Selvedge Thinking: What Japanese Denim Got Right About Quality

Last September, I was in Kurashiki, Okayama. I went to buy denim. That was the whole plan. Walk into a shop, try on a few pairs, bring something home.

What I did not plan for was how much time I would spend just looking at the jeans. Turning them inside out. Running my thumb along the seams. Checking the selvedge line inside the cuff. Asking the shop staff about the mills and the looms and the dyeing process. I spent more time learning about how the jeans were made than I did picking a pair.

I went to Okayama because I like denim. I left thinking about product quality, craft, and what we are actually doing when we ship AI-generated everything at scale.

The Slowest Jeans in the World

Okayama Prefecture is the birthplace of Japanese selvedge denim. The story starts in Kojima, a district in Kurashiki, where cotton cultivation and textile manufacturing go back to the Edo period. In 1965, the first Japanese-made jeans were produced here using imported American materials. By 1972, local manufacturers had figured out how to dye their own indigo yarn, and a fully domestic product was born.

What makes Okayama denim different is not one thing. It is the accumulation of many small, deliberate decisions.

The mills still use shuttle looms from the mid-20th century. These machines weave fabric slowly, roughly 30 centimetres per hour. Modern projectile looms can do the same in minutes. But shuttle looms produce a tighter, denser weave with a unique irregular texture. The imperfections are the point.

Then there is the dyeing. Okayama is known for a technique called "shinpaku," where only the surface of each yarn strand absorbs the indigo. The core stays white. This means the denim fades beautifully over years of wear, revealing contrast patterns unique to the person who wore them. Your jeans become yours in a way that mass-produced denim never can.

Every step in the process is intentional. The cotton is sourced from specific regions. The yarn tension is calibrated by hand. The chain stitching on the hems uses single-needle machines that take three times longer than modern alternatives. A single artisan can craft one pair of made-to-order jeans from pattern cutting to final stitch.

None of this is efficient. All of it is deliberate.

And it is under threat. A Business Insider Documentary from late 2023 captured the paradox perfectly: Japanese denim is more popular than ever, with tourists and luxury fashion houses scrambling to get their hands on it. LVMH-backed private equity recently acquired a stake in Kojima-based brand Kapital. Sales are soaring. But the master weavers who actually make the fabric are aging out, and almost nobody is stepping up to replace them. Between 10 and 20 factories have shut down in recent years because there was simply no successor to take over. The vintage shuttle looms from the 1960s can only be operated by a handful of artisans alive today.

The world figured out it wanted the product. It just did not invest in the people and process that make the product possible. Sound familiar?

What We Lost When Speed Became the Product

I work in Generative AI. Specifically, I lead a team building AI experiences at SAP. I spend my days thinking about how to help people generate things faster. Screens, content, workflows, code. The whole pitch is speed and scale.

And I believe in that pitch. Generative AI is genuinely transformative. The ability to go from idea to prototype in minutes instead of weeks changes what is possible for builders everywhere. I wrote a whole article about this Everyone Can Generate a Screen Now. So What? after a product design event in Singapore. The barriers to creation have collapsed. That is a good thing.

But standing in that Okayama shop, holding a pair of jeans where I could feel the texture difference in the weave, where the shop staff could tell me which mill made the fabric and what loom it came off, I felt the tension between what we are building and what we might be losing.

When everyone can generate a screen, a document, a marketing asset, a codebase in seconds, the output itself stops being the differentiator. The craft does. The decisions behind the output do. The understanding of why this word and not that one, why this layout and not that one, why this feature exists at all.

Speed is the easy part now. Taste is the hard part.

Three Things Denim Taught Me About AI-Era Product Quality

1. Constraints are a feature, not a bug.

Okayama mills choose shuttle looms knowing they are slower. The constraint forces a different kind of attention. The fabric is better because the process is harder.

In product work, the temptation with AI is to remove all friction. Generate more options. Produce more variations. Ship faster. But some friction is valuable. The friction of writing a product spec by hand forces you to clarify your thinking. The friction of designing a screen pixel by pixel forces you to understand hierarchy. When you skip the friction, you often skip the understanding too.

It does not help when organisations are spinning up teams specifically to run speedboats on every new AI tool that drops. I get it. AI is an enabler. It empowers innovation. But not everything has to be run through it. Not every workflow needs to be accelerated. Sometimes the slower path is the one that produces something worth shipping.

Use AI to accelerate, not to bypass. There is a difference.

2. The person behind the process matters more than the process itself.

The artisans in Kojima who have been making jeans for decades are not following instructions. They are reading the material. Adjusting in real time. Making judgment calls that no manual could capture. You can see it in the product even if you never see the process.

AI is a tool. A powerful one. But the person directing it still needs to know what good looks like. A designer who has spent years understanding typography will use AI-generated layouts differently than someone who has not. A product manager who has shipped and failed and shipped again will prompt differently than someone reading a playbook for the first time.

Generative AI does not replace the need for expertise. It raises the bar on what expertise means. Knowing the tool is not enough. You need to know the craft the tool is serving.

3. The best products age well.

Shinpaku dyeing is slow and expensive because it optimises for how the denim will look in five years, not how it looks on the shelf today. The craftsmen are building for a future version of the product that the customer will shape through use.

Most AI-generated outputs optimise for the moment of delivery. They look great in the demo. They pass the first review. But they do not hold up under real use because nobody thought about edge cases, or maintenance, or how the user's needs would evolve.

Quality is not just about the first impression. It is about the hundredth use. Build for that.

The Self-Edge

Here is the thing about selvedge denim. The word itself is a corruption of "self-edge." It has been in use since the 16th century. On a shuttle loom, a single continuous weft thread passes back and forth across the warp. When it reaches the edge of the fabric, it does not get cut. It loops back. It turns around and keeps going. That looping creates a tightly woven band along both sides of the fabric that will never fray or unravel. The fabric finishes itself.

On a modern projectile loom, the weft thread is cut at every pass. The edges come out frayed and raw. A machine has to stitch them shut afterward to keep the whole thing from falling apart. It is faster. It is cheaper. But the fabric cannot hold itself together on its own.

That difference is invisible to most people. The self-edge sits inside the outseam of the jeans. You will never see it unless you cuff the hem or flip the fabric inside out. In Japan, they call it "akamini," which translates to "red ear," because of the coloured thread that often runs through that finished edge. Different mills use different colours. Levi's was red. Lee was blue or green. Wrangler was yellow. Each colour was a quiet signature from the people who made the fabric. A mark of origin hidden in a place only the curious would look.

The self-edge does not make the jeans look better on a shelf. It does not show up in a photo. It is not a feature you would put in a pitch deck. But it is the reason the jeans hold together after years of wear. It is structural integrity disguised as a small detail.

I keep coming back to this idea.

In the age of generative AI, most of what we build is projectile loom output. Fast, wide, good enough. The edges get stitched shut after the fact. We ship the thing, then patch the seams when they start to fray. We call it iteration. Sometimes it is. Sometimes it is just a lack of intention dressed up in agile vocabulary.

What I want to build are self-edged products. Things that hold themselves together because the process that created them was continuous and considered. Where the quality is not bolted on at the end but woven in from the start. Where someone cared about the parts you cannot see.

The question used to be "can you make it?" Now anyone can make it. The question now is whether the thing you made can hold its own edge.

I think about those jeans from Okayama a lot. The looms behind them are loud. The work is slow. The world has moved on to faster, cheaper methods. But people still fly from all over the world to buy from those shops. They cuff the hem. They look for the coloured thread. They check the self-edge.

That tells you everything you need to know about quality.