Embracing the 3-Month Rule: A Practical Approach to Unscalable Solutions

In the world of startups, we often hear the seasoned advice of industry veterans like Paul Graham: “Do things that don’t scale.” But while this mantra is well-known, the specifics of applying it—especially in the coding arena—are rarely discussed. After dedicating the past eight months to my AI podcast platform, I’ve devised a straightforward framework that I like to call the “3-Month Rule.”

This strategy is simple: any unscalable hack I implement is given a lifespan of three months. At the end of this period, it must either validate its worth and transition into a robust solution or be retired without regret.

As engineers, our instincts often push us toward scalable designs right from the start—think complex architectures featuring microservices, distributed systems, and all the intricate design patterns we’ve been trained to follow. However, in the startup landscape, focusing on scalable code too early can amount to costly procrastination. Often, we find ourselves optimizing for hypothetical user bases while overlooking the very immediate needs of our actual users. My 3-Month Rule encourages me to embrace straightforward, sometimes “messy,” code that gets delivered quickly and helps refine my understanding of user needs.

Current Infrastructure Strategies: A Case for Simplicity

1. Single VM Utilization

All elements of my infrastructure—from the database and web server to background jobs and caching—reside on a single virtual machine costing just $40 per month. While this approach lacks redundancy and requires manual backups, it has afforded me invaluable insights into my resource requirements over the past two months. I’ve discovered my AI-centric platform typically peaks at 4GB of RAM. The elaborate Kubernetes architecture I almost pursued would have merely managed empty containers. When crashes (yes, two so far) occurred, the insights gained about failures were enlightening—revealing weaknesses I could not have anticipated.

2. Hardcoded Configuration

In my codebase, configuration values are represented by simple constants:

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

Rather than investing time in creating configuration files and managing environment variables, hardcoding these values allows for swift searches and easy tracking through version control. Modifying any setting results in a prompt redeployment. Over three months, my adjustments