@danlovesproofs lockfiles are new thing now and also reducing actions dependencies.
tbh, the safest setup may be going with a different system, like @buildkite and self host
Join us in celebrating our biggest sponsor and @evilmartians client: @buildkite.
Buildkite is a CI/CD platform that lets you run your pipelines on your own infrastructure, so you can scale builds and tests as far as you need without giving up control. It’s trusted by big teams, including Shopify, Airbnb, Uber, and Anthropic.
Buildkite’s core app is built on Rails, and they’re longtime supporters of the Ruby community. Get to know Buildkite: https://t.co/jeTskNdI1k
Stop playing the game of switching CI to the “lowest cost” provider. That is a first order solution.
You need to be building loops around CI optimization and automatic test selection. You should be compute-neutral. Just use @buildkite.
Use a compute-neutral CI provider and run your own compute. That's what the big boys are doing, which is why they all run on @buildkite.
Basically CI providers have turned into high-interest checking accounts or rewards credit cards. Each one makes a bunch of promises they can't keep to try to lock in customers, but it's all on borrowed time because their costs are not sustainable.
PART III: RE-RUN OF THE LIVING DEAD
You know the ritual. Re-run failed jobs. Re-run all jobs. Still red.
Stop feeding the undead. Buildkite finds flaky tests and quarantines them automatically, before the infection spreads to every build.
Shouldn’t your team spend the week shipping vs. fighting off the horde?
That’s a wrap. Happy Halloween 💀
PART II: BURIED
Ancient curse, modern invoice. Your default repo CI rounds every partial minute up to a whole one, on every job. Fan out a few hundred jobs and the curse will bury you...
Buildkite bills per second for hosted runners. Curse, lifted. No incantation necessary.
Up next: the undead return to close the trilogy.
PART I: PARANORMAL AVAILABILITY
“Waiting for a runner to pick up this job…”
Is your CI ghosting you when it matters most?
Now showing (in a queue near you).
Stay seated. Something ancient rises tomorrow.
Lights. Camera. Actions?
The unrest begins this week: a trilogy of CI horror that won’t stay buried rises in San Francisco.
Rated U for Unreliable.
Brought to you by @buildkite: the only CI that survives.
What I want to know from the 5 new CI and sandbox providers I see every day spruiking how they are x% cheaper than everyone else: What are your margins like? If you're a thin wrapper around someone else's compute, what is your actual plan to retain folks when the next provider undercuts you? How can you convince customers you'll actually be here in 12 months and won't go broke or lose their data?
I love the pull request. But reviewing agent-written code needs a loop. With Trails, evals feed the next round of agent work. Sessions, code changes, and CI checks from @buildkite and @depotdev stay together.
Entire already runs its own CI on @buildkite. As of today, every team on @GoEntire can.
Push → build → check on the Trail.
Congrats @ashtom and the Entire team. Now every team on Entire has access to the CI orchestration platform frontier AI labs rely on to build the next wave of pioneering software.
Agents reinvented how we produce code. It changed our tools and half of our workflow. It’s time for the other half to catch up.
Today, we’re shipping Entire: the Git forge for agent-human collaboration. We’ve built @GoEntire using Entire. Now we want to see what you'll build with it.
Here’s why and what’s different. 🧵