Couscous

Podcast 2 — In conversation

Speaker A: I made couscous for the first time last night and I cannot believe how quick it was. I just poured boiling water over it, covered it for five minutes, and it was done. I have been overcomplicating dinners for years.

Speaker B: That is the instant couscous you get in most supermarkets — the pre-steamed kind. It is convenient, but traditional couscous is a completely different process. In Morocco they steam it multiple times in a couscoussier, breaking up the clumps by hand between each steaming. It takes well over an hour.

Speaker A: So the instant version is to real couscous what instant noodles are to ramen — technically the same category but a very different experience.

Speaker B: Exactly. And here is the thing most people get wrong: couscous is not a grain. People put it alongside rice and quinoa as if it belongs in the same family, but it is actually pasta. It is made from semolina — crushed durum wheat — mixed with water and rolled into tiny granules. It is closer to orzo than to rice.

Speaker A: That genuinely surprises me. I have always thought of couscous as a grain. Every salad bar I have ever been to files it under grains next to bulgur.

Speaker B: And that is a useful comparison. Bulgur is a cracked wheat product — it actually is a grain, because it is whole wheat kernels that have been parboiled and crushed. Couscous is manufactured from milled flour. They look similar, they cook differently, and nutritionally they are distinct. Bulgur has more fibre because the bran is intact.

Speaker A: When I ordered couscous at that Tunisian restaurant in Soho, it came as the base of a massive stew — lamb, vegetables, chickpeas, a spicy broth poured over the top. They called it couscous but the word seemed to refer to the entire dish, not just the grain — the pasta, rather.

Speaker B: That is exactly how the word functions in North Africa. Couscous refers both to the semolina granules themselves and to the complete dish — the steamed base plus the stew on top. When a Moroccan family says we are having couscous on Friday, they mean the entire meal, not just the grain component. Context determines which meaning is active.

Speaker A: So when I say I added couscous to my salad, I am talking about the ingredient, but when a Tunisian says come for couscous, they mean a full ceremonial dish.

Speaker B: Precisely. And the Friday couscous tradition is deeply significant — in Morocco, Algeria, and Tunisia, couscous on Friday is a communal family meal, often after the midday prayer. It has social and cultural weight far beyond its nutritional content. It is the dish that gathers the family.

Speaker A: The word semolina keeps coming up. What exactly is semolina in relation to couscous?

Speaker B: Semolina is the coarse flour milled from durum wheat. It is the raw material from which couscous is made. You sprinkle water over semolina, roll it between your palms to form the tiny pellets, sieve out the ones that are the right size, and steam them. Semolina is also used to make pasta, certain breads, and puddings. So couscous is one product of semolina, not the other way around.

Podcast 3 — Prompt engineering

Speaker B: Today we are going to write AI prompts that revolve around couscous — not as a metaphor, but as the actual subject. Couscous is ideal for this exercise because it sits at the intersection of food, culture, and data. A recipe app, a food delivery platform, a cultural education tool — these are real, buildable products, and couscous gives us rich domain content to work with.

Speaker A: So we are using couscous as the concrete domain for real application prompts — UI design, database architecture, the full stack?

Speaker B: Exactly. Let us start with a recipe management dashboard.

Speaker B: PROMPT: Build a recipe management dashboard where users can search, filter, and save couscous recipes by region — Moroccan, Tunisian, Algerian — using a warm earthy colour palette with card-based layout and smooth filtering animations.

Speaker B: The prompt gives the AI a clear domain — couscous recipes — a data structure — regional categorisation — and a visual direction — warm earthy tones and card layout. The regional filter is not arbitrary; Moroccan, Tunisian, and Algerian couscous genuinely differ in preparation, grain size, and accompaniments. The AI gets real-world categories to work with.

Speaker A: The earthy palette choice mirrors the ingredient itself — golden semolina, warm spices. That kind of domain-appropriate design direction helps the AI produce something coherent rather than generic.

Speaker B: Good observation. Now a database schema prompt.

Speaker B: PROMPT: Design a database schema for a food delivery platform that categorises couscous dishes by grain size — fine, medium, and large — with relational tables linking ingredients, preparation time, and dietary tags.

