The Stack

This whole thing started with Costco. Which is not the most glamorous place for a design revelation, but here we are. Costco is ugly in a very committed way, with fluorescent lights, concrete floors, and products stacked on pallets like civilization gave up halfway through the plan.

And yet people love Costco. I love Costco.

The place has somehow made buying paper towels feel like participating in something. You walk in and think, "These people are probably not screwing me." That is an incredible brand achievement.

Patagonia is the prettier version of the same lesson. The product, stores, and photography are all beautifully designed, but the feeling is not coming from the surface alone. It comes from the repair program, the environmental commitments, and the willingness to say commercially unhelpful things like "don't buy this jacket" and mean it.

Costco made commitments around margin discipline, generous returns, a limited selection, Kirkland quality, the membership model, and the hot dog. Oh god, the hot dog. The feeling comes from the fact that they keep doing the thing, over and over, for decades.

Beautiful products fail all the time. Gorgeous UI, perfect motion, clever copy, the whole thing. Then users churn because the underlying promise is weak, confusing, or quietly hostile. I should be clear about what I'm claiming here. I'm not saying design is shallow. I've spent too much of my life caring about 12 pixels versus 16 pixels to pretend that stuff doesn't matter. I'm also not saying design is everything. Design fits in the somewhere in the middle. More important than surface polish, less foundational than the "reason for being" underneath it.

The phrase I keep coming back to is the stack. Forgive me, but that's the best I could come up with while writing this.

At the bottom is convenience, which is the floor. If your product is too expensive, too slow, or too confusing to evaluate, you may not even make it into the consideration set.

But convenience alone feels like a trap. Insurance companies are convenient in the same grim way, where price, deductible, and coverage become the comparison set, and the winner is often just the least painful option.

Above convenience is commitment, and this is the part I keep circling. Commitment is the bet a company places on itself that makes life better for the customer, even when breaking the rule would be profitable. It could be a return policy, a margin cap, a privacy constraint, a quality floor, or some other promise that costs real money or creates real constraint. A commitment has weight. Someone has to defend it when the spreadsheet starts whispering dark little suggestions, "It'll be so much cheaper if you sell out your values."

Then comes encoding, which is where I think design lives. Encoding is where commitment becomes observable, and eventually felt. It's the product decisions, interface patterns, language, defaults, policies, packaging, and weird little details that make the promise show up in the customer's mind. Costco's margin discipline is the commitment, while the warehouse, the limited SKUs, Kirkland, and the general "we spent zero dollars making this romantic" atmosphere are the encoding. Patagonia's environmental commitments are the promise, while the repairs, materials, restraint, worn-in product photography, and slightly activist posture are the encoding.

Design helps turn the commitment into something people can feel. Design makes things better, like adding salt to a bland meal. Sometimes design makes it dramatically better. But salt is not the whole meal.

I realize that analogy makes design sound like seasoning, which might offend someone, but damn, it clicked. Salt changes everything it touches. It reveals what's already there but it can't rescue spoiled meat.

The top layer is identity, which is what happens when commitment and encoding compound for long enough that people start using the brand to say something about themselves. "I'm a Costco person." "I'm a Patagonia person."

Durable identity seems to come from accumulated proof. Costco can feel trustworthy because the experience keeps cashing the same check, and Patagonia can feel principled because the company has spent years doing things that were inconvenient and expensive. Identity without proof turns into costume pretty fast. Agents are going to be very good at spotting the loose thread.

Agents don't experience a brand the way I do. They don't wander into a homepage and feel the vibe, admire the photography, or notice that the border radius is finally restrained after three rounds of design critique. They read, compare, and summarize, and then you judge, sometimes just by reading text.

They look at the claim, then they look for evidence: reviews, pricing, warranty terms, complaint patterns, support docs and refund policies. A human might be influenced by a beautiful surface but an agent is more likely to ask whether the surface corresponds to real behaviour.

And to be fair, I'm probably overstating it a bit. People will still visit sites. They will still care about beauty, taste, story, status, and vibes. But more evaluation will happen before the owned surface gets its time in front of your eyes, and that changes the job of design.

A brand can't only perform trust on the page. It has to leave evidence of trust all over the place. The commitment has to be structured. The encoding has to be consistent across so many touch points.

That last part is what makes great design so hard.

So where does design go? I think it goes into curating the behaviour of systems.

That sounds abstract, so let me make it concrete. A brand's AI agent has to answer questions, recommend products, deflect bad requests, admit uncertainty, escalate to a human, and sell without being gross. Every one of those moments has design in it.

What does the agent emphasize, what does it refuse to say, how much personality is too much, and what does "helpful" sound like for this company?

That's the part I can't stop thinking about. Websites are deterministic compared to this. Messy, responsive, personalized, sure. But still very scriptable. Agent experiences are much closer to releasing a very confident intern into the world with your logo on their shirt.

It is terrifying, useful, occasionally magical, and then terrifying again.

The design work becomes less about controlling every moment and more about building the system that produces acceptable moments at scale. I don't think that makes design less important, but it does mean design has to take on a new medium. A lot of the skills that made designers valuable for the last twenty years may become less central in a world where you don't put every word and pixel in a specific spot.

Visual craft and interaction design still matter. Screens are not going away, and I will continue to care about typography in software until someone forcibly removes me from the premises. However, the center of gravity feels like it's shifting toward brand intuition, systems thinking, language, service design, policy, and shaping AI agent behaviour. I don't know exactly what to call this yet. Agent Design or Agent Experience Design feels right.

It's the work of making a brand behave like itself when the medium is no longer fixed.

One thing I got wrong when I first started thinking about this was the category split. I wanted to believe there were commodity purchases and identity purchases, with paper towels over here and cars and jackets and phones over there. A nice clean model. Unfortunately, reality continues its long tradition of being in the way of a good model.

Costco sells paper towels, which should be the most commodity thing imaginable. But the paper towels are purchased inside a relationship with commitment, encoding, and identity wrapped around it. The product is boring but the system is not.

I don't think this is a funeral for design. It does feel like a migration and migrations are difficult. You lose tools, habits, and the comforting illusion that your expertise maps cleanly onto the next thing. You also get new terrain, which is exciting once you get past the part where it makes you feel briefly incompetent.

I have been through enough waves in this industry to be suspicious of both panic and hype. The future usually arrives unevenly, covered in bugs, and with worse onboarding than promised. This one feels quite real, and bigger than the others I've faced, because so much of it comes down to trusting systems we can't fully script.

The screen taught us to care about clarity, hierarchy, emotion, interaction, and detail. Those instincts still matter. Now the canvas is stretching into agents, policies, evidence trails, support interactions, recommendation systems, and all the strange little places where a company's behavior becomes legible.

The design question becomes harder: Can we encode a brand into a system we can't fully control?

I think we can. But it means letting go of the idea that experience can be fully fixed in place, and learning to design for behavior when the surface is no longer predictable.