After playing with AI-assisted coding tools for a couple of months, a pattern becomes hard to ignore.

At first, it feels like autocomplete on steroids. Then it starts to feel like pair programming. Eventually, something clicks: the way you think about writing code changes.

You stop obsessing over the exact control flow and start spending more time on:

* clearly describing the problem * identifying edge cases upfront * defining what "correct" actually means

That shift mirrors something we’ve already seen before in programming history.

---

## A Familiar Parallel

There’s a useful mental model for understanding AI coding tools:

> *AI‑assisted coding : human coding :: declarative programming : imperative programming*

This analogy isn’t perfect—but it’s surprisingly powerful.

Imperative programming is about *procedures*. Declarative programming is about *intent*.

AI‑assisted coding pushes software development in the same direction declarative paradigms did: away from micromanaging execution, toward expressing goals and constraints.

---

## Imperative Coding: Telling the Computer How

Imperative code is explicit, linear, and mechanical:

```js let sum = 0; for (let i = 0; i < nums.length; i++) { if (nums[i] % 2 === 0) { sum += nums[i]; } } ```

You specify:

* control flow * intermediate state * order of operations

This mirrors *traditional human coding*:

* Read requirements * Decompose into steps * Implement each step manually

The human is responsible for every decision: loops, branches, data flow.

---

## Declarative Coding: Stating What You Want

Now compare that to a declarative style:

```js const sum = nums.filter(n => n % 2 === 0) .reduce((a, b) => a + b, 0); ```

Here you describe:

* what qualifies (even numbers) * what result you want (a sum)

You don’t care how iteration happens.

SQL, React, CSS, functional pipelines—all follow this pattern.

The system figures out execution details.

---

## AI‑Assisted Coding Is Declarative Human Input

When you use AI, your role subtly changes.

Instead of writing:

> “First do X, then Y, then Z…”

You write:

> “Given this context, implement a function that does *X*, respecting *Y* and *Z*.”

Example prompt:

> “Write a Go function that parses DNS trace output and builds a graph of resolution hops.”

That’s declarative.

You specify:

* intent * constraints * shape of the output

The AI fills in loops, conditionals, data structures.

Just like a SQL engine or React reconciler.

---

## Humans Move Up the Abstraction Ladder

With imperative coding, humans live at the execution layer.

With AI assistance, humans move upward:

* defining invariants * describing interfaces * naming concepts * evaluating correctness

This is the same shift that happened with:

* assembly → C * C → higher‑level languages * manual DOM → React

AI doesn’t remove abstraction—it *adds another layer*.

---

## The New Skill: Prompting Is Specification Design

Good AI usage looks a lot like good declarative API design:

Bad prompt:

> “Write some code to do DNS stuff.”

Good prompt:

> “Given this input format, produce a deterministic DAG where each hop is a node; preserve ordering; handle NXDOMAIN explicitly.”

This mirrors:

* good SQL schema design * good React component boundaries * good FP type signatures

The better the spec, the better the output.

---

## AI Doesn’t Replace Imperative Thinking—It Compresses It

The imperative steps still exist. They’re just compressed into the model.

You’re trading:

* line‑by‑line control for * higher‑level leverage

Just like calling `Array.map` instead of writing a loop.

Eagle analyzes your Windows programs and suggests a Wine setup