Modern Retro started with a stupid question: what would a Supreme store look like in 1974?
Not as a joke. As a genuine creative exercise. What would the signage look like? The interior? The products on the shelves? What would feel right about it, and what would feel completely wrong? That tension - between a brand that couldn't possibly exist in that era and the physical reality of a 70s high street - was the whole idea.
I made one image with AI. Then another. Then I couldn't stop.
The creative constraint
Every project I've built has a constraint at its core, and Modern Retro's is specific: Wes Anderson symmetry, Kodachrome film warmth, 1970s retail architecture. That's the visual language. No exceptions. No "let's try a different style for this one." The discipline of staying within that constraint is what makes the whole thing cohere.
It also makes the AI outputs better. When you give an image generator a vague prompt, you get vague results. When you give it precise art direction - "symmetrical storefront, tungsten interior lighting, period-appropriate signage, warm brown and beige tones, slightly kitschy" - you get something that actually looks considered.
The scoring system nobody asked for
Here's where my strategy brain kicked in. I didn't want to just pick brands randomly. I wanted a framework - a way to evaluate which brands would create the most interesting images. So I built the MR Score.
Every brand gets scored 0-100 across five dimensions:
Infrastructure Invisibility - how dependent is the brand on digital systems that didn't exist? A hardware store scores low. A fintech app scores high.
Product Intangibility - can you physically hold the product? LEGO scores low. Spotify scores high.
Cultural Temporal Distance - how alien would this brand feel in 1975?
Retail Translation Friction - how hard is it to imagine this as a physical shop?
Semiotic Shock - how jarring would the brand name and visual identity feel in that era?
The higher the score, the better the content. Brands in the "Deep Red" tier (81-100) - like ChatGPT, Perplexity, Cursor - create the most visually absurd images. Brands that score below 20 got removed entirely. Nike and Adidas? Too boring. They already could've existed in the 70s. That's not the point.
The taste decisions
96 brands live on the site now. But the interesting number isn't what I included - it's what I removed. Nike, Adidas, and a few others were deleted because they scored too low on the absurdity scale. They weren't creating the cultural tension the project needs.
That's a hard creative call. Removing work because it doesn't match the vision. But it's the same muscle I used in advertising for 15 years: knowing what to leave out is as important as knowing what to put in.
The site design follows the same principle. Warm, minimal, period-appropriate colour palette. Beige backgrounds, warm browns, dark charcoal. The restraint lets the images do the talking.
From gallery to business
Modern Retro has a print shop. Museum-quality giclée prints - A3 from £29, A2 from £49, framed options available. Stripe checkout, Printful fulfilment, push notifications on every order. The whole pipeline works.
But here's the thing: the gallery came first. Modern Retro is primarily a taste statement, secondarily a revenue stream. That sequencing matters. If I'd built it commerce-first, it'd be a print shop with nice images. Instead, it's a creative project that happens to sell prints. The difference is everything.
What I learned
Modern Retro taught me that constraint is freedom. The tighter the creative brief, the more interesting the output. It also taught me that AI isn't replacing creativity - it's amplifying taste. The tool generates images, but the taste decisions - which brands, which aesthetic, what to include and what to cut - are entirely human.
It's the project I'm proudest of. Not because it's the most technically complex (it isn't), but because it's the purest expression of an idea. One question, one constraint, one aesthetic. Everything else follows from that.
