The Three-Month Rule: A Practical Approach to Unscalable Solutions in Tech
In the tech community, Paul Graham’s insight—”Do things that don’t scale”—is often cited, yet few discuss how to actualize this advice within the realms of coding and software development. After spending eight months developing my AI podcast platform, I’ve crafted an effective framework that I call the Three-Month Rule: any workaround or hack that isn’t designed to scale gets a trial period of three months. After that, it either earns its place through demonstrated value or is discarded entirely.
Embracing Unconventional Methods
As engineers, we frequently find ourselves entrenched in the mindset of creating scalable systems right from the outset. We immerse ourselves in advanced architecture like microservices and distributed systems, thinking of future growth and vast user bases. Yet, this is often a detriment in a startup context, where scalable solutions can end up being a costly form of procrastination. We risk investing time and resources into optimizing for hypothetical users while neglecting current needs.
The Three-Month Rule pushes me to implement straightforward, albeit imperfect solutions that can be deployed quickly. This approach enables me to gain meaningful insights about user requirements without getting bogged down in complex structures.
My Current Simplified Infrastructure Strategies
1. All Services from a Single Virtual Machine
I’m currently running my entire stack—including the database, web server, background processes, and caching—on a single $40/month virtual machine. While this may sound reckless, it has allowed me to efficiently measure my actual resource needs. For instance, I discovered that my platform peaks at just 4GB of RAM, indicating that a complex setup like Kubernetes would have only added unnecessary overhead. When issues arise (and they have), I gain direct feedback on unexpected failures.
2. Hardcoded Configurations
My configuration values are hardcoded into the codebase. For example:
python
PRICE_TIER_1 = 9.99
PRICE_TIER_2 = 19.99
MAX_USERS = 100
AI_MODEL = "gpt-4"
This method means that every change requires a redeploy, but the upsides are substantial. I can quickly search for constants across my entire codebase and keep a clear history through git. Instead of dedicating time to build a separate configuration management tool, I simply make minor adjustments when necessary, saving countless hours in development.
