01 introduction

All machines deserve a fair stack trace.

I build software — mostly MERN, increasingly with a model somewhere in it — and I run the infrastructure it lands on. The second part is why the first part ships.

The three Articles

I. The right to a fair stack trace

No process shall be terminated without a record of what it was doing and why it stopped. Silent failure is a due-process problem wearing a logging costume.

II. The right to a graceful shutdown

Fifteen seconds and a SIGTERM handler. Every service written here closes its listener, drains its work, and exits on purpose. It costs four lines.

III. The right to be understood before replaced

Nothing gets rebuilt because it was inconvenient to read. If it survived nine years in production, it knows something you do not.

02 operator

More and more of what I build has a model in it somewhere. Fifteen years of infrastructure is the reason I can ship it and still be the one who gets paged when it breaks.

React and Node on the front, Express and Mongo behind it, a model wherever it earns its place. The consoles I build get used by people who do not care what a container is — which is the correct attitude, and my problem to absorb.

The robot rights angle is the hook, and it is also not entirely a joke. The habits we form around machines that cannot object are the habits we will bring to the ones that can. I would rather practise now.

~500k lines of code shipped · 15+ years of infrastructure · 1 exposed port

See the projects · Read field notes