Not only Swift

Can you build entire SDKs using Agentic Engineering? That’s what I tried to find out in this live stream.

I was so inspired by all the articles about Agentic Engineering in this newsletter that I decided to give it a try myself, and why not do it on stream, so we can all learn from my experiences?

So, grab a cup of coffee - or maybe a hearty snack, as this stream is about Pickle Rick - kick back, and watch as I try to build a Swift version of Mozilla’s Readability SDK using Gemini and Pickle Rick!

Pickle Rick, by Gal, is an extension for Gemini CLI that implements a development methodology based on continuous AI agent loops. It’s similar to the “Ralph Wiggum” technique, but with a couple of twists.

It will first put together a product requirements document (PRD), and then split the work into smaller tasks, which it will then execute one by one with sub agents.

To see it in action, check out the livestream where I used it to build a Swift version of Mozilla’s Readability SDK.

In issue 94, I shared an article in which an OpenAI team describes how they implemented an entire platform without writing a single line of code and instead vibe-coded the entire thing, including all CI/CD.

Here is another article of a single engineer rebuilding an entire framework (Next.js) in just a couple of days with the help of LLMs.

If you have good specs and tests, modernising your app / SDK / system is no longer a pipe dream.

Here is another article demonstrating how to build an entire product with an autonomous team of agents.

Wilson Lin describes how they went from “prodding an agent” to continue working on a task (which failed) to an autonomously running system of thousands of agents almost entirely runnable without human intervention.

Margaret-Anne takes a step back to look at the broader implications of agentic coding. In her article, she explores how generative AI is shifting our primary concern away from traditional technical debt and towards “cognitive debt”—the mental overhead required to understand and maintain AI-generated codebases.

How do you measure real-world battery drain of your app? Using the Simulator is not an option, so you need to do it on a real device. But - walking around with an iPhone tethered to your Mac running Instruments is not always practical…

Thankfully, there is a much easier way - Vladimir explains how it works.

In issue 95, I wrote about CLIs gaining in popularity. Justin (creator of the Google Workspace CLI) shares some insights into building CLIs that work well for humans and agents alike.

Yes, you should write your CLIs for different audiences, as the DX is different: humans prefer flags, agents love JSON.

Greenlight is an open-source pre-submission compliance scanner for the Apple App Store. It runs completely offline, and performs static checks against App Store policy violations. This article describes how it works in more detail. It also includes a skill, so you can invoke it from your favourite coding agent.

App Store Connect CLI by Rudrank Riyam is a command line tool to automate interactions with App Store Connect, and people really seem to love it. Rudrank recently added a skill for it, which makes using it even easier with AI coding agents.

Now, if you’re an Android developer and you look at your iOS colleagues, you might be a bit jealous of the tools they have at their disposal.

But you don’t have to be! Dalton has built gpd – Google Play Developer CLI, the equivalent of App Store Connect CLI for the Google Play Console. It supports everything you need to manage your app releases, including beta tracks, staged rollouts, and more. And, thanks to its JSON-first output, it’s agent-friendly.

Here is your updated list of conferences for iOS, Swift, and AI.