[Dev Catch Up #69] - Gemini CLI, Anti Clean Code, API Versioning, Unregistry, OpenAI API Cost Hack, Dead Code cleanup for Go, o3-deep-research API, mcp-scan, JavaScript Full E-book, 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 69th 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.
Dev Shorts is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.
Must Read
Google launched Gemini CLI, an open-source terminal AI agent, competing with Claude Code. It offers free usage limits and includes features like MCP integration, Google Search and more. Read Google's announcement for more details on Gemini CLI.
I've developed APIs for many projects, and I know how important API versioning is. This comprehensive guide covers API versioning strategies, implementation methods, and lifecycle management. Check the full guide for practical examples and tools to manage API versions.
Most developers aim for clean code but often end up with chaotic codebases that are hard to maintain. This guide explains the F.L.U.I.D. trap, when code becomes Fragile, Lax, Untested, Insecure, and Disorganized and provides actionable steps to address it. Check the full post for solutions to fix them.
OpenAI shipped a couple of new capabilities in their API. The update includes deep research models like o3-deep-research now available in the API, webhooks for async tasks, and web search capabilities for reasoning models. Check the OpenAI community post for more details.
OSS Highlight of the Week
This week we're featuring Unregistry, a tool that simplifies transferring Docker images to remote servers. Instead of dealing with Docker Hub where code goes public, self-hosted registries where you pay, or rebuilding remotely, it lets you push images straight to remote servers over SSH. Check out the GitHub repo if you want simpler Docker image transfers without the usual overhead.
Good to know
Dead code is common in all languages, and Go is no exception. The Go team released a dead code tool that finds unreachable functions by analyzing execution from main. Check the full post to know how to remove dead code in Go.
Fine-tuning is popular in the LLM space, but many people still don’t know the complete approach. This guide covers complete details about fine-tuning including base model selection, data preparation, and framework selection, etc. Check Heavybit's guide to know complete details about fine-tuning.
Database queries can sometimes run forever, but proper indexing can turn those slow searches into faster lookups. ByteByteGo breaks down different index types and when to use each one for faster performance. Check their guide for indexing strategies that actually work.
Google released Gemma 3n, an AI model designed for on-device applications that can handle text, audio, and images. It features MatFormer architecture that provides efficient performance on various device capabilities. Check Google's guide for technical details and implementation.
Notable FYIs
Developers can now build, host, and share AI-powered apps directly in the Claude App. When someone uses the app, they authenticate with their existing Claude account. No one needs to manage API keys. Check Anthropic's announcement for full details.
JavaScript developers looking for a comprehensive reference, should check out "Exploring JavaScript," a free online book covering basic to advanced JavaScript topics. Check the full book for detailed JavaScript learning.
If you're using OpenAI API for audio transcription, here's the cost hack. OpenAI charges per minute, so make your input audio minutes shorter. Running files at 2x or 3x speed through ffmpeg reduces costs by 33%. Check the full post for the script and implementation details.
People who are using vLLM should check out Nano-vLLM, a lightweight vLLM which delivers high inference speeds. Check the GitHub repo to learn more about Nano-vLLM and implementation details.
MCP servers are becoming popular, but they can introduce security vulnerabilities like prompt injections and tool poisoning attacks. MCP-Scan is a security tool that scans your installed MCP servers for common vulnerabilities. Check the GitHub repo if you're using MCP servers and want to audit their security.
You either treat open source like a serious competitive edge or you treat it like a weekend hobby and only one of those builds real momentum.