The B2B Podcast Index
Index
All categories
MarketingSalesSaaSFinanceHROpsLeadershipCustomer SuccessAI & DataProductStartups & FoundersRevOpsEngineering & DevTools
MethodologySubmit
Best of:MarketingSalesSaaSFinanceHROpsLeadershipCustomer SuccessAI & DataProductStartups & FoundersRevOpsEngineering & DevTools
An independent project byFame
SearchBest episodesGuestsInsightsMethodologySubmit a podcast
Index/Engineering & DevTools/Scaling DevTools
Scaling DevTools artwork

AI broke traditional infra - with Kyle from Depot.dev @ AIE Europe

Scaling DevTools · 2026-05-18 · 11 min

0:00--:--

Key moments - from our scoring

Substance score

57 / 100

Five dimensions, 20 points each

Insight Density12 / 20
Originality10 / 20
Guest Caliber15 / 20
Specificity & Evidence9 / 20
Conversational Craft11 / 20

Depot CI, launched two weeks before this conversation, represents a fundamental rethinking of continuous integration for an AI-driven world. Kyle explains that traditional tools like GitHub Actions, Git, and existing CI/CD platforms weren't built for scenarios where autonomous agents generate 10x code volume, overwhelming systems designed around human-centric workflows. The core problem: Depot previously owned only 30% of the pipeline with GitHub Action runners; the remaining 70% stayed with GitHub, limiting their ability to optimize. Depot CI changes this by taking 100% control of the pipeline with their own engine and sandbox technology.

The conversation centers on infrastructure disruption. As code generation accelerates exponentially - Kyle cites DevTools growth rates doubling in just three months - traditional collaboration paradigms fail. Source control, pull requests, testing, and code review become bottlenecks. Depot's answer involves not just faster CI, but exploring open-source plugin systems to support CircleCI and Jenkins syntaxes, and eventually extending optimization down to the source control layer itself. The broader vision is enabling five-person teams to operate like 500-person teams through agent augmentation, requiring fundamentally different infrastructure primitives than what existed for the past two decades.

Key takeaways

  • →Depot CI controls 100% of the pipeline end-to-end, versus the previous 30% with GitHub Action runners, enabling full optimization for AI-generated code volumes.
  • →DevTools growth rates have doubled in the past three months entirely due to AI agents and code generation explosion, making infrastructure a critical bottleneck.
  • →Traditional tools built around human-centric workflows - Git, GitHub Actions, PR review systems - are crumbling under 10x code volume and can't scale to teams augmented with autonomous agents.
  • →Depot is exploring open-source plugin systems to support multiple CI syntaxes (CircleCI, Jenkins) and plans to extend infrastructure optimization beyond CI into source control and code review layers.
  • →No existing vendor has solved the collaboration paradigm needed for AI-augmented teams; the industry is collectively figuring out how to maintain code quality at agent-generated scale.

Guests

Kyle (Depot)

Topics in this episode

AI agentsSource controlGitHub Actionscode generationDepot CICI/CD infrastructureSandbox technologyPull requestsDevTools growthAgent-augmented engineering teams

Questions this episode answers

Why did Depot build their own CI engine instead of improving GitHub Action runners?

Depot's GitHub Action runners only controlled 30% of the pipeline while GitHub controlled 70%, limiting optimization. Depot CI takes 100% control of the pipeline with their own engine and sandbox technology, enabling full end-to-end performance for AI-generated code volume.

How much is DevTools growth accelerating due to AI?

DevTools growth rates have doubled in the past three months, and Kyle attributes literally all of that acceleration to AI agents and the explosion of code generation they're producing.

What infrastructure components are breaking under AI-generated code volume?

Source control (GitHub struggling with reliability), CI/CD, testing, and pull request systems are all crumbling under exponential code volume that traditional tools weren't designed to handle.

What collaboration paradigm does Depot think engineers and agents need?

Depot is exploring how five-person engineering teams augmented with agents can operate like 500-person teams, requiring new infrastructure primitives for code review, source control, and CI/CD that support agent-human collaboration at scale.

How will Depot CI support different CI/CD syntaxes?

Depot plans to build an open-source pluggable system for Depot CI so users can bring CircleCI, Jenkins, and other pipeline syntaxes, with community-led support for different platforms.

What our scoring noted

Our reviewer’s read on each dimension, with quotes from the episode.

