From APIs to DSLs: How AI Is Reshaping the Way We Build Software

For decades, programming has been shaped by a simple assumption:

Humans write code, and computers execute it.

That assumption is breaking.

With the rise of AI coding systems like ChatGPT, GitHub Copilot, and frameworks like LangChain, we are entering a new paradigm:

Humans describe intent. AI writes and operates the system.

This shift does not just change who writes code. It changes what code is.

The Evolution of Interfaces: CLI -> API -> AI

Software interfaces have evolved in layers:

• CLI (Command Line Interface)
  Humans typed commands designed for human memory and usage.
• API (Application Programming Interface)
  Structured endpoints designed for machines to call.
• AI interfaces (emerging)
  Systems where humans express intent, and AI translates it into actions.

We are now moving toward what can be called:

Intent-driven interfaces

APIs Became the Foundation

In the API era, systems exposed capabilities like this:

GET /users?age=18

APIs are powerful because they are:

• Structured
• Composable
• Machine-friendly

But they still require step-by-step orchestration.

Enter DSLs: Languages for Intent

A DSL (Domain-Specific Language) is a language designed for a specific problem.

Examples include:

• SQL for querying data
• Configuration files such as YAML and JSON for defining systems

Instead of writing procedural code, DSLs let you declare intent:

checkout:
  payment: stripe
  currency: EUR
  retry: true

The Key Shift: From Execution to Description

Old model	New model
Code = instructions	Code = intent
Human writes logic	AI generates logic
APIs are used directly	DSLs describe what APIs should do

Why AI Prefers DSLs

DSLs are:

• Constrained
• Predictable
• Structured

This makes them ideal for AI systems that map intent to implementation.

DSL + API: Not Competing, But Layered

Human / AI
   ↓
DSL (intent)
   ↓
AI / Interpreter
   ↓
API calls
   ↓
System execution

What This Means for Builders

1. APIs are your foundation.
2. DSLs are your interface.
3. Systems should be designed for AI operability.

Final Thought

The most important language is not the one humans write, but the one AI understands best.

TL;DR

The future of software is not about writing better code.
It is about designing better ways to describe what should exist.