Zademy

It Just Needs to Work

Opinion; Pragmatism; Engineering
words words

The pull of the new

I catch myself doing it. I open the tech side of the internet, and within ten minutes I'm convinced my entire stack is obsolete. A new runtime, a meta-framework I'd never heard of yesterday that apparently solves everything, a database with a slick landing page. The pressure is constant, and most of it is manufactured.

Dan McKinley gave a talk at Etsy in 2015 called "Choose Boring Technology," and the core of it has stayed with me for a decade. Every company gets about three innovation tokens. The supply is fixed. You spend those tokens on the one or two bets that genuinely move the needle, and you let everything else be boring on purpose. MySQL is boring. Postgres is boring. Python and cron are boring. They are also good. Boring shouldn't be conflated with bad.

Boring has a track record

The companies quietly handling absurd load are usually the ones nobody writes thinkpieces about. Stack Overflow served billions of pageviews off an IIS and SQL Server monolith with a handful of engineers. Basecamp runs a Rails app with a tiny team, and Rails turned twenty-one this year. Etsy's activity feeds were PHP, MySQL, and Memcached, scaled roughly twenty times with nobody really watching them for years.

I bring these up because people treat "boring" like a confession of falling behind. It isn't. "Best tool for the job" is the real fallacy. Your job is keeping the company in business. The best tool is the least-worst one for most of your problems, regardless of how many GitHub stars it racks up this quarter.

The cost of chasing trends is real, and nobody bills you for it clearly. Two to four weeks of learning curve per new toy. Migrations that swallow a hundred, sometimes five hundred hours. Resume-driven development dressed up as engineering excellence. The senior engineers I trust most are the ones stepping back from the cutting edge rather than chasing it.

The caveat I won't skip

I want to be careful here, because "boring" gets misused as cover for negligence. Boring-good is a maintained, patched, battle-tested thing. End-of-life is boring-bad. Running PHP 5 or an unpatched Windows Server 2012 because "it still works" isn't pragmatism, it's a liability with a timer on it. Aspen Digital called this pattern out earlier this year, and they're right to. The point is to spend your limited attention where it actually matters, not to freeze the stack in amber and call that discipline.

It just has to work

There's something almost funny about the last couple of years. The most genuinely innovative companies on the planet, the labs doing real frontier AI work, ship it wrapped in the dullest infrastructure you can imagine. Flask or Express in front of Postgres with pgvector bolted on. No bespoke distributed framework. No manifesto on a stage. It just has to work.

Technology for its own sake is snake oil. Use the boring tool, ship the thing, keep it patched, and move on. If what you built is still serving a person or a company six months from now, you already did more than most of the people chasing the next big thing.