Natural Language Coding: Writing Software with Plain English Prompts
Baljeet Dogra
The barrier to entry for coding is dropping, not because the syntax is getting easier, but because we are moving away from syntax entirely. "Natural Language Coding" is the practice of describing what you want in plain English, and letting AI handle the how. It's time to treat pseudocode as your primary programming language.
The "Pseudocode First" Approach
In traditional development, pseudocode was a planning tool—something you scribbled on a whiteboard before writing the "real" code. With Copilot, that whiteboard sketch is the code generator.
Instead of diving into function definitions, try writing a step-by-step plan in comments. Copilot will read your plan and implement it line by line.
Example: CSV Parsing
// Function to parse a CSV string
// 1. Split the string by newlines
// 2. Extract the headers from the first line
// 3. Map over the remaining lines
// 4. For each line, split by comma and create an object mapping headers to values
// 5. Filter out any empty lines
// 6. Return the array of objectsBy pressing `Enter` after these comments, Copilot will generate the exact implementation that matches your logic flow.
Iterative Refinement
Rarely does a human explain a complex task perfectly on the first try. The same applies to AI. If Copilot generates code that is almost right, don't delete it and start over. "Talk" to it.
Add a comment above the specific block that needs changing.
Refining Error Handling
try {
await database.connect();
} catch (error) {
console.error(error);
// Retry the connection 3 times with exponential backoff
}Copilot will see the existing catch block and inserting the retry logic you asked for.
One-Shot Prompting
Sometimes, the best way to explain what you want is to show an example. This concept, known as "One-Shot Prompting" in LLM terms, is incredibly effective for data transformation tasks.
Data Transformation by Example
const input = ['John Doe', 'Jane Smith', 'Bob Jones'];
// Transform to: [{ first: 'John', last: 'Doe' }, { first: 'Jane', last: 'Smith' }, ...]
const output = ...Copilot immediately understands the pattern and generates the `.map()` function to achieve the transformation.
Conclusion
We are moving towards a future where the primary skill of a developer is not memorizing standard library functions, but clearly articulating logic and requirements. Natural Language Coding allows you to code at the speed of thought. The next time you sit down to write a feature, start with the English, and let the AI handle the syntax.