Embracing the 3-Month Rule: A Strategic Approach to Unconventional Coding Practices
In the startup world, adhering to seasoned investor Paul Graham’s principle of “doing things that don’t scale” can be transformative. Yet, there’s often a missing piece in the conversation—how do we apply this philosophy to the technical realm, particularly in coding? After eight months of developing my AI podcast platform, I’ve devised a compelling strategy: every unscalable approach gets a lifespan of just three months. If it doesn’t demonstrate its worth during that timeframe, it’s time to move on.
As developers, we’re conditioned to create scalable solutions from the outset. We tend to focus on grand architectures—design patterns, microservices, and distributed systems—suitable for enterprises. But in a startup context, pursuing scalability often turns into expensive delays where we optimize for hypothetical users, diving into complexities that may never manifest. My three-month rule compels me to produce straightforward, sometimes “messy,” code that gets deployed swiftly so I can genuinely comprehend user needs.
Current Infrastructure Strategies: A Look at What Works
1. Consolidation on a Single VM
I run my entire operation—database, web server, background processes, and caching—on a single $40/month virtual machine. While this setup may seem perilous due to its lack of redundancy, it’s been an enlightening experience. In just two months, I’ve gained invaluable insights into my actual resource demands. Surprisingly, my platform peaks at only 4GB of RAM. The intricate Kubernetes architecture I initially considered would have been overkill, managing phantom resources instead of tangible user interactions. Each time my system fails—twice so far—I glean real insights about points of failure, which often differ from my assumptions.
2. Hardcoded Constants for Configuration
Instead of using configuration files or environmental variables, I work with hardcoded constants:
python
PRICE_TIER_1 = 9.99
PRICE_TIER_2 = 19.99
MAX_USERS = 100
AI_MODEL = "gpt-4"
This method may lack flexibility, but it has its benefits. I can quickly search through my codebase for any value, ensuring that every change is captured in version control. Over three months, I’ve only had to make three adjustments, which took me 15 minutes of redeployment, far less than the week it would have taken to implement a dedicated configuration service.
