A Small Defense of Boring Software

I have spent enough years around interesting software to be suspicious of it. Interesting software is a pleasure to announce and a burden to inherit. Boring software — the kind that logs in, writes a row, sends an email — is what most of the working world actually needs, and it is strangely hard to keep boring.

The pressure is social. Nobody gets a conference talk for choosing Postgres again. Nobody gets a spike of dopamine for leaving a cron job as a cron job. The industry rewards the sentence that starts with “what if we.” Boring software is the sentence that ends.

What I mean by boring

I do not mean ugly, and I do not mean stagnant. I mean predictable failure, obvious state, and a restore path you could explain to someone half-asleep. I mean dependencies you can hire for. I mean features that refuse to arrive until they have an owner for the 3 a.m. version of themselves.

Agents make this argument sharper. A loop on top of a chaotic system is chaos with better grammar. If you want an agent in the building, first make the building boring: stable APIs, crisp receipts, tests that mean something. Then the loop has a world it can trust.