[Dev Catch Up # 74] - ChatGPT's study mode, How to build SDKs, Nine VSCode Extensions for easy coding, Dyad, mcp-ui, Gemini's Deep Think, Anthropic's skillJar Courses, HRM 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 74th 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
OpenAI introduced study mode in ChatGPT. It helps you solve problem step by step instead of getting direct answers. It works in an interactive way. Check OpenAI’s Post on how to get started, key features, ways to use study mode, limitations and other details.
Google added Deep Think mode to Gemini App for Ultra plan users. Deep Think uses parallel thinking techniques. It combines different ideas over time, before arriving at the best answer. Read Google’s blog to understand how Deep Think works, and how to use it in Gemini App and more.
As a developer, I have used SDKs to build applications. This Substack post explains what SDKs are and how to build them. Quentin Pradet, software engineer at Elastic and maintainer of Apache Libcloud, shares practical insights from his experience about working with SDKs. Read the post to learn more about building and maintaining SDKs.
Context Engineering is becoming popular. We covered many articles on this topic in our previous editions. This Substack Post on Context Engineering explains about difference between prompt and context engineering and shows the approach in action. Read the post to learn more.
OSS Highlight of the Week
This week we're featuring Dyad, a free, local, open-source AI app builder. It works as an alternative to tools like v0, lovable and Bolt. You can build AI apps using local models through Ollama and LM Studio or connect to APIs like OpenAI and Claude. Check the GitHub repo if that sounds interesting.
Good to know
Code with deep nesting is harder to read and understand. This post shows how early returns can simplify your functions and make code easier to read and maintain. Read the Substack post to learn how to avoid deep nesting with this simple pattern.
Many of us know Heroku as a go to platform for deploying applications. It recently faced a 23-hour outage caused by a routine Ubuntu update. This post covers what went wrong, why the response was delayed, and how it differed from Heroku’s earlier outage. Check the Substack post to know more details.
If you are using an MCP server, check out mcp-ui. It provides web components to the MCP Server. So that Server can deliver UI which can be rendered by the client. This takes AI interaction to the next level. Check the GitHub repo for more details on mcp-ui.
If you are exploring data science, check Positron. It’s a free, data science IDE for writing code and exploring data. It combines IDE with data science tools. Check docs page to learn more about Positron IDE.
Notable FYIs
A software engineer who uses VS Code and Cursor shared a list of 9 helpful extensions. These tools improve search, refactoring, and make every day coding tasks faster and easier. Read this Substack post to know about those extensions and see how it can help on your daily workflow.
Anthropic has launched a separate learning page with free courses. It includes topics like Claude with Amazon Bedrock, Claude Code in action, advanced MCP, and more. Check the page to see all available courses.
Sapient has released the Hierarchical Reasoning Model (HRM). It is a brain inspired architecture for self-evolving, multi-level reasoning and adaptive planning. The full code is available on GitHub. Check the repo to learn how to set it up and run.
Anthropic introduced persona vectors to track and control model behavior. Their preventative steering approach helps the model to avoid unwanted behaviors during training. Check the Twitter thread for full details and read the research paper to know more about Persona Vectors.