Insight Density

12 / 20

The episode contains some substantive claims about infrastructure challenges in an AI-agents era (DevTools growth doubling, 10x code volume, existing tools breaking), but these are largely assertions without deep exploration or novel mechanisms. The conversation skims surfaces rather than drilling into root causes, timelines, or implementation specifics. Most insights boil down to 'old tools weren't built for this load' without the architectural or operational depth a technical founder would extract.

DevTools growth rate has more than doubled in the past three months
what do we do when there's 10,000 agents on a 10 person engineering team producing, like, 10 x the amount of code that now needs to be reviewed

Originality

10 / 20

The thesis - that AI agents are breaking legacy devtools infrastructure - is current but not contrarian or deeply original. The framing of CI/CD bottlenecks, GitHub reliability issues, and the need for agent-aware tooling are logical extrapolations of present trends rather than first-principles rethinking. The guest recycles common startup-founder language ('step function leap,' 'right infrastructure primitives') without articulating a genuinely novel perspective on how to solve these problems.

the volume of code has turned up exponentially, and we're trying to figure out, like, how do we manage the quality of it
we're starting to think a lot more about, like, what's the new paradigm for how engineers and agents need to collaborate

Guest Caliber

15 / 20

Kyle is founder of a reasonably established B2B DevTools startup (Depot) with apparent traction (100+ KubeCon conversations, recent product launch). He has operational visibility into customer pain and market signals. However, the conversation does not probe his depth as a builder - no discussion of architectural decisions, scaling challenges overcome, or hard-won lessons. He speaks more as a product marketer than as a practitioner who sweated implementation details.

I was just in a meeting this week here in London where somebody was asking me, like, how much of that can be attributed to, like, AI and agents? And then our response was, like, literally all of it
we're working on a lot of cool stuff. I think DPO CI is definitely in its early days

Specificity & Evidence

9 / 20

The episode is sparse on concrete examples and data. Claims like 'DevTools growth doubled' lack context (which companies? what baseline?), 'GitHub struggling with reliability' lacks incident details, and 'agents producing 10x code' is a scenario, not an observed metric. No named customer wins, no specific benchmarks, no metrics on Depot CI adoption post-launch. Mostly abstract framing without the grounded evidence that would validate the thesis.

the response has been off the charts. Everybody's moving, their GitHub actions workflows over to Depot CI
GitHub is struggling to keep one nine of reliability

Conversational Craft

11 / 20

The host asks open-ended questions and shows curiosity but rarely pushes back or probe deeper. When Kyle makes bold claims ('literally all [growth] is from AI'), the host doesn't ask for evidence, breakdown, or caveats. The follow-up 'how much can be attributed to AI?' is soft and accepts the answer without friction. Questions about Europe and future plans are warm and surface-level, missing opportunities to extract concrete operational details or test assumptions.

What you know, things are changing a lot. What are you, what are you seeing in, like, the way people are changing, how they write code, and how they deploy it?
how much of that can be attributed to, like, AI and agents?

Conversation analysis

Computed from the transcript - who did the talking, and the words that came up most.

Most-used words

depot14code8agents7source7github6tools6control6living6infrastructure5across5london5happening5answer5france5engineers4kyle4

Episode notes

In this episode, Kyle Galbraith from Depot shares the story behind building Depot CI and why traditional infrastructure is "crumbling" under the weight of AI-generated code. We discuss the shift from human-centric pipelines to agent-augmented workflows, the challenge of managing a 10x increase in code volume, and Kyle’s perspective on the rising tech hubs across Europe. Links: - Depot - Kyle Galbraith

Full transcript

11 min

Transcribed and scored by The B2B Podcast Index.

Hi, everyone. We're AI Engineers, and I'm joined today by Kyle from Depot. Hey, Kyle. Thanks for having me.

I'm excited to be here. This is, quite the event. I think this is the first one in Europe. And Yes.

Well, I think there was a, like, kind of There's one in here. Ours. Yeah. In your your neck of the woods.

Mhmm. So what are you, what are you excited about at the moment? What's, yeah. So we we just released Depot CI.

I think that was two weeks ago, and the response has been off the charts. Everybody's moving, their GitHub actions workflows over to Depot CI. So DevTools CI is our own CI engine that we've built from the ground up. Really, it's like the CI engine we've wanted from the very beginning.

