Kyle Thompson

Software Engineer @rwx_cloud ✨

[Kyle Thompson @kylekthomp]

@thdxr Highly recommend taking a look at RWX (https://t.co/4RbIJNhnwy)! CI can be fast without the old stateful approach.

[Dan Manges @dan_manges]

GitHub just shocked engineering team budgets with a change to GitHub Actions pricing.

For people using self-hosted runners, or runners provided by a third-party, GitHub previously offered their CI/CD control plane for free. This will no longer be the case.

Many people using third-party runners chose that option for the cost advantages, and this changes the equation.

For engineering orgs that really want to reduce costs though, there’s a better way. GitHub Actions still uses the traditional approach of running jobs on parallel VMs. However, running pipelines as a graph (DAG), while utilizing automatic, content-based caching, provides the most substantial cost reductions.

This approach is the key to how the big tech companies like Google and Meta run their pipelines.

We built @rwx_cloud to make this approach accessible for regular engineering teams. RWX runs pipelines faster, and more cost effectively, than any other CI/CD platform.

We completed a cost comparison today where a pipeline on RWX used a total of 84,004 seconds of execution, while the identical pipeline on GitHub Actions consumed 300,787 seconds of compute, for a 70%+ reduction on RWX.

If anybody wants to relieve some pressure on their engineering budget in 2026, send me a DM.

[Kyle Thompson @kylekthomp]

Check it out! https://t.co/0lqiEPvGju

[Dan Manges @dan_manges]

We deleted our Dockerfiles at @rwx_cloud and moved to a better, faster way to build container images.

[Dan Manges @dan_manges]

Next Monday, @rwx_cloud will be launching the biggest evolution in container technology since the release of Docker’s BuildKit, which is now over 7 years old.

Docker changed how applications are deployed. Containers bundle system dependencies with application code, providing a portable image that can run on any generically configured server.

Unfortunately, building container images can be painfully slow and inefficient. Too many engineering teams are hindered by a CI/CD pipeline that takes 10+ minutes to build and push images before even the smallest code change can run.

The sluggish performance impedes human productivity, and it stands in the way of unleashing agentic coding.

Until now. We looked at the problem from first principles and built an entirely new platform for building container images. It is up to 10x faster and produces the same Docker images that everybody runs today.

Join us on a Zoom call on Monday November 24th to be one of the first to see it 👇

https://t.co/vc65AGkore

[Kyle Thompson @kylekthomp]

@dan_manges RUUX

[Kyle Thompson @kylekthomp]

@ayaz_hafiz_ A starting point: https://t.co/SEUPWoA7zR

Greg young talks about it a lot as well, if you'd rather a talk format: https://t.co/dDJtnjJKJr

[Nate Vick @nate_vick]

Everyone should be checking out @rwx_cloud to replace their CI. This team is killing it!

[Ayaz Hafiz @ayaz_hafiz_]

still blown away every time a merge into main instantly deploys because of @rwx_research Mint's cache hits.

[RWX @rwx_cloud]

We agree. One reason among many that Mint provides the best developer experience in CI/CD 😎

[RWX @rwx_cloud]

We raised a $7M seed round led by @QuietCapital, and today we're launching Mint, the Continuous Integration and Continuous Deployment (CI/CD) platform with the best developer experience.

https://t.co/BqDU5g02BZ

[Dan Manges @dan_manges]

Why should CI runs have to be finished before they're marked as having failed? We separated result status and execution status in Mint. Now runs are marked as failed as eagerly as possible, even if tasks are still running. @TAGraves just made it happen 🎉

[Dan Manges @dan_manges]

We built a breakpoint debugger into Mint to solve this problem. Call `mint-breakpoint` anywhere you want, and then run `mint debug` using the local CLI to open a shell. Shouldn’t have to push and guess at changes in CI/CD workflows.

[Dan Manges @dan_manges]

When you revert a PR and CI is entirely cache hits 😎

After using Mint, I can't go back to CI without content-based caching

[RWX @rwx_cloud]

We have been a long time customer of @depotdev. It's the fastest way to build Docker images, and we use it for all of our applications. We've built Mint to be the fastest CI platform, so we knew we needed to make Depot available. And now it is!

https://t.co/ggY6QaadT2

[Dan Manges @dan_manges]

@HamelHusain Mint solves all of those problems! Is this the DX you're looking for?

[RWX @rwx_cloud]

We’re sharing the first public preview of Mint, our new build tool, at the end of May. Join us at one of these sessions:

🗓️ Tue May 30 @ 4pm Eastern: https://t.co/zkjOlAbvVx

🗓️ Wed May 31 @ 10am Eastern: https://t.co/6IYDE9N4z9

[Nate Berkopec @nateberkopec]

If each of our tests have “five nines” of stability:

Each test will pass 99.999% of the time and flake 0.001% of the time.

Then, a build of 100,000 tests will flake and fail 64% of the time (0.99999^100000).

Test stability is hard.

[RWX @rwx_cloud]

Captain is now open source! It's is a CLI that can detect and quarantine flaky tests, automatically retry failed tests, partition files for parallel execution, generate comprehensive test failure summaries, and more.

https://t.co/22l5Xs3j0L

[RWX @rwx_cloud]

Last year, we began building a universal test runner that executes test suites as distributed workloads. After a thorough beta testing period over the past several months, we’re excited to announce the generally available, open source release of ABQ.

https://t.co/Qj9w1U4K4P