The 3-Month Experiment: A Pragmatic Approach to Unscalable Coding

In the entrepreneurial realm, it’s common to hear the advice from Paul Graham: “Do things that don’t scale.” However, implementing this mantra within software development remains a less explored topic. After spending the past 8 months developing my AI podcast platform, I’ve adopted an intriguing approach: every unscalable solution I implement is given a lifespan of just three months. At the end of this period, I evaluate its effectiveness—if it proves its worth, it transitions into a more robust solution; if not, it’s time for it to go.

As developers, we often default to creating scalable architectures from the outset, delving into elaborate patterns like microservices and distributed systems—designs specifically crafted to accommodate millions of users. However, this mindset can be detrimental for startups, where aiming for scalability too early can equate to an exercise in costly procrastination. We find ourselves preemptively solving problems for users who aren’t even on the horizon. My three-month rule encourages me to produce straightforward, even “inelegant,” code that allows me to understand the genuine needs of my users.

The Value of My Current Infrastructure Hacks

Here are some of the unconventional strategies I’ve employed and why I’ve found them to be more insightful than you might think:

1. Consolidated Operations on a Single VM

I operate my entire stack—a database, web server, background jobs, and Redis—on a single $40/month virtual machine. While this approach lacks redundancy and relies on manual backups, it has offered invaluable insights into my resource requirements. Over the past two months, I have learned that my usually “AI-heavy” platform runs comfortably at 4GB of RAM. The intricate Kubernetes configuration I contemplated would have only served to maintain idle containers. Each crash (yes, it’s happened twice) provides real-world data on what actually fails—often surprising me with its unpredictability.

2. Simplicity with Hardcoded Configurations

My approach to configuration is decidedly low-tech—settings are hardcoded directly into the code:

python
PRICE_TIER_1 = 9.99
PRICE_TIER_2 = 19.99
MAX_USERS = 100
AI_MODEL = "gpt-4"

No configuration files or environment variables complicate things; instead, I can easily grep my entire codebase for any setting. Every change gets logged in Git history and