And the best way to think about it is it supports different languages effectively. So the first syntax it speaks is GitHub Actions, obviously, because that's what we're most familiar with nowadays. But, like, we're building our own SDK for it so you'll be able to, like, write your own TypeScript code or Python code to, like, orchestrate your pipelines. The entire system is built on top of our own sandbox tech, and we're working on, like, exposing the lower level details of that.

But, yeah, really, like, being here at this conference and kinda, like, seeing the current state of coding agents, and, really, we've made this, like, step function leap in what agents can do. And we're starting to think a lot more about, like, what's the new paradigm for how engineers and agents need to collaborate and, like, what's the right tools and right infrastructure components that we need to do that at scale. Yeah. And I've seen you guys do getting a lot of attention.

I was watching one of Fia's videos that was, like, that Sunil shared, and I was watching it, and then, hey. I see. Yeah. Like, you guys popped up as well.

Yeah. Yeah. Yeah. We're definitely, it's wild to, like, be in this place now, like, three years later and be like, I was just at a conference.

I was at KubeCon, back in Amsterdam, and I think a 100 people, like, walked up to me and were like, oh, yeah. Like, I've heard of Debooted. I'm like, this thing or that thing or this other thing or, like, I've seen your stuff on Twitter or seen your stuff on LinkedIn. Yeah.

To, like, go through that journey as a startup and be at that state now with Depot is, is wild for sure. Yeah. One of the things I really wanted to ask you because I think, with Depot, you have, like, I think a lot of you have a really, like, strong view into people's development flow and, like, what they're actually doing, how they're building things. What you know, things are changing a lot.

What are you, what are you seeing in, like, the way people are changing, how they write code, and how they deploy it? And Yeah. I think it's kinda like saying water is wet, in 2026, but, like, DevTools growth rate has more than doubled in the past three months. And that to be clear, right, that's growth rate has doubled.

So it's the growth rate. Growing I I I'm making these numbers off. If you were growing at 10%, you're now growing at 20%. Yeah.

That's what yeah. And that's across everything. So that's across revenue. That's across product usage.

You know, I was just in a meeting this week here in London where somebody was asking me, like, how much of that can be attributed to, like, AI and agents? And then our response was, like, literally all of it. And, like, the explosion of code is real. It is happening.

And we're seeing it across all digital customers. Like, everybody is facing what has, like, always been the bottleneck, which is, like, really all of your infrastructure tools. So, like, source control, CICD, like, tests, pull requests, like, all of those things have always been a bottleneck in various different ways. And now they're just, like, crumbling under this new load, right, where it's like, what do we do when there's 10,000 agents on a 10 person engineering team producing, like, 10 x the amount of code that now needs to be reviewed?

What do we do when there's, like, 10 x the amount of code, like, flowing into source control and GitHub is struggling to keep one nine of reliability. Right? And all of these things are, like, ripe for disruption. Like, how do you handle this new world?

And nobody has the answer, by the way. Like, I don't have the answer. Git doesn't have the answer. Git Butler doesn't have the answer.

Like, everybody's trying to figure this out. But it's very clear that, like, the volume of code has turned up exponentially, and we're trying to figure out, like, how do we manage the quality of it? Like, how do we build solid infrastructure primitives, like source control underneath, NCI, ACD providers that can really, like, support that load at scale. Interesting.

Okay. And then how do you think like, what's what's going wrong for developers now, and what needs to go right? I think what's going wrong for developers right now is just really, like, the tools and systems that we've been using for the past twenty years aren't built for this new world that we're living in. Either, like, the fundamental, like, collaboration paradigm is, like, wrong with them because they were built around this concept of a human needs to be involved at every step, and that's just simply not true anymore, to like, the actual tools themselves can't handle this new load.

Right? They are literally falling over daily, as we sit today. And we need better tools. We need better primitives.

We need a different paradigm, and that's what we see with depot. That's why so many people were hungry to, like, switch to depot CI is because the existing status quo around GitHub Actions, even with Depot's own GitHub Action runners, like, we only owned, like, 30% of the pipeline with those runners. Right? Like, 70% is still with GitHub.

And so there's only so much we could do to, like, accelerate that and, like, make that stable. And bringing Depot c Depot CI online is really, like, us taking over a 100% of the pipeline. So, like, now it's our engine. It's our control plane.

It's everything. And people want that people want to like reliable tools. They want reliable infrastructure, that can also like serve not only their engineering team, but like all of the agents that are spinning up locally or inside of cloud environments and stuff like that. Okay.

