benchmarks · methodology

Smaller definition.Lighter review loop.

The examples below are computed from source that ships in the page HTML. They show how a compact axint definition compares with the Swift an agent or developer would otherwise need to generate, inspect, and regenerate for the same Apple-native feature. Less generated code to review, less churn to diff, more confidence when the team has to decide whether a feature is safe to ship.

live comparison

Three definitions, three Swift outputs.

defineIntent()

App Intent that runs in Siri

TypeScript16 lines
116tokens
import { defineIntent, param } from "@axint/compiler"; export default defineIntent({ name: "CreateCalendarEvent", title: "Create Calendar Event", description: "Create a calendar event",...
Swift36 lines
225tokens
import AppIntents struct CreateCalendarEventIntent: AppIntent { static var title: LocalizedStringResource = "Create Calendar Event" @Parameter(title: "Event Title") var title: String...
Token reduction ratio
1.9x

defineView()

SwiftUI view component

TypeScript20 lines
164tokens
import { defineView, prop, state, view } from "@axint/compiler"; export default defineView({ name: "ProfileCard", props: { displayName: prop.string("User's display name"),...
Swift61 lines
318tokens
import SwiftUI struct ProfileData { let name: String let bio: String let avatar: String...
Token reduction ratio
1.9x

defineWidget()

iOS/macOS widget

TypeScript20 lines
137tokens
import { defineWidget, entry, view } from "@axint/compiler"; export default defineWidget({ name: "StepCounter", displayName: "Step Counter", description: "Shows your daily step count and progress toward your goal.",...
Swift144 lines
817tokens
import WidgetKit import SwiftUI struct StepData: Codable { let steps: Int let goal: Int...
Token reduction ratio
6.0x

workflow size

What ends up in front of the reviewer.

225
116
Intent
318
164
View
817
137
Widget
TypeScript
Swift

the numbers

What this changes in practice.

69%

Less definition overhead to carry

Across the public intent, view, and widget fixtures

~314

Less generated Swift to inspect per feature

Average gap between axint source and emitted Swift

~31,400

Tokens kept out of 100 repair + review loops

How the smaller definition compounds across repeated iterations

what changes for a team

Lighter review, cleaner handoff.

01

Smaller review burden

When AI output regenerates, reviewers do not have to reread the full Swift output from scratch. The compact axint definition stays smaller, so the review loop stays lighter.

02

Cleaner handoff to Apple reviewers

Teams still get ordinary Swift at the end, but they carry less generated code through the authoring and repair loop before it reaches a shipping decision.

03

Better fit for Cloud validation

Smaller definitions are easier to regenerate and compare, which makes validation, regression review, and shipping checks much more legible inside Cloud.

methodology

Inspect every number.

01

What this page measures

Smaller Apple-native definitions. Each comparison estimates tokens for the axint source and the emitted Swift using a fixed 4-characters-per-token approximation.

02

What it does not measure

This page does not claim compile throughput, universal savings on every Apple feature, or exact provider billing. It is a narrow methodology page, not a catch-all performance claim.

03

How to rerun it

Use the repo benchmarks for compiler throughput, inspect the examples we use as reference fixtures, and read the methodology note that explains where the public numbers come from.

why this matters beyond compression

Apple-native iteration gets heavy fast.

This page isn't about shaving five dollars off a bill. It's here to show that smaller definitions leave less output to inspect, resend, diff, and debug.

axint changes the workflow shape: one defineIntent() call generates production Swift. The agent or developer works against a compact definition, then carries the generated Swift only when it's time to validate, review, or ship.

Result: 69% less definition overhead and 3.3× average compression across the public samples on this page. That doesn't replace validation about bugs caught or shipping readiness — it supports it.

ready to ship

Move from benchmarksto workflow.

Start with the compiler walkthrough, then use Cloud when you want validation, report URLs, and saved history around a real Apple-native feature.