The Cook Does Not Want Pages

I was not thinking about media economics. I was making supper.

In my kitchen in northern New Mexico, that means real constraints, not the decorative sort. Two servings, not six. Half-pound bison patties, because that is what I intended to cook. Hatch green chiles, fried whole, one on the burger and one on the side per plate, because that is the canonical way it is done at The Mine Shaft Tavern in Madrid, New Mexico. A sourdough brioche bun already developed for high desert altitude and strong enough to hold a half pound bison cheeseburger without turning into supermarket mush. Tomato wide on top, arugula from the garden. Sharp cheese on the patty, Mendocino mustard and ketchup below. Topping order adjusted because topping order matters. This is what actual cooking looks like. It is not the retrieval of a sacred text. It is iterative fit under local conditions, plus inspiration and taste. A tuned recipe, herein appended.

That is why recipe websites have become such a clear casualty of the new computational regime. They were built for a different machine. The older search economy assumed that knowledge lived on relatively static pages and that the user’s task was to locate the correct page, extract the relevant procedure, and adapt it by hand. The economic model followed from that assumption. Publishers produced documents. Search engines ranked them. Users clicked through. Money was made from pageviews, display ads, affiliate links, and the various swamp flowers of search engine optimization (SEO). Recipe sites became one of the purest expressions of this arrangement. Over two decades they evolved into long, ad-heavy scroll engines padded with autobiography, lifestyle theater, substitution notes, and sentimental throat-clearing. This was not merely bad taste. It was business logic. The cook’s delay was the product; eyeballs trapped into scrolling across pages before arriving at a recipe that was not quite right.

The familiar ritual was absurd but stable. One observed a search result page, oriented through titles and snippets, decided which blue link might conceal the needed answer, acted by clicking, and then repeated the cycle across tabs and ad-ridden pages. In Boyd’s terms, the user’s OODA loop was stretched and taxed at every stage. Orientation took time. Action was delayed. The path from intention to useful result had to be made inefficient enough to throw off a profitable number of impressions, a dark pattern that emerged over the years as an economic basis. Thus the web achieved that singular form of progress in which one consults an article called “Easy Weeknight Pasta” and first receives a minor Tuscan Bildungsroman, three autoplay ads, and a testimonial to resilience in the face of basil.

But conversational AI changes the structure of the task. The first answer is not the answer. It is a draft. The real work lies in tuning. Make it for two. No cilantro. I only have thighs. My oven runs hot. Convert to grams. Less sweet. At 7,000 feet. Explain why the emulsion broke. What the user wants is not primarily retrieval of a static artifact, but rather help fitting procedure to circumstance. Recipes are especially exposed because they are both highly compressible and highly parameterizable. They are not novels. They are compact procedural objects to be adjusted against reality. AI can do that.

That changes the economic model because it changes where value resides. Under the old arrangement, value sat in the page and in the traffic sent to the page. Under the new arrangement, value migrates toward the adaptive layer that can hold context, preserve constraints, and revise the answer in real time, conversationally, with pauses and iterations. The old web monetized traversal; the new layer monetizes convergence. The recipe page is demoted from destination to source material, often raw material. The monetizable pageview is bypassed and eyeballs no longer traverse the cluttered landscape. The memoir remains loyally in place, like an officer standing at attention after the fortress has already fallen, but no one reads its southern girl anecdotes.

To describe this only as a faster search engine misses the real break. Search assumed that the answer already existed in finished form somewhere out there, waiting to be found. Conversation assumes that the answer will be produced through interaction. The valuable unit is no longer the static page but the answer trajectory: the evolving exchange in which constraints are added, clarified, and satisfied. The old web organized itself around discoverability. The new systems organize themselves around tunability.

This is where a Boydian frame helps, but only if it is revised as a structural inversion. The user still has an OODA loop. But now so does the computational entity. The system observes the user’s requests, pantry, preferences, equipment, corrections, and local facts. It orients by building a provisional model of the task and of the user’s situation. It decides on a next move. It acts by producing a revised answer. The user then observes that answer, re-orients by noticing what is right, wrong, missing, or newly possible, decides on a correction or refinement, and acts again. Two loops are in play. Each is part of the other’s environment.

But this relation is not best understood as competition. It is not a dogfight. Here the better biological analogy is Lynn Margulis and symbiogenesis. Margulis’s great insistence was that durable novelty can arise not only from struggle and selection in the narrow sense but from distinct organisms entering into cooperative relation and producing a new level of capability. The entities do not melt into one another. They remain distinct. Their substrates, histories, and failure modes remain different. Yet under certain conditions they form a composite that can do something neither one, operating alone in the old way, would do as well. It is emergent.

That is a better description of the kitchen case. I remain the cook. The machine does not smell the chile oil, does not notice the exact spring of the bun under my thumb, does not eat the finished burger. But for the duration of the task it enters into a local cooperative coupling with me, different from anyone else’s interaction. I provide goals, constraints, taste, ingredients, corrections, and judgment. The system provides rapid recombination, memory across turns, procedural variation, and immediate adaptation. Each becomes part of the other’s active environment. Over several turns, the pair can settle into something like a cooperative lock. Not merger, and certainly not romance. Symbiogenesis is the right smell; Hallmark is not.