Speaker B: Grain size is a genuine classification in couscous — fine grain is common in Libya, medium is the Moroccan and Algerian standard, and large pearl couscous, sometimes called maftoul or Israeli couscous, is a distinct product. The AI gets a real taxonomy that maps naturally to database categories, not an invented one.

Speaker A: And the relational structure — linking ingredients, prep time, and dietary tags — is a standard pattern for food platforms but grounded here in couscous-specific data. The schema writes itself because the domain is concrete.

Speaker B: Exactly. A UI component prompt next.

Speaker B: PROMPT: Create a restaurant menu component with a dedicated couscous section featuring an expandable accordion panel, showing each variation with thumbnail images, spice level indicators, and a one-click order button.

Speaker B: This prompt gives the AI a specific UI pattern — accordion panel — a specific content domain — couscous variations — and concrete interactive elements — spice indicators and an order button. The couscous section is not a generic food section; it implies multiple variations, which drives the need for the accordion pattern in the first place.

Speaker A: The spice level indicator is a nice touch — North African couscous varies enormously in heat. A harissa-heavy Tunisian dish is a completely different experience from a mild Moroccan vegetable couscous.

Speaker B: Right. Now a standalone utility application.

Speaker B: PROMPT: Generate a couscous cooking timer application with a minimal interface — a single circular progress indicator, adjustable steaming duration, and a notification bell that chimes when the couscous is perfectly fluffed.

Speaker B: This is a focused, single-purpose application prompt. The AI knows the domain — couscous cooking — the interface constraint — minimal, one circular indicator — and the completion signal — a notification chime. The phrase perfectly fluffed is intentional; it tells the AI that the timer is not just counting down but is tied to a specific culinary outcome.

Speaker A: It is a small app, but the specificity means the AI will produce something polished rather than a generic timer with a food label stuck on it.

Speaker B: An inventory management prompt next.

Speaker B: PROMPT: Build an inventory tracking module for a wholesale food supplier where couscous stock levels trigger automatic reorder alerts through a colour-coded side panel — green for sufficient, amber for low, red for critical.

Speaker B: This is a business logic prompt. Couscous stock levels as the trigger gives the AI a concrete product to track, and the colour-coded alert system — green, amber, red — is a universal UX pattern applied to a specific supply chain context. The side panel constraint shapes the layout without over-specifying.

Speaker A: The wholesale context is important — couscous is shipped in bulk, so stock thresholds, reorder quantities, and supplier lead times are all real concerns that the AI can model.

Speaker B: Good. A cultural education platform next.

Speaker B: PROMPT: Design a cultural food education platform where the couscous page includes an interactive map showing North African regions, with hover tooltips explaining local preparation traditions and a step-by-step visual recipe guide.

Speaker B: This prompt combines data visualisation — the interactive map — with educational content — regional preparation traditions — and practical instruction — the recipe guide. Couscous is perfect for this because the regional variations are genuine and well-documented. The hover tooltips give the AI a specific interaction pattern to implement.

Speaker A: The map anchors everything geographically. A user hovering over Tunisia gets different information from someone hovering over Morocco, and that regional specificity is real, not fabricated for the exercise.

Speaker B: One final full-stack prompt.

Speaker B: PROMPT: Create a meal planning calendar application where users can drag and drop couscous-based meals into weekly slots, with automatic nutritional breakdown calculations and a smart grocery list generator.

Speaker B: This is the most complex prompt — drag-and-drop interaction, nutritional computation, and list generation. Couscous-based meals as the content domain means the AI has a specific food category to work with, and the grocery list generator implies ingredient decomposition — breaking a couscous dish down into its component ingredients and aggregating across the week.

Speaker A: That grocery list feature is the most technically interesting part. If I plan three couscous meals in a week, the generator needs to consolidate overlapping ingredients — semolina, chickpeas, harissa — rather than listing them three times.

Speaker B: Exactly. And that is the power of using a real food domain in prompts. Couscous is not a placeholder word in any of these prompts — it is the subject that drives the data model, the UI decisions, and the business logic. Every prompt we wrote today would produce a different, weaker result if you replaced couscous with a generic term like food item. The specificity is what makes the prompt effective.