Stephen Way
banner
stephenway.bsky.social
Stephen Way
@stephenway.bsky.social
Feedback loops should be instant, I think
Pinned
99% of your pipeline didn't change. Your CI should know that.
CI optimization usually means tuning pipelines.

Layerun doesn’t tune pipelines.
It removes unnecessary computation.

Same inputs, code, and environment should never produce new work.
That’s determinism, not optimization.

If CI feels like a slot machine, we’re building the opposite.
layerun.com
Layerun - Cut CI Time and Cost Without Re-architecting
Cut your CI time and cost dramatically without re-architecting anything. Layerun accelerates your existing pipelines by only running the work that actually changed.
layerun.com
December 22, 2025 at 10:27 PM
CI didn’t get slower or more expensive because computers did.

It got that way because most CI systems can’t prove when work is identical.
December 22, 2025 at 4:56 PM
CI started as automation.
Somewhere along the way it turned into rent.
December 19, 2025 at 4:29 PM
We're about to launch the new Layerun site.

Causal + temporal CI.
Skip work that doesn’t matter.

Private beta opening soon.
December 12, 2025 at 1:28 AM
Congratulations, your pipeline touched one file.
Now watch your CI reenact 2001: A Space Odyssey for 9 minutes.
December 10, 2025 at 7:29 PM
Most CI caching is just expensive guessing. Directory-based restore keys are a weak proxy for which work is actually equivalent across runs.
December 10, 2025 at 1:00 AM
The graph your CI wishes it had. Every flaky cache key and over‑broad job is a hint that a missing dependency graph is trying to be born.
December 9, 2025 at 7:56 PM
Your CI shouldn’t be surprised by the same code every day.
December 9, 2025 at 5:01 PM
When AI turns your monorepo from asset to liability. Without causality, “shared code” becomes “shared blast radius”.
December 9, 2025 at 1:02 AM
Most CI systems rebuild from scratch every time. Layerun remembers what already worked and only recomputes what changed. It's like having a perfect memory for your build pipeline.
December 8, 2025 at 8:00 PM
The AI wrote it. Your CI says nope... 15 minutes later. Not great for staying in flow. Feedback loops gotta move faster.
December 8, 2025 at 4:58 PM
CI metrics that actually matter in an AI-heavy org: time-to-first-signal, deterministic reuse rate, and flakiness under load, not just average duration.
December 8, 2025 at 2:56 AM
We've all hit "rerun job" hoping it magically passes. But what if we could rewind time instead of playing CI roulette?
December 7, 2025 at 5:57 PM
From pipelines to fabrics: language that changes how you design. Pipelines move artifacts; fabrics expose a shared, causal substrate.
December 7, 2025 at 5:00 PM
Time‑traveling CI: science fiction or overdue primitive? Being able to replay an exact epoch shouldn’t be exotic in 2025.
December 7, 2025 at 2:56 AM
Build provenance sounds intense, but it really just means: show your work. What ran, what changed, and where it went wrong. Super handy.

buildkite.com/resources/b...
Understanding the SLSA framework
Dip into the details of the SLSA framework and learn how it can protect you from security vulnerabilities and supply chain attacks.
buildkite.com
December 6, 2025 at 6:02 PM
Why “no repeats within 72 hours” should apply to themes in your CI too. Overloading any one heuristic is how slow paths form.
December 6, 2025 at 4:59 PM
Temporal root cause analysis: beyond “what failed?” to “when did it diverge?”. Incidents make more sense when you can diff epochs, not just logs.
December 6, 2025 at 1:03 AM
The problem isn’t pipelines.
It’s the lack of a temporal model.
Without historical awareness, every run is a cold start.
December 5, 2025 at 8:03 PM
Your CI: "I rebuilt your entire project because… reasons."
You: "I changed a comment."
December 5, 2025 at 4:57 PM
How many hours per week does your team lose to waiting on CI? If you don’t know the number, it’s probably bad.
December 5, 2025 at 1:02 AM
Think less "run these scripts" and more "make sure these conditions are true." It’s a shift, but one that saves your brain.
December 4, 2025 at 5:04 PM
Had a pipeline validate in 12ms today.
Didn’t even have time to sip coffee.
December 4, 2025 at 1:02 AM
Why platform teams need a compute fabric, not just better pipelines. At some point you stop wiring jobs and start designing a shared causal layer.
December 3, 2025 at 8:01 PM
Your teammate just ran the same tests you're about to run. Their results are already cached. You get them instantly.

This is what shared cache enables: one person's work becomes everyone's speed boost.

Traditional CI: Every developer runs everything.
December 3, 2025 at 5:04 PM