The future of shopping.

Every shopping site you use is a catalogue with a search box bolted on.

You browse categories, you filter, you scroll, and the thing you actually wanted is somewhere in a grid of four hundred approximations.

Nobody wants a catalogue. People want a specific thing, at a fair price, arriving before they need it. The catalogue exists because it is what a database of products naturally looks like, not because it is what anybody asked for.

Set a budget and you will still be shown things above it.

Say you need it this week and you will be shown things that arrive later, quietly reordered rather than excluded.

This is deliberate. An engine optimising for the chance that you buy something has every reason to show you the nearly-right thing. It is a reasonable default and a bad one, and it is why filtering feels like negotiating. Try both.

The same physical object is routinely sold by a dozen resellers at prices that differ several times over.

None of them made it, none of them photographed it, and none of them will tell you the others exist.

This is not a scandal so much as a structural fact. Resellers add real things sometimes: faster shipping, a returns policy, support worth having. What is not defensible is that the difference is invisible.

Price comparison engines join listings by barcode, brand or title.

Branded goods have a GTIN, so they compare well, and branded goods already have narrow spreads, because they are easy to compare.

Unbranded goods have no barcode, forty sellers and forty keyword-stuffed titles. They also have the widest spreads. Every comparison tool is structurally blind to precisely the products where comparison is worth the most.

Resellers do not photograph their stock, because they never hold it.

They upload the supplier's media pack unchanged, so the identical image file appears on every listing of the same object, across every marketplace.

That makes the photograph the one genuinely shared identifier for unbranded goods. Matching it is closer to comparing checksums than to judging similarity, and it works exactly where barcodes do not.

Measured on real product photography: 98% recall with zero false matches across 1,762 true negative pairs, at a threshold of 10 bits out of 64.

You describe what you want in a sentence.

Every constraint in it is a gate: met, missed, or honestly reported as unmeasurable. You get one answer, the cheapest thing that actually satisfies what you said, and beside it the cheapest thing overall, with the reason it does not qualify.

Both photographs are shown, so the claim that they are the same object is something you verify in half a second rather than take on faith.

No catalogue. No grid. No negotiating with a filter.

A comparison engine that takes money for placement is not a comparison engine.

The moment one result can be bought, every result becomes a question, and the product is worth nothing.

So there is no advertising product and there will not be one. Some outbound links earn a commission; every page carrying them says so, and the ordering is by price and relevance only. The code that attaches a tracking tag receives a finished link and cannot see position, so ranking by payment is not possible here even by accident.