AI and worktrees are filling our disks: kache storage, measured
The way we build is shifting to many parallel worktrees, one per task and often driven by AI agents, and each is a full build tree on disk. A measured look at how kache shares those bytes: on APFS a second Firefox worktree restored 13.5 GB without copying one, and reflink sharing saved 27.4 GB.
Where Does My Traffic Go? Services, Selectors, and Readiness in Kubernetes
A Service can look completely healthy and still route to nothing. Here is how a Service decides which pods receive traffic, why a valid Service can route to nothing, and how to look at the set of pods behind it directly.
Designing a correct compile-cache key
A compile-cache key can fail two ways: too specific (needless recompiles) or too loose (wrong artifacts). How kache v0.5.0 corrects both, and what it does about inputs the compiler can never see.
kache v0.4.1: what it takes to trust a build cache
Two releases covering cache key correctness, three reliability hardening changes, experimental C/C++ caching, and first-class Windows support.
Cloud Repatriation Explained: Current Trends in Hybrid Infrastructure
Cloud repatriation has moved from niche debates into mainstream enterprise conversations. We look at what the data says, what engineers are actually struggling with, and why the real issue is operational clarity, not cloud vs on-prem.
Content-addressed Rust builds (or, what kache actually caches)
How kache works under the hood: a RUSTC_WRAPPER with a content-addressed cache keyed by source, dependencies, and toolchain, restoring hits with reflinks and hardlinks across worktrees and CI.
kache 0.3.0: C/C++ build caching, crates.io, and the storage win
kache 0.3.0 brings C/C++ object caching behind a conservative allow-list, reflink-first restores, Windows groundwork, and an in-progress crates.io name transfer, plus why storage savings matter as much as speed.
The Moment You Pipe kubectl logs Into grep, You've Already Lost
Why the grep-jq-pipe-repeat loop fails at 2:47 AM, and what a structured log interface changes about Kubernetes incident response.
Rolling Updates and Rollbacks in Kubernetes
A Deployment doesn't manage pods directly. It manages ReplicaSets. Here's why that two-layer model makes zero-downtime updates and instant rollbacks possible, and how to watch the handover happen.
Cluster updates, in your inbox.
Kubernetes deep dives, GitOps field notes, and platform-engineering essays from the team building Kunobi. Two posts a month. No fluff.