Thinking in Platforms

What it’s about

Two people from the platform engineering community make the case that platform engineering is more than an infrastructure team with a portal. They frame it as an operating model: identify the work an organisation does over and over, then design the paths that work takes so teams get it done safely and with less effort. The AI era sharpens the argument, because agents now do some of that work too.

What I learned

A platform is a production system, not a toolbox. Buying good tools and putting them in a catalogue doesn’t make a platform. A platform is designed around recurring work and the outcome each path should produce. I’d been saying “platform” loosely; this gave me a stricter definition.

Design for the paths work actually takes. Start from what teams repeatedly need to do (start a service, ship a change, get access to data) and design those journeys end to end, rather than starting from the tools you already own. This is the same instinct behind Golden Paths.

Agents are users too. If an agent is going to open pull requests, run checks or query systems, it needs the same things a new developer needs: clear defaults, guardrails, good feedback and least-privilege access. A platform that only works for people who already know the tribal knowledge won’t work for agents at all.

Platform thinking applies beyond engineering. Any recurring knowledge work (writing, reviewing, publishing) can be treated as a designed system with paths and guardrails.

What I applied

The clearest application is personal, and you’re looking at it. My second brain runs as a small production system: I capture freely in a daily log, a nightly command distils it into notes, and a publishing step turns some of those notes into pages on this site. Reading this book made me treat that as a platform rather than a pile of scripts: one constitution file as the contract, small named commands as the paths, git as the safety net, and a privacy gate as the guardrail on the publishing path. It’s written up in How I Take Notes: Capture, Distil, Resurface.

This site was also built by several coding agents working in parallel, each in its own git worktree with a shared brief (Git Worktrees: A Desk per Branch). That was a small, real test of “agents as platform users”: the clearer the brief and the checks, the less steering they needed.

At work, where I’d use it next is platform discovery: asking teams about the recurring work they do before proposing any tooling.

Where it connects

Golden Paths · Internal Developer Platform · What Makes an AI Agent · Golden Paths and the Recommended Routine · Platform Strategy · Domain-Driven Platform Engineering