Not only Swift

In this live coding session, I’m building a “Smart Tagging” feature for Sofia, my personal knowledge management app. I’ll analyze the pros and cons of on-device vs. cloud implementation and then live-code a solution.

Want to guess which one I choose? Cast your vote on this poll, and then join the livestream to see which one I actually choose, and why!

Swift’s AsyncStream makes consuming asynchronous streams of values easy - or so you might think. But it can be dangerously easy to get it wrong.

If your codebase contains code similar to the following, you should read Michael’s article to learn how to avoid the pitfalls:

Task {
  for await value in stream {
    handle(value)
  }
}

Natalia explains the nuances of initializing @Observable classes in SwiftUI views. She covers why storing them in @State is crucial for lifecycle management and how to use .task to defer initialization for performance.

Antoine van der Lee and Omar Elsayed have created a comprehensive agent skill for SwiftUI best practices. It distills practical SwiftUI best practices into actionable, concise references for agents and code review workflows.

In this video, I explain Function Calling (also known as tool use)—the missing piece that allows generative AI models to interact with the real world. Learn how to use it in your own apps with Firebase AI Logic.

Warning: don’t watch this when hungry - there’s a parmigiana recipe in there that will make you crave it immediately.

Have you ever ended a coding session with your coding agent and thought “ah, I really should create a skill for this?”

Well, now you can!

Jediah created a skill that helps you capture insights from your current coding session as a skill.

Apple’s Foundation Models allow you to perform on-device AI tasks. While this has the advantage of privacy and speed, it also has the disadvantage that the models are limited in their capabilities - specifically in their context window.

This requires careful prompt design. In this article, Artem goes over a comprehensive list of best practices for prompting Apple’s Foundation Models, focusing on limited context windows and specific model characteristics.

Vibe coding a todo-list app? Sure! But how about storing the data in GitHub issues? Martin Woodward decided to find out.

The source code is available on GitHub.

One of Antigravity’s unique features is its agent manager. But how does it work, and how can you use it to perform multiple coding tasks in parallel?

Mete explains it in this post.

Luca interviews Gonçalo Silva, CTO of Doist, about how they built Todoist. They discuss the Doist philosophy: small, bootstrapped, and distributed, as well as async-first communication and how AI impacts engineering work.

Is Git unsuitable for agentic development, asks Pedro. Instead of using branches, maybe the concept of a “session” is more suitable. And what happens with CI/CD, and who pays for any tokens used in an open source contribution?

Jules is Google’s autonomous coding agent - and given its tagline, “Jules does coding tasks you don’t want to do”, I should probably use it more often!

Google PM Kathy created a native macOS companion app for Jules that brings better session control and desktop integration.

I love that we’re seeing more and more native apps for AI agents :-)

Git worktrees are a great feature, and super useful for quickly switching to a different branch when you want to do some quick fixes or experiments.

Daniel created Trees, a macOS menu bar app that simplifies managing and switching between Git worktrees.

A couple of conferences have updated their dates, and a few more have just been announce, like Agentic Conf Hamburg in March.

Here is the updated list of conferences for 2026: