The Office of Steampunk DevSecOps

A department of Chimney Top Digital Services that does not exist, for work that does: running a small software company on hardware we already own and on the cheapest tool that passes the test.
The office is fictitious. There is no separate team, building or org chart; every guide below is written and published by Chimney Top Digital Services, like everything else on this site. The seal marks a kind of article, not an author.
What the office files
- Free-cycled and salvaged hardware put back to work: an old desktop as a build runner, a 2017 iMac on Linux, retired phones as servers.
- Pressure tests of paid tools against free ones, with the numbers, before we commit money.
- Moving work from the cloud to our own machines and back again when the measurements say so.
- How we build with AI coding agents: two models reviewing each other, questions batched in QnA MD files, an MCP interface on every app.
Each note reports what we tried, what we measured and where the result stops applying. Some experiments made things slower; those are filed too.
Where the seal came from
We wanted a mark for these notes that looked like the work: old parts doing new jobs. A first round of film-inspired symbols was too generic, so we pointed the design at photographs of handmade workstations built from typewriters, camera bellows, gauges and valve tubes. Polished color versions looked machine-made, so the drawing moved to charcoal on graph paper, like a page from a workshop notebook.
The salvaged typewriter terminal became the center of a round seal, with our name around the top and the office on the nameplate. Its screen shows an activity-decay curve labeled Ac-225 Activity, a nod to the radiation calculators: activity falls by half each half-life, so two half-lives leave 25 %. It is an illustration, not a calculation.
We made the seal with an AI image model over one evening's design session, in a charcoal-pencil style; it was not drawn by hand. The full design record is kept with our project files.
Filed by the office
14 articles, newest first.
An SSD, more memory and a gigabit switch: measuring three upgrades to our patchwork lab
Engineering note
We moved a 2017 iMac onto an SSD, doubled a 2014 laptop's memory and cabled three machines to a switch, then timed the jobs the lab really runs.
We turned two retired Samsung phones into always-on Linux servers
Engineering note
Two old Samsung phones become headless Termux Linux servers: SSH, auto-start on boot, a day-zero hardware baseline, and why memory bandwidth decides the rest.
When trimming CI was not enough: moving the jobs to a desktop we already owned
Engineering note
Why more pipeline trims could not reach our GitHub Actions target, four paths compared on cash, and the two-hour setup of a self-hosted runner on a desktop.
How we gave every app an MCP interface, and what an AI assistant can do with it
Product help
Why each of our apps has a Model Context Protocol server, how one is made from the app's own functions, a shielding question worked through it, and the rules.
How we audited GitHub Actions cost and run time
Engineering note
How we audited repeated CI work, separated waiting time from runner minutes and chose simpler checks with measurable trade-offs, with a worked example.
We split one AI model across two home computers. In our tests, it got slower
Engineering note
We split Llama 3.1 8B and 3.3 70B across a desktop and 2017 iMac over Wi-Fi. All splits generated text more slowly; bottleneck causes need more tests.
Our 2017 iMac couldn't run Ollama on macOS. On Linux it reached 70% of a modern desktop
Engineering note
In our October 2026 test, Ollama needed macOS 14; on Ubuntu, a 2017 iMac ran 3B and 8B models at about 70% of our Dell’s speed.
Batch questions with QnA MD and the agentic development life cycle
Product help
How agents compile open questions into one QnA MD file, an owner answers them in one sitting, and the answers are applied and recorded before the next round.
How we first set up Claude and Codex in one repository
Engineering note
The first arrangement: Codex built, Claude reviewed, the owner merged, signed and deployed. Verification records, handoff reports and one merge queue.
Two AI models, one repository: early speed and quality data
Engineering note
Measured results from the first 25 hours of Claude and Codex reviewing each other: pull requests, review findings and test counts. Early, thin data.
We built a private AI server without a dedicated GPU. Here's what our first benchmarks showed
Engineering note
Three local models on a Windows desktop with no dedicated graphics card, run with Ollama: 15.67, 7.67 and 4.26 tokens per second, and how we measured.
What we changed after a day of two-model development
Engineering note
The 2026-10-19 changes: approvals delegated to the other model, what the owner keeps, automation held to the same rules, and the build-review flip.
Why we built QnA MD: agent questions that outlive the chat
Product help
Why an AI agent's questions and decisions belong in a Markdown file as hidden HTML-comment tags: git-friendly, invisible when rendered, offline and phone-first.
Why we use two AI models to build and review each other's work
Engineering note
One model reviewing its own work misses things. We had Claude and Codex check each other's work and batched the owner's decisions to what matters.