Skip to content
Jack Smyth/Differentiation/Facts with Flair
Experiment 01 · differentiation

Facts with flair.

Same product. Same specs.
A clearer reason to choose.

A practical test of differentiation: help a shopping agent see a buying decision, not another list of interchangeable features.

01 / See the difference

A product, reframed.

Eight Australian retailer examples. See the source, the rewrite and exactly what changed before trying your own.

Sony WH-1000XM5 Wireless Noise Cancelling Headphones Black

Officeworks · checked 2026-10-29
The source

You can pair with up to 2 devices at a time via Bluetooth or connect directly via cable. Enjoy up to 30 hours of playtime on a single charge, and quick power thanks to USB-C quick charging. The noise cancelling feature uses 4 microphones on each earcup to keep out more high and mid frequency sounds than ever. A carrying case, 1.2m headphone cable and USB-C charging cable are also included.

Open the retailer source ↗
A different frame · same source facts

Keep your headphones. Change the connection.

Make the connection the thing you change, not the headphones. They pair with up to 2 devices at a time via Bluetooth. For listening between charges, there's up to 30 hours of playtime on a single charge, plus USB-C quick charging when it's time to plug in. Noise cancellation uses 4 microphones on each earcup.

Playtime · up to 30 hoursPairing · up to 2 devicesNoise cancellation · 4 microphones on each earcupCharging · USB-C
What changed

Makes multi-device pairing the organising idea, while separating listening duration, charging and noise-cancelling hardware.

Pre-generated with Flint, editorially refined and separately reviewed. This is not the live six-gate QA certification.

02 / Try the same thinking

Your product. The same constraint.

Flint reads the page, writes from its extracted facts, then checks the result. Usually 1–3 minutes. Review before publishing.

Powered by Flint from Springboards

The method is part of the point.

The examples are dated source snapshots, not live prices or stock. Flint drafted; editorial refinement and a separate Flint review improved the framing. No retailer endorsement, market-uniqueness proof or ranking uplift is implied.

Live runs preserve exact source values and return specifications, narrative, three “Why this one” lines, agent-readable attributes and six QA gates. A failed gate stays visible. Automated QA checks the extracted source; it is not independent verification. Page reads and generated output are not persisted by this core app.

The experiment runs here; the existing Facts with Flair service handles page extraction, model calls and signed snapshots. Flint from Springboards provides the generative model. Shopping research and ShoppingComp inform the emphasis on explicit, comparable facts—not a claim that a phrasing change causes better rankings.

Use this experiment with your agent

The identity MCP's get_work tool, id w-fwf, explains the experiment. The Markdown guide and example data expose exactly the same source-backed comparisons.