Build Worse Software

Published July 10 2020
A screenshot of my first software project, X96 Panel.

One of my first software projects -- probably the first project I did that wasn't just building a website theme -- was building my own CMS. The year was 2010, I was a middle school student doing web design on the side, and I was tired of WordPress.

I am reminded of this project from time to time; usually I remember how much I enjoyed building it. I used this custom CMS for some local business websites during my consulting days. The project was just a pile of procedural PHP scripts, with HTML and PHP just smashed together. MySQL was the database of choice, with unprepared SQL strings littered everywhere.

Somewhere along the way since then, I learned the "right" way to build things: carefully, with lots of planning, tests-first, deployed in a scalable way, stateless if possible… the list goes on. And with that knowledge, I lost the ability to just build something. My first question wasn't "What features should it have?", but rather "What tools and architecture should I choose?".

In my path to becoming a "professional programmer" I lost the most important goal of writing software: to solve problems. I became enamoured with the tools, processes, and dogmatic "correctness" of various communities that I blinded myself to what I was actually trying to accomplish.

My projects languished in various states of incompleteness as every time I would try and make progress, I would be hindered by a crippling narrative that I needed to do it the right way. I was paralyzed by my own irrational standards.

My solution to this was to build worse software. All the barriers of code quality, tests, flawless architecture -- they're are all constructs that affect production software; but they do not affect software that's just getting started. I had to unlearn my impulses and force myself to take naïve risks and shortcuts to getting a working solution. Visible progress is the finest motivator.

Since then I've made an effort to purposefully write bad software first, and good software later. By prototyping a messy, mostly-working version of something, you explore the solution organically and have a starting point for improvement.

This practice of care-free prototyping is what has allowed me to build Zaiko, my inventory management software. I needed to track server parts as I built and shipped some machines to a datacenter, maybe 4 or 5 years ago. None of the existing solutions seemed particularly nice to use, so I started building my own.

I could have spent decades planning, scheming, and architecting a textbook-worthy application that followed all the rules; instead, I slapped together a Rails application that worked "enough" and solved my problem. Five years and one rewrite in Elixir later, I understand so much more about what I want from the product and have slowly built up a production-ready SaaS application that is ready to scale and extend, and has pretty decent test coverage.

Was that Rails prototype a waste of time, since I rewrote it? Absolutely not. It let me quickly explore features and ideas, and got me using my own product as soon as possible. Would I have felt comfortable opening it up to the world? Also absolutely not. The goal isn't to build worse software and call it done; the goal is to build worse software so you get the context you need to evolve it into good software over time.

If I had to sum up the lesson I learned in a sentence, it'd be… Time heals all software? You have to build the village before the roads? Something like that, anyway; I'm still working on it.

Addendum: A while back I found a snapshot of the code for my first CMS archived on Google Code (RIP). I managed to export the SVN repo into Git and have archived the project on Github for those who are curious to see some godawful PHP. Only took a few seconds to find an SQL injection vulnerability. Oh, how times change.