Building the internet’s next infrastructure layer | Cloudflare's Brendan Irvine-Broque

**Andrew Zigler** (0:06)
Welcome to Dev Interrupted. I'm your host, Andrew Zigler.

**Ben Lloyd Pearson** (0:10)
And I'm your host, Ben Lloyd Pearson.

**Andrew Zigler** (0:13)
This week, we're diving into MCP servers at scale, with Brendan Irvine-Broque, Director of Product at Cloudflare. But first, we're going to be diving into some news that caught our eye this week, and we got to start with a big one. Dora 2025 is here. Ben, did this catch your attention at all in the last week?

**Ben Lloyd Pearson** (0:32)
Well, considering that I was at the big event this week, like Enterprise Technology Leadership Summit, where Dora always makes a big announcement around this report, yes, it did catch my attention.

**Andrew Zigler** (0:43)
Nice.

**Ben Lloyd Pearson** (0:43)
In fact, yeah, I did get to meet some of the people from the team, including Nathan Harvey, who's been one of the big voices in the Dora community for quite a long time now. For anyone that's not familiar with this event, it's Gene Kim's annual event for engineering leaders. Gene Kim's an author. His latest book is Vibe Coding. Many people know him for The Phoenix Project. He's also a past guest here on Dev Interrupted. But it was a really fun event. Lots of conversations around Vibe Coding. It was really cool seeing a bunch of engineering leaders get introduced to Vibe Coding for the first time. Particularly people who haven't coded in a while, learning these new tools and getting really excited about coding for the first time in quite a while. And it was really nice to see that being celebrated. But then also a lot of discussions about MCPs as well. So we've been building MCPs at LinearB. And it's really nice with this emerging technology just to share strategies around what it takes to bring these stuff to market. But yeah, so the Dora Report, really exciting stuff in the report this year. They made a lot of big changes. And the headline that I think has really been making the rounds, and Nathan Harvey brought this up, that even CNN covered this, that 90% of software teams now use software on a daily basis.
And in his talk, Harvey mentioned that because of how widespread AI usage is, they're effectively just going to assume that it is ubiquitous now, and that question is probably not even going to be on future reports anymore. So pretty interesting to just think that AI is now just ubiquitous within software development life. And I actually fully expect we'll dive into this report quite a bit more in the coming weeks. I mean, it's 140 pages, so I've only barely scratched the surface on it. But there are a few early insights that I think our audience would love to hear about already. And really the biggest takeaway that I think everyone should learn from this is that AI is an amplifier both for the good and the bad. So if your organization has really strong engineering practices and good decision-making processes, AI enables you to leverage those benefits even more. But on the other hand, if you have poor internal practices, bad engineering infrastructure, AI is going to make your problem worse. So it's not even like the benefits of AI will be smaller. It's AI will actually make your organization worse. The thing that really illustrates this is within the report, there's a whole bunch of spider charts on one page. It looks a little overwhelming at first, but if you really take a moment to just understand what you're looking at, it's got a lot of really interesting insights. If you just look at two different types of organizations within those spider charts, teams that have foundational challenges, so they surveyed a bunch of people and asked, what is your organization like? They classified them based on the responses that the participants gave. They found organizations that have foundational challenges, they have lots of problems with their engineering. Then they have organizations on the other hand, who have super harmonious high achievement practices. The teams that had foundational challenges, when they adopted AI, they had higher rates of burnout, more friction, they had more software instability and worse product in engineering performance. But then that, on the other hand, that team of harmonious high achievers was practically the exact opposite. Like it was, it's very stark how different those two organizations were. And, you know, the report also unveiled what the Dora community is calling the AI capabilities model. So they've identified seven core capabilities that enable organizations to more successfully leverage AI. And what I really like about this capabilities model is really just how well it aligns with the conversations that we're having here on Dev Interrupted. So it's things like having healthy data ecosystems that are accessible to your AI system, maintaining quality platforms, having engineering best practices like working in small batches and strong version control practices. Like if you've listened to Dev Interrupted enough, you've almost certainly heard at least one or more guests talk about these exact things in the context of AI. So like I mentioned, it's a really long report. There's a lot to unpack. We're going to keep covering this here on Dev Interrupted. We'll probably have a sub-stack article about it soon enough. And I would really love to get Nathan Harvey back on to the show again, just to help break down some of the details of it at some point in the future too. So stay tuned for that because I think we'll make it happen.