Establishing a Baseline: The Anchor of Reality
Definition: A reference model for comparing how well another model performs, helping quantify minimal expected performance for new approaches.
The “Vibe” Trap
In Vibe Coding, it’s easy to get lost in optimization. “Can we make this code 5% faster?” “Can we make the UI 10% prettier?” The Baseline is your reality check. “Does it work better than the simplest possible solution?”
How to Set a Baseline with AI
Before you ask the AI to build a “Scalable Microservices Architecture”:
- Prompt: “Write the simplest, ugliest, single-file script that solves this problem.”
- Run it: This is your Baseline.
- Iterate: Now ask the AI to improve it.
Why this is critical
AI loves complexity. If you ask for a “Hello World” in Java, it might give you a Spring Boot application with 50 files.
- The Baseline Approach: “Give me the baseline implementation first.”
- Result: You get
public static void main.... - Comparison: Now you can decide if you need the complexity of Spring Boot.
Performance Baselines
- Scenario: “Optimize this SQL query.”
- Action: First, measure the current speed (Baseline: 200ms).
- AI Goal: “Make it faster than 200ms.”
Without the baseline number, the AI is just guessing at what “fast” means.
