How should you manage AI contributions to an open-source project? How do you measure the impact of using AI tools for development, and what generative features do end users want from a content management system? This week on the show, we speak with Thibaud Colas and Meagen Voss from Wagtail about the complex considerations software organizations are currently facing.

Meagen is the Wagtail community manager, and Thibaud is a member of the core team. We discuss the impact of AI on a popular open-source project. We reference multiple articles they’ve written about managing these challenges.

We dig into wrangling AI-assisted contributions to the project. They share how they research options for open-weight AI models and inference providers. We also discuss building AI features into the Wagtail CMS. Finally, we touch on the recent release of Wagtail 8.0.

Topics:

• 00:00:00 – Introduction
• 00:01:49 – What is Wagtail?
• 00:08:35 – Flood of AI submissions to open-source projects
• 00:11:38 – Guidance for AI-led contributions
• 00:20:43 – Security submissions
• 00:21:56 – Ethical decisions about internal use of AI
• 00:26:31 – Resurgence of web sites
• 00:27:41 – Measuring benefits and performance
• 00:32:49 – Models understanding your framework
• 00:38:05 – Video Course Spotlight
• 00:39:42 – Adding AI features to Wagtail for creators
• 00:48:47 – Balancing the use of AI internally and the cost of inference
• 00:55:29 – Advice for other open-source projects
• 01:00:00 – Wagtail 8.0
• 01:02:10 – What are you excited about in the world of Python?
• 01:04:10 – What do you want to learn next?
• 01:05:56 – How can people follow your work online?
• 01:06:43 – Thanks and goodbye

Show Links:

• Wagtail CMS - Django Content Management System
• Wagtail Space 2026 - Wagtail CMS
• Showcase - Wagtail CMS
• Guidance for AI-led contributions - Wagtail CMS
• What I learned from two days of hanging out with AI experts - Wagtail CMS
• AI saved me 10 minutes on a blog post: Was it worth it? - Wagtail CMS
• What AI tools get right and wrong with Wagtail - Wagtail CMS
• Comparing open weight AI models and providers - Wagtail CMS
• AI in the CMS: steering the ecosystem - Wagtail CMS
• The carbon footprint of Wagtail AI - Wagtail CMS
• Open source AI we use to work on Wagtail - Wagtail CMS
• Results of the 2026 Wagtail DX with AI survey - Wagtail CMS
• open-code-review: Secure, fast, efficient, battle-tested - hybrid architecture code review tool
• CMS with AI, not AI CMS: Wagtail 8.0’s new API
• Roadmap - Wagtail CMS
• Executive Director Search Extended to September 22 - Django Weblog
• git-worktree Documentation
• Wagtail newsletter
• Meagen Voss (@vossisboss@fosstodon.org) - Fosstodon
• Thibaud Colas (@thibaudcolas@fosstodon.org) - Fosstodon

Level up your Python skills with our expert-led courses:

• How to Set Up a Django Project
• Getting Started With Google Gemini CLI
• Using the Claude API in Python

Support the podcast & join our community of Pythonistas

if you want to check out all the things ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠torc.dev⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ has going on head to ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠linktr.ee/taylordesseyn⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ for more information on how to get plugged in!

Comment

Last week I attended the iPhone Duo Group Lab, and while none of my questions made it to the panel, I came out with a few insights. I was as interested in the questions other developers asked as in Apple’s answers. It’s clear to me that the new hardware sparked a lot of imagination. Developers are eager to explore creative app ideas that the new device enables.

For example, are third-party StandBy apps that use the Tent pose, like Apple’s glowing alarm clock, supported? Or can apps receive touch input from both the inner and outer displays simultaneously when in the Laptop pose (for hot-seat gaming perhaps)? The answer to both questions seems to be ‘No’, at least for now as the panelists acknowledged that they themselves don’t yet know all the ways people will physically use Duo.

Then there’s the hinge. Can the hinge angle be creatively used as input? For this, the panelists suggested a hilariously brilliant idea of a virtual band app, where the hinge can be used as an electric guitar whammy bar, or an accordion bellows, or as a piano when in the Laptop pose. Honestly, somebody should just build this!

Speaking of the hinge angle change, in the middle of the Lab, one of the panelists mentioned that the new Device Hub comes with a realistic model of Duo that can be freely rotated in 3D space, along with a hingle-angle slider. Wait, what?! Did we miss it in the latest Xcode 27.1 beta? Turns out it’s just hidden behind a feature flag, and folks already figured out how to enable it. Take it for a spin (literally), it’s really cool! I initially had mixed feelings about Device Hub, but seeing that Apple approaches it not merely as another simulator wrapper with menus, but as an immersive device simulation environment, it starts to make a lot more sense.

Wrapping up thoughts on the iPhone Duo Group Lab, for all the creativity and imagination reflected in the community questions, the overarching message from Apple was to treat Duo as an iPhone that folds and resizes, not as a separate platform. The new hardware creates opportunities, but developers shouldn’t try to invent a gimmick for every pose and hinge angle.

But there’s no harm in having a bit of fun, right? 😉

– Alex Ozun

Sponsored Link

GitHub-hosted macOS runners love to queue, and Xcode builds take forever once they finally start. WarpBuild runs your existing workflows on dedicated M4 Pro machines with the latest Xcode up to 3x faster and at half the price. Switching takes just a single line of YAML. Get faster builds with $50 in free credits.

Tools

I love SF Symbols! And I love unapologetically native macOS apps (aka Mac-assed apps). Amy Worrall combines both of these beautifully in McKinley.

Creating custom SF Symbols and syncing them with Xcode’s asset catalogue is still more fiddly than it should be, which makes McKinley a compelling all-in-one tool, check it out!

The Nil Coalescing team has been a trusted source of SwiftUI and Swift knowledge in Apple Community for years now, so it only makes sense for them to distill and package that knowledge into a tool for AI agents. I like that it’s not just another skill, but an MCP-enabled tool that can plug into an agentic harness for better experience and improved token efficiency.

Code

Serg Dort addresses an increasingly common problem for Apple developers: switching between agentic and hands-on workflows while sharing the same Xcode build environment.

The article shows how to configure Xcode’s compilation cache so that a handoff from a headless AI agent to a human in Xcode still benefits from incremental build times.

Harking back to the iPhone Duo Tech Talks and Labs, Apple clearly positions ArrangementView as one of the heavy lifters for iPhone Duo adoption, so it’s worth taking the time to understand it properly.

What I found particularly valuable in this article by Fatbobman is that it focuses on the semantics behind ArrangementView, helping us build intuition rather than merely learn a new API. The key insight is that ArrangementView declares the relationship between two views rather than geometry. As the title suggests, it pushes us to think before we arrange. I love this message!

It’s that time of year when we get tons of new APIs that aren’t available on previous OS versions. Backporting is a common way of dealing with this challenge. There are many ways to approach backporting, but I like the type-safe, idiomatic, and lintable Backport API that Majid proposes in this article. I also like how it demonstrates the new anyAppleOS availability qualifier!

Figuring out how to deterministically test Unstructured Concurrent Swift code is no mean task (trust me, I tried)!.

Michal aptly points out how both humans and even the latest models often resort to sprinkling sleep calls into tests, hoping to shoehorn racing tasks into some semblance of an ordered queue. We all know how this usually ends. Michal outlines a better solution based on the Test Spy technique, where objects performing racing tasks can be swapped for controllable test doubles that can produce values or advance time deterministically in tests. Neat!

And finally...

macOS 27 comes with the Liquid Glass transparency slider. But what if we could keep pulling the slider out of the settings window? 🤔