The 3-Month Rule: A Pragmatic Approach to Building Scalable Code for Startups
In the world of software development, Paul Graham’s mantra, “Do things that don’t scale,” is echoed by many. However, discussions around how to effectively apply this advice, particularly in coding practices, are often lacking.
After eight months of diligently working on my AI podcasting platform, I have refined a framework that I adhere to: each unscalable innovation is given a lifespan of just three months. At the end of this period, the solution must demonstrate its value and evolve into a more robust system, or it is discarded.
The Challenge of Scalability
As engineers, we are accustomed to planning for scalability from the outset. We dive into design patterns, microservices, and distributed systems, crafting intricate architectures capable of serving millions of users. However, this approach often suits larger organizations better than startups. In the early stages of a venture, building scalable infrastructure can resemble procrastination—it focuses on optimizing for an audience that may not yet exist and attempting to solve problems that may not arise.
By implementing my three-month rule, I am encouraged to produce straightforward, even “imperfect,” code that can be deployed quickly, allowing me to glean insights into user needs in real time.
Current Infrastructure Strategies That Work
1. Unified VM Architecture
My platform operates on a single virtual machine, hosting everything from the database to the web server and job processing for a modest monthly fee. This decision, often seen as reckless due to the lack of redundancy, has yielded immense educational value. I’ve quickly understood my resource requirements, discovering that the “AI-heavy” platform functions efficiently with just 4GB of RAM. The complex Kubernetes configuration I initially contemplated would have led to unnecessary management of idle resources.
When crashes occur (and they have), I gain essential insights into performance bottlenecks. Remarkably, the issues that arise are not what I anticipated.
2. Hardcoded Configuration
Rather than relying on configuration files or environment variables, I have chosen to hardcode constants throughout my codebase. This approach might seem archaic, but it offers significant advantages. I can easily locate and modify configuration values, and each change is documented in version control. Since deploying a configuration service would take over a week, using hardcoded values has saved me considerable time. My adjustments have been minimal—just three changes in three months.
