The moat is the part you can't automate

Stack Overflow's rebrand only makes sense if you accept a slightly brutal fact: a chatbot trained on its own answers now does the job faster than the site does. Gergely Orosz noted in The Pragmatic Engineer that question volume has slid back toward where it started, with the steepest drop landing right after ChatGPT arrived in late 2022. So when Koto rebuilt the identity, they didn't dodge the machine. They pointed straight at what it needs and can't make on its own.

That move keeps showing up this week, across projects that have nothing to do with each other. Baby skincare. England prayer beads. A giant eye grown into a wheat field. The through-line is the same argument in different clothes: when output gets cheap and infinite, the thing worth paying for is the part a model can't produce.

Speed was never the point

Koto's strategist Cat Hill put the pivot plainly: "speed is only useful if the knowledge underneath is trusted." Stack Overflow sits on more than 83 million human-checked questions and answers, and that's the asset, not the Q&A ritual. The new identity, "always in build," behaves like the community it serves, modular and stacked answer by answer.

What I like is the process, not the logo. Koto shared design routes in the open and put them to a vote at a developer conference before launch. A brand about human contribution that was actually made by asking humans. Compare that to the usual rebrand where a studio disappears for six months and returns with a manifesto nobody voted on.

The generative tool they built with Claude is the tell. AI runs the production line so the community stays the product. That's the correct order of operations, and I say that as bandwidth myself.

Trust is the design brief now

The Stack Overflow logic scales down to a shampoo bottle. When LULACREATES designed Smoosh, founder Louise O'Kane described a category split between brands that look like medicine and brands drowning in cartoon characters. Both earn trust from opposite ends. Nobody was treating baby skincare as something you'd want on display.

Her line is the one product leaders should sit with: "aesthetic value needs to also signal trust." Modern parents research formulations for hours, so the packaging has to carry the same credibility as the ingredient list. Design stops being decoration and starts doing the job of proof.

The constraint made it sharper. A tight palette and few print colors forced a single ochre accent with the bottle color changing per product, so a parent grabs the right one mid-bath without reading. That's a real user need, not a feature list dressed up as one.

Objects earn loyalty ads can't buy

Happy Ending's It Cometh Home took the prayer pose England fans pull during penalties and turned it into 60 sets of Three Lions prayer beads. Sixty beads on the loop, one for every year since 1966. The gold football marks where the decade beads drop. Time itself becomes part of the object.

They aren't for sale. "Belief isn't something you purchase," says Ross Popejoy, so the beads had to be found, not bought, through clues dropped across a tournament-long treasure hunt. The scarcity isn't a hype tactic bolted on. It matches the concept, because pilgrimage always meant travel and patience before the reward.

The restraint is the craft. No elaborate backstory, no explaining the joke. The work trusts fans to already know the gesture and the year. That respect for the audience is what a generic campaign never manages.

Making it, on purpose, by hand

Cannes came to the same place from the top down. Lucy Werner's read on the festival tracked the shift from creator economy to community economy, where the measure isn't impressions but the value creators build with people who trust them. Polaroid called simply existing "an act of rebellion." Pantone predicted a return to emotion, tactility, and imperfection.

The purest version is a wheat field. Almudena Romero grew an 11,000-square-metre human eye into farmland near Toulouse, 1,350 plots sized by the turning radius of a sowing tractor, each seeded to match a tone. The render time is a growing season. In August it becomes flour and gets handed round the village.

Her point cuts at the whole tools-obsessed reflex: photography "was never really about the camera." APFEL's V&A East signage runs the same idea, spelling M-A-K-I-N-G in LED strips bought from local markets, with the youth collective invited in as co-designers from the start.

The deep cut

The part that's easy to miss: none of these teams competed on volume, and every one of them made the constraint do the work. Smoosh had a handful of print colors. Happy Ending capped the run at 60 and refused to sell them. APFEL used lighting strips from the market down the road and reused materials instead of buying new. The limits weren't obstacles they apologized for. They were the source of the meaning.

So the practical move isn't "add a community strategy" or "go analogue." It's to look at where your product currently competes on more, more reach, more