Seen this way, the economic shift becomes clearer. The search-era web made money by forcing the user to traverse pages. It monetized obstruction, and slowed action. It rewarded the ownership of bottlenecks. The conversational system makes its bid by joining the task and shortening the path through it. That does not make it altruistic. It merely means the profitable loop has moved. The user is no longer prey wandering through a field of optimized snares. The system wins by helping the user converge, and by keeping that convergence inside its own adaptive layer.

This is why classic SEO and PageRank begin to look archaic. They belonged to a world in which the central problem was to determine which fixed document should be surfaced first. Their hidden constitution was page-centric. In a conversational regime, the page remains as a source of evidence or authority, but it no longer serves as the primary unit of value capture. The important question is no longer, “Which page should this person visit?” It is, “How should this adaptive system continue the exchange so as to converge on a useful result?” Ranking yields to orchestration. PageRank may survive as one signal among others, but it no longer governs the whole geometry of the web. It is an ornate canal map in the age of railroads.

The recipe site is not a trivial edge case. It is the pure demonstration of a broad transformation. It shows what happens when a monetized presentation layer is overtaken by a conversational tuning layer that better matches the actual structure of the task. The cook does not want pages. The cook wants supper. The old web could only approximate this by publishing a frozen recipe plus a great trailing wedding train of SEO bait and display inventory. The new system can collaborate in the actual adjustment process. Once a user experiences that, the old format begins to feel not authoritative but cumbersome, even faintly ridiculous.

And that takes me back to my own stove.

In my kitchen, the recipe was never really a static page. It was always a starting point, an inspiration, and an instrument under revision: adjusted for altitude, ingredients at hand, desired structure, serving count, and taste. I found recipes, tuned them, edited them, and saved them. What has changed is not that cooks suddenly discovered variation. Cooks have always varied. What has changed is that the computational system can now join that revision loop in a usefully active way. The old web made money by forcing the cook to traverse pages. The new system makes its claim by helping the cook converge on delight.

The recipe that follows, then, is not merely an appendix or a decorative flourish. It is evidence. It is an artifact of this change: a tuned Hatch green chile bison burger, developed not by passive lookup but by iterative fit in a real kitchen. The hearth is not separate from the argument. It is where the argument becomes true, and gets baked in.

The Recipe

Yield: 2 burger plates Serve: one whole fried chile on each burger, one whole fried chile on the side, per serving

Intent

A whole roasted Hatch green chile, lightly scored, dredged, battered, and fried crisp enough to sit on a half-pound cheeseburger without collapsing. This is not chopped green chile and not a chile strip. The canonical plate is two whole fried chiles per serving: one on the burger and one on the side.

Ingredients

Chiles

• 4 whole roasted Hatch green chiles, peeled, stems left on if possible
• 1/4 cup all-purpose flour, for dredging
• Flaky salt, for finishing

Batter

• 1/3 cup all-purpose flour
• 2 tablespoons cornstarch
• 1/4 teaspoon baking powder
• 1/4 teaspoon kosher salt
• Pinch garlic powder
• Pinch smoked paprika, optional
• 1 small egg, or 1/2 large beaten egg
• 1/3 cup very cold beer, club soda, or sparkling water
• Neutral oil for frying: avocado, peanut, or canola

Burger Plate

• 2 brioche buns, toasted
• 2 half-pound bison patties
• Irish grass-fed aged cheddar
• Onion, Mendocino mustard, catsup, and garden arugula

Preparation

Dry and score. Pat the roasted, peeled chiles very dry. Lay each chile flat and make 2 or 3 shallow diagonal cuts on one side only, about 1/4 inch deep at most. Do not crosshatch and do not cut the chile into pieces.

Heat the oil. Heat 1 1/2 to 2 inches of oil to 365°F. Use a small heavy pan so the chiles fry quickly and the oil recovers fast.

Mix the batter. Combine flour, cornstarch, baking powder, salt, garlic powder, and paprika. Beat egg with cold beer or sparkling water. Stir wet into dry only until mixed. Lumps are fine.

Dredge and fry. Lightly dredge each chile in plain flour and shake off excess. Dip in batter, let excess fall for a second, then slide into the oil. Fry 2 to 3 minutes total, turning once, until golden and crisp. Drain on a rack. Salt at once.

Build the burger. Sear bison hot and fast. Add cheddar late and cover briefly. Toast brioche. Assemble with mustard and catsup on the bun, onion, patty, melted cheddar, one whole fried Hatch chile, arugula, and the top bun. Put the second whole fried chile on the side of the plate.

Notes

The batter should be lighter than pancake batter but heavier than tempura. If it runs off the chile, add a spoonful of flour. If it turns bready, thin it with a splash of cold beer or sparkling water. The decisive move is dryness: wet roasted chile sheds batter and steams.

The arugula goes above the chile, not under the hot patty. It stays greener and gives a sharp bitter snap against the cheddar and fried Hatch.

Nova Lux kitchen sheet