Every post leads with a technical problem, carries the evidence, and points at the repository. Never a bare link.
Week 1
LinkedIn
The origin story, with the evidence
The pipeline went red because the bill was due
“Twenty-eight builds in a row failed. Not one of them failed on code.”
On 21 October we used the last of our 2,000 included GitHub Actions minutes, with eleven days left in the cycle. Every job from that point either cost money per minute or did not start at all. The largest repository in the workspace had 28 failing runs out of the last 30, all of them in about six seconds, all of them with the same annotation: the job was not started because the spending limit needed increasing. It looked exactly like a broken build. It was an empty account. CI Kit is the answer: the checks that do not need a shared machine now run on the developer's own laptop, in about a hundred seconds, for nothing.
CTA: The full write-up, with the annotation and the numbers, is on my site. Link in the comments.
Week 2
LinkedIn
Problem → evidence → CTA
A red build is not always a red build
“Our pipeline was red for a week. Not one of those 28 failures was a code failure — the jobs never started.”
GitHub had refused to start the jobs because of a spending limit. Every run failed in about six seconds. At the pull request level a red build is a red build, so the week looked like ordinary breakage. What I built to find out what was actually happening is now open source, and I wrote up the incident — including the measurement I threw away.
CTA: The full write-up, with the numbers and the harness, is on my site. Link in the comments.
Week 3
X
Terminal screenshot + one number
Benchmark release
“Idle CPU, four mock containers vs one dependency-free process:”
7.9% → 0.0%. Memory 250 MB → 50 MB. First request 20 s and rising → 1.4 s. Same mocks, one process, no dependencies. The footprint is a build gate now — the build fails above 100 MB, so the number cannot quietly stop being true.
CTA: Open source: github.com/omeiza-workspace/api-mock-services
Week 4
LinkedIn
Architecture breakdown + directory-tree image
Business logic as data, on a host with no internet
“Here is the layout of an air-gapped automation engine that runs enterprise workflows from JSON instead of controller code.”
The engine core is immutable and lives in its own module scope. Workflows are pure JSON, so a business rule changes by editing a file rather than by shipping a deploy. New integrations are *.plugin.ps1 files that get auto-discovered — the core is never edited to add one. Every step attempt, including retries, is written to SQLite so an auditor can trace a self-healed run. Zero external dependencies, no build step, no internet in the request path.
CTA: The full architecture write-up, with the layout and the design decisions, is on my site.
Week 5
X
Milestone + thanks
Repository milestone
“CI Kit just crossed [N] stars.”
Started this because I watched a team fail 28 builds in a row without anyone noticing the jobs had never run. Thank you to everyone who has filed an issue — particularly the ones that told me the toolchain check was hiding a class of real failures. Open to issues and PRs.
CTA: github.com/omeiza-workspace/ci-kit