AI, Agents & Human Judgment
How many app issues fell off your radar this week?
Mobile teams deal with a firehose of incoming issues: regressions from QA, beta feedback, App Store reviews, crashes, etc. AI code makes managing this even harder. Triage by Runway pulls every issue into one inbox, de-dupes, assigns, and tracks fixes so nothing is missed. See how it works
This message is brought to you by a sponsor who helps keep this content free for everyone. If you have a moment, check them out. Your support means a lot!
Welcome to issue #80 of the iOS Coffee Break Newsletter 📬.
Lately, I have been thinking a lot about how AI is changing the way we build software, but also how it changes the way we feel about our work.
For this community edition, I picked six articles that explore that change from different perspectives: building an iOS app in six hours, improving agentic workflows, keeping humans involved in important decisions and even designing better technical interviews.
Some are practical, others are more personal, but all of them made me stop and think about how I want to use these tools in my own work.
Have a great week ahead 🤎
CURATED FROM THE COMMUNITY
Xcode Projects for the Command Line and Coding Agents
Artem explores Xcode 27.2's new JSON project format and the command-line tools Apple introduced to create, inspect and edit Xcode projects.
I have always found project files to be one of the most fragile parts of iOS automation, especially when coding agents are involved. A readable format, smaller diffs and an official CLI feel like a meaningful step forward. The tools are still in beta, but this is definitely something I want to try.
Building an iOS app in 6 hours with Firebase and Antigravity
Marina and Tim share how they built Mixtape, an iOS app that lets friends exchange song links across Apple Music and Spotify, in about six hours.
What I liked most is that this is not another story where one prompt magically creates a finished app. The agent helped them move quickly, but they still had to review the code, replace a failed AI approach with real APIs and test everything repeatedly.
5 stages of (developer) grief
Costa reflects on the last few years of AI-assisted development through the five stages of grief: denial, anger, bargaining, depression, and acceptance.
This one made me laugh, but it also felt uncomfortably familiar. Becoming more productive does not automatically replace the joy we get from writing code and thinking through a problem. I suspect many developers will relate to this one.
Agentic Engineering: 11 Tips for iOS Development with AI
Christian shares eleven practical lessons from using coding agents on production iOS projects.
There is a lot here that matches my own experience: keep the context focused, plan larger changes, use logs for debugging and keep pull requests small. The key point is to review strategically, paying closer attention to architecture, business logic and security-sensitive code.
Using Jev to Help a Human Decide
Alejandro explains how he uses a classifier model inside a merge gate that helps his team decide what caused a failing CI run.
What stood out to me is how small and focused the model's job is. Regular code gathers the evidence, the classifier suggests the likely cause and a human makes the final decision. It is a great reminder that not every AI feature needs to become a fully autonomous agent.
A Tech Interview That Doesn't Suck
This is not a recent article, but it is one I revisit from time to time. Jacob proposes replacing LeetCode-style interviews with a small, imperfect codebase that the candidate and interviewer improve together.
With AI making take-home exercises harder to evaluate, the idea feels even more relevant today: make the interview resemble the job and pay attention to how the candidate thinks.
Have a great week ahead 🤎
