Embracing the 3-Month Rule: A Practical Approach to Unscalable Hacks in Software Development

In the entrepreneurial realm of startups, the adage “Do things that don’t scale,” championed by Paul Graham, resonates profoundly. However, a less-explored aspect of this wisdom is how to systematically implement it, particularly in the coding process.

Over the last eight months of developing my AI podcast platform, I’ve crafted a straightforward yet effective framework: every unscalable technique I employ gets a trial period of three months. If the method demonstrates tangible benefits within this timeframe, it will be properly integrated into my architecture; if not, it will be discarded.

As software engineers, we are often conditioned to prioritize scalable solutions from the outset. We become enamored with sophisticated design patterns, microservices, and distributed systems, envisioning our applications catering to millions of users. However, this ambitious mindset can lead to expensive delays, especially for startups. My three-month rule compels me to write simpler and more immediate code, which facilitates shipping and offers valuable insights into my users’ needs.

Exploring My Strategic Infrastructure Hacks

1. Consolidating Operations on a Single VM

I run my complete stack—including the database, web server, background jobs, and caching—on a single $40/month VM. This setup lacks redundancy, and manual backups are saved locally.

While this may sound reckless, it has proven to be a brilliant move. Within just two months, I’ve gained deeper insights into my actual resource requirements than any theoretical capacity planning document could provide. I’ve discovered that my platform’s “AI-heavy” processes peak at only 4GB of RAM. The complex Kubernetes architecture I nearly deployed would have wasted time managing non-existent workloads.

When my system crashes—it’s happened twice—I gather invaluable information about the failure points, which often surprise me.

2. Directly Hardcoded Configuration Values

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

I opt for hardcoded constants scattered throughout the code, eliminating the need for configuration files or environment variables. To make any changes, I simply redeploy.

The beauty of this approach lies in its efficiency. I can quickly search my codebase for configuration values, and any modifications are immediately part of my version control history. In three months, I’ve changed these constants only three