And what are you excited about, for the future of deco? We're working on a lot of cool stuff. I think DPO CI is definitely in its early days. There's a lot of things that we wanna add to that.

We definitely want to explore, like, how can we do more things open source. So this idea of, like, Depot CI can support different syntaxes. Like, how can we make that a pluggable system that's open source so that if you wanna bring CircleCI pipelines or you wanna build bring your Jenkins pipelines or things like that, like, how can Depot CI support that in a community led way? Also, like, where can we go beyond this?

Right? How can we offer, like, the depot level of infrastructure performance all the way down to the source control layer? And so for us, like, we're thinking a lot about source control, CICD, obviously, with Depot CI, code review. But, ultimately, like, what is that collaboration paradigm that we need for the next five, ten, fifteen, twenty years, for a world where a team of five engineers is operating like a team of 500 engineers once you, like, augment them with agents at their side as well.

K. Amazing. One last side question. You're from The States originally, living in France.

We're here in London for Air Engineer Europe. What are you making of, like, the European tech scene and the differences, the the strengths, the weaknesses? Yeah. No.

It's a really good question. I get that question fairly frequently, as an expat living in France. The best way I can answer it is really, like, London feels like a hub for tech. I would say France less so, especially in the South Of France.

South Of France has its own, its own culture, its own speed. Sometimes there's some friction there for me personally of, like, I don't I don't like that slowness. But every time I come up here to London, it's like, yeah, this feels like a hub. And I think part of that is, like, historically, it's, like, been a financial hub as well.

And so that's carrying over into, like, becoming a tech hub as well. But, like, I've also been to Berlin and Barcelona and things like that and and Amsterdam recently as well. And those those places feel like hubs as well. Like, it feels like there's something rising up here in Europe.

And I think, like, it was presented maybe in today's keynote or yesterday's keynote of, like, no. No. No. Like, all of this, like, AI research and all of the, like, advancements we're making with LLMs is coming out of, like, research that's happening here.

Right? It's happening in Berlin. It's happening in London. It's happening, really, like, all across the EU and The UK that really feels like you're you're living in something that's it's gaining ground.

It's gaining traction. Sure. It might not, like, be a 100% like San Francisco, but San Francisco has its own hub of things that they're focused on, and it also has its own downsides as well. And so I think it's an exciting time to be living over here and, like, still be, living in this world of new technology.

Yeah. Amazing. Where where do you wanna send people? Yeah.

I mean, like, go check out devtools.dev. Obviously, I'm on Twitter as well, or X or whatever they call it nowadays. Just Kyle Galbraith.

And, yeah, check us out. Amazing. Cool. Thanks, Kyle.

Awesome. Thank you.

Related episodes across the Index

Other episodes covering the same guests and topics, from across The B2B Podcast Index.

  • Utilizing AI internally to iterate faster and empower smaller teams to upskill w/ Vivek Raghunathan #263The Engineering Leadership Podcast · on CI/CD infrastructure96 / 100
  • AI Agents, False Productivity, and the Sales Team Reset with Gabe LarsenMake It Happen Mondays · on AI agents91 / 100
  • Pricing in the Age of SaaSpocalypse | Emanuel MartoncaProductized Podcast · on AI agents89 / 100
  • Why a $1.2B exit felt like his biggest failure, and the customer-obsession thesis behind AgencyThe GTMnow Podcast · on AI agents86 / 100
  • Unresolved.cx - Resolution means something different at every company - Craig Stoss - KODIFUnresolved.cx · on AI agents84 / 100
  • AI-Accelerated Supply Chain Attacks with Mackenzie JacksonRunAs Radio · on GitHub Actions83 / 100

More from Scaling DevTools

All episodes →
  • Robby Russell on Oh My Zsh, Developer Experience, and Open Source70 / 100
  • Swyx and Louis Knight-Webb @ AI Engineer Europe: DevTools, code mode, and the future of AI engineering68 / 100
  • Acquiring a startup: what to do AFTER the deal closes - with Dan Moore from FusionAuth86 / 100
  • Joel Griffith from browserless: from GitHub issue to bootstrapped business76 / 100
  • Cloudflare devs @ AI Engineers Europe (Sunil Pai, Matt Carey & Thomas Ankcorn)77 / 100
Explore the best B2B Engineering & DevTools podcasts →
All Scaling DevTools episodes →