July 2026CompositionThe market

Your Menu Already Contains Products You Haven't Found

Popeyes' chicken sandwich existed as a valid combination of things they already stocked — for years — before a trend found it by accident and made them a fortune. We map a business's whole option space in advance, so that when a trend hits, the answer is a lookup, not a product-development cycle. This is the worked demonstration, with the real numbers.

The Popeyes problem

In 2019, Popeyes launched a chicken sandwich that broke the internet, sold out nationally, and reportedly added hundreds of millions in value. Here is the uncomfortable part: every ingredient in that sandwich had been in their kitchens for years. The product existed as a valid point in their own option space the whole time. Nobody had enumerated the space, so nobody knew it was there. A trend found it by accident.

That is not a Popeyes story. That is the normal condition of almost every business: you do not know your own configuration space. Your products, your services, your bundles, your pricing tiers — they are combinations drawn from option sets you already own, and the combinations you currently sell are a tiny, historically-accidental subset of the ones you could sell tomorrow at zero new supply cost.

The demonstration

To show the method, we ran it on a menu everyone knows: In-N-Out's. (To be clear: this is a demonstration built from publicly known menu information. In-N-Out is not a client and is not affiliated with us — we picked them because their menu is famously small and their secret menu is famously real, which makes them the perfect test case.)

We modeled the burger as six option dimensions — patties, cheese, bun, onions, spread, extras — and had the engine enumerate every valid combination: 972 configurations. Then we marked what's already claimed:

• The published menu claims 3 points — Hamburger, Cheeseburger, Double-Double.
• The secret menu claims about 3 more — and this is the part that should stop you cold. Animal Style was never on the menu. Customers found it in the space, named it, and it became a cultural asset worth millions in free marketing. The secret menu is real-world proof that the valid space exceeds the published menu, and that the market will pay for points in the dark region.

That leaves roughly 966 valid combinations nobody has claimed. Not hypothetical products — combinations of ingredients already in every store, buildable today, with zero new supply chain.

Then a trend hits — and this is where it gets good

In the demonstration, we simulated a trend event: triple-patty stacks start trending. Watch what the response looks like when the space is already mapped:

1. Filter the mapped space: valid combinations, three patties, not already claimed by the menu or the folklore.
2. Rank the candidates and certify the winner as the trend response.
3. Done.

No concept phase. No product development cycle. No supply-chain change. The trend was answered by lookup, because the product already existed in the mapped space — exactly the thing Popeyes could not do, because their space was unmapped. Trend response collapses from months to a query.

Three more properties of the method matter to an operator:

• The core menu never mutates. New waves are generated as separate, certified collections — the flagship space is provably untouched (in the demonstration we verify this by comparing a full digest of the space before and after; byte-identical). Zero brand risk.
• New ingredients get their own sibling space. Want to pilot a BBQ sauce in limited stores? That's a new space scoped to those stores — and the engine counts the delta exactly: adding one spread option took the space from 972 to 1,296 combinations, and the 324 new points are precisely the BBQ configurations. The pilot is contained by construction.
• The same move works above the menu. In the demonstration we composed the product space with a pricing space and generated the ad-campaign spec for the chosen trend response — the whole chain from "what could we sell" to "here is the campaign," on one mapped substrate.

What's real here, exactly

The engine, the enumeration, the certification steps, and the immutability checks are real — the demonstration is a captured, replayable run on the live engine, every step passing its validation gate, and it can be re-run end to end from a script. The cost and margin figures used for ranking in the demo are illustrative (we don't have In-N-Out's real costs, obviously), and the trend event is simulated. What the demonstration proves is the method: map the space in advance, and trend response becomes selection.

"Why wouldn't I just ask ChatGPT or Claude to do this?"

Fair question — and you can. Hand a frontier model the In-N-Out menu and it will happily write you a combinatorics calculator. Here's what you'll actually have, and what you won't:

• You'll have an answer. You won't have a place. The model's calculation lives and dies inside one chat. Our engine reifies the space as a durable, addressed object: every combination has a permanent coordinate, certifications are recorded and never silently undone, and the map is still there — identical — next quarter. The model re-derives; the substrate remembers.
• You'll have a calculator. You won't have navigation. A mapped space is something you walk: stand at your current configuration, see the valid neighbors, inspect any point's full decision trail, filter the unclaimed region, watch the frontier where certified meets unexplored. That interactivity changes what the AI can do with you — every next step is grounded in where you actually stand, instead of re-explained from scratch each conversation.
• You'll have one tool. You won't have one grammar. Ask a model for a menu tool, then a pricing tool, then a campaign tool, and you get three bespoke, incompatible artifacts. Our engine is one encoding for every domain — the same coordinates, the same certification, the same math — which is why the demonstration composes the menu space with a pricing space and generates the campaign spec on the same substrate. It stays organized on its own, structurally — organization isn't a chore the model must remember to do; it falls out of how the space is built.
• And to be direct about the elephant: the frontier labs don't make this. It isn't a bigger model — it's the world a model thinks in. We use frontier models inside this system every day; the model is the generator, the map is ours. A better model makes our engine better. It never replaces it.

What this means for your business

You are not a burger chain — but you have a configuration space. Service packages. Product bundles. Feature tiers. Onboarding paths. Price points. Whatever you sell, it is combinations over option sets you already own, and right now you are selling the handful of points history happened to hand you.

What we do: model your option space, enumerate it, certify the points you already claim, and hand you the map of the dark region — so the next time the market moves, you respond with a lookup while your competitors schedule a product-development meeting.

The secret-menu lesson is the one to keep: if you don't map your space, your customers — or your competitors — will find the valuable points in it before you do.