Embracing the 3-Month Rule: A Pragmatic Approach to Technical Decisions

In the startup world, the advice from Paul Graham to “do things that don’t scale” resonates deeply. While many discuss this concept philosophically, few delve into the practicalities of applying it in a technical landscape. After spending the last eight months developing my AI podcast platform, I’ve adopted a methodology that I refer to as the “3-Month Rule,” which allows unscalable hacks to exist for a limited timeframe. If they aren’t providing value after three months, it’s time for them to go.

Rethinking Scalability in Startups

As engineers, our instinct often leads us to craft scalable solutions right from the outset—considerations of microservices, distributed systems, and other complex architectures emerge as we envision supporting vast user bases. However, in the startup environment, such approaches can lead to expensive delays, as they anticipate user demands that simply don’t exist yet. The 3-Month Rule prompts me to prioritize simplicity and speed in my coding practices, enabling me to quickly learn what my users genuinely need.

Current Infrastructure Hacks: Navigating the Unconventional

Allow me to share some of my current infrastructure choices that may seem unconventional, yet have proven to be surprisingly effective:

1. Consolidated Virtual Machine Infrastructure

For a mere $40 a month, I’m operating everything—from the database to the web server and background jobs—on a single virtual machine (VM). This setup lacks redundancy, and I manually back up data to my local machine.

This choice has yielded invaluable insights about my resource requirements. In just two months, I’ve identified that my “AI-heavy” platform only peaks at 4GB of RAM—information that would have been lost in traditional capacity planning documents. When my system crashes—which has happened twice—I gather real-time data on failures, revealing issues that were often unexpected.

2. Hardcoded Configurations

My configuration management is straightforward: I use hardcoded constants throughout my codebase.

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

Instead of relying on configuration files or variables, this method enables rapid searches for any configuration value. Making adjustments requires a quick redeployment rather than extensive engineering efforts—a task I’ve only needed to do three times in the last three