// Product Inputs
System
~240
User Input
—
Max Output
800
[ _ ]

Fill in the product details and click generate to produce three A/B/C copy variants.

Generating Variants
Single API call · structured output
3 Variants Generated
— tokens used
// Selected variant — full copy
//

How It Works

01

System Prompt Design

A ~240 token system prompt enforces output schema, caps every field at hard word limits, locks tone to professional/aspirational apparel, and requires clean JSON — no preamble. This is the PM craft: maximum output quality from minimum token spend.

02

Structured Output

Each of the 3 variants returns 4 fields: tagline (≤8 words), headline (≤12 words), body (≤30 words), CTA (≤5 words). Schema enforcement via prompt ensures parseable JSON every time — no post-processing regex needed.

03

Token Efficiency

Total budget: system (~240) + user input (~60) + output (≤800) = ~1,100 tokens per call. No multi-call pipelines, no RAG, no tool use. Single synchronous call via Cloudflare Worker proxy — demonstrating that great AI products don't need complexity to deliver value.