tool called s3grep. It helps you search text directly inside S3 buckets without downloading files. Handy if you often work with logs or big data in the cloud. Check its GitHub repo for more details.
If you're building agents, ACI.dev is worth checking out. It connects 600 plus tools with unified auth and permissions, all accessible through a single MCP server or Python SDK. Helps you skip writing custom OAuth flows or API clients. Check the GitHub repo for more details.