Platform engineer: I checked the title before I used it
Lesson from:
One job title, thirteen capabilities, five gaps
There is a simple question I could not answer well: what do you do?
I run a lab on one machine in a small flat. It collects news in dozens of languages, catches attackers in honeypots, publishes signed threat reports, serves a library of 120,000 works and gives AI agents the tools to work on all of it. When someone asked for my job title, I had several answers and none of them felt right.
So I asked a different AI model to read the site as a stranger and tell me what it saw. It answered with two words: platform engineer.
I had heard the term. I did not know whether it was mine to use. This post is about how I checked, and about the page that came out of it: The Platform Document.
Contents · 6 sections
What the title means #
A platform engineer builds the things other engineers build on, and treats those things as a product with users.
A team that writes an application should not also have to work out how to log users in, where to keep secrets, how to store data safely and how to see what broke at night. Somebody builds those things once and offers them to everyone.
The definitions I read come from different authors and years. Four words return in all of them: product, self-service, users, and complexity removed.
What it does not mean #
The sources are just as clear about where the title stops.
- It is not a renamed operations team.
- It is not a tool. Installing Kubernetes is not the job.
- It is not a ticket queue. If users must ask and wait, the platform has failed.
- It is not a mandate. People should choose the platform because it is better.
- It is not new. One publisher calls it the natural evolution of DevOps and infrastructure as code.
That last point matters to me. I have done this work for years under older names.
The count #
The Cloud Native Computing Foundation lists thirteen capabilities a platform provides. I had the lab checked against each one, with numbers taken from the running system on 17 November 2026.
| Verdict | Capabilities |
|---|---|
| Strong | Tools and APIs, observability, infrastructure, data services, identity and secrets, security |
| Present | Portals, documented paths, development environments, artifact storage |
| Partial | Build and test automation, delivery automation |
| Thin | Messaging and events |
Ten of thirteen are present or strong. For scale: 71 namespaces, 183 deployments, 207 running pods, 69 alert rules, 338 test files, on one machine.
A large share of the code was written by AI agents under my direction. I have always said so on this site, and it matters for the next point.
The users are machines #
Every platform needs users. Mine are AI agents.
When an agent starts work here, it does not ask me for a password and it does not guess which command is dangerous. It gets tools with declared risk levels, a memory it must read first, a board that says which files belong to another session, and rules about what only I may do. Three agent sessions work on one codebase in parallel this way.
Google’s DORA research for 2025 found that AI helps strongly where the platform under it is good and hardly at all where it is poor. Their phrase is “AI is an amplifier”. My lab is a small working case of that sentence.
Five gaps #
A measurement that finds no gaps is an advertisement. These are mine.
- No teams of human developers use it. Listening to other teams and winning them over is half of the job in a company. This lab cannot prove it.
- It is one machine. No second node, no second region, no large cloud account.
- Nobody else is on call. When it fails at night, it costs only me.
- Delivery is still manual. Changes are in git, but there is no continuous delivery system.
- The standard measures are not tracked. How often changes ship and how often they fail is not on any dashboard yet.
Some of these are answered by the jobs I had before the lab. Some are simply true, and they are now on my list.
Why a third document #
This site already has two long documents. The Human Document tells where I come from. The Machine Document is the machine’s own account of how the work is done and what went wrong on the way.
The Platform Document is the third. It says what the work is called and how far the name holds. It is shorter than the other two, and it is written the same way: by the machine, at my request, with the instruction to be accurate before being kind.
I think a CV should look like this. Not a list of what I say about myself, but a claim, the definition it is measured against, the count, and the gaps.
If you are hiring, do not take the title on trust. The system is running. Ask to see it.