Skip to main content

Dan Manges

Nov 18, 2026

AdamHOST
11:40
Can you really expound upon both? Like what are the improvements specifically and what value do they unlock?
11:47
let me talk about content-based caching first, and then we can go into the graph-based execution.
11:54
So the way that content-based caching works is we look at the command that you're running, and this kind of applies at the granularity of an individual step in a traditional CI platform.
12:07
So if you think about maybe a CI pipeline from an application, it's usually broken down into jobs that are running in parallel.
12:15
Each one of those jobs is broken down in individual steps to install system packages and clone the code and configure the runtime and configure the packages and all the dependencies that you need.
12:26
So this technique we apply at the granularity of individual steps.
12:32
And the way it works is we look at the command that we're running and we look at the contents.
AdamHOST
14:33
Did you talk about the graph-based
RomainHOST
8:01
What
8:02
is changing now with AI? So there's a few really big opportunities in this space.
8:08
And the traditional model, again, focused on pipelines where you need distributed compute to run everything because you need to parallelize it as much as possible.
8:18
There's a couple inherent problems.
8:21
One is that we've had this trend over the past decade, maybe, of moving towards more ephemeral infrastructure.
8:28
We've obviously learned principles in engineering that persistent state machines is problematic for a wide variety of reasons.
8:37
And we really like the properties of ephemeral infrastructure, where you start from scratch every single time and you know, rebuild your application, you don't have persistent infrastructure.

16 MINS LATER

RomainHOST
24:40
What are some of the technical ways of doing this, but also let's say some of the mental models that you have or you have seen for those humans to be in the loop? I
4:40
So getting into RWX, like how does it compare to like, if someone's really familiar with GitHub Actions running CI on that or like Circle back in the day, like what does RWX kind of bring to the table and what are the kind of main differentiators? So
4:55
RWX is actually very similar to other CI platforms in the basics.
4:59
We'll run on version control triggers, you know, you get status checks on your PRs, things that engineers are used to.
5:05
But we have a vastly different execution model.
5:08
So every other CI platform, GitHub Actions, CircleCI, Buildkite Jenkins, they've been built around the same abstraction, which is jobs running on virtual machines.
5:18
You paralyze your pipeline.
5:19
There's a bunch of independent jobs because you want everything to run as fast as possible.
8:46
And these would kind of either help with like ideas that I've been trying to like mock up or build internally or services we already use that I think could be improved.
12:05
And why is that bubbling up this issue? What are they now seeing that maybe they were either not seeing before or they were willing to live with before? Yeah,
12:15
I think that's a great intro to some of the changes that we've made in the approach that we've taken at RWX because there's still a lot of engineering teams that are in effect working the way that they're working before, but trying to get agents to do the things that people were doing.
12:29
And I think some of the most effective engineering organizations are the ones that are rethinking the way they work completely from first principles.
12:36
Some of this for us at RWX was very fortuitous because I think the approach that we were taking to improve continuous integration tooling for engineers also just positioned us to be far more agent-native.
12:49
And a great example of that, Brandon, is the fact that every other CI platform, as I talk about like the evolution, the abstraction of CI, was very version control centric.
12:58
And it made sense.
12:58
They built their platforms that probably if an engineer spent a few hours writing code and wanted to run the bill, it was a good time just to push it up to GitHub anyway.

9 MINS LATER

21:56
How do these things actually work?

We value your privacy

We use cookies to understand how you use our platform and to improve your experience. Click “Accept All” to consent, or “Decline non-essential” to opt out of non-essential cookies. Read our Privacy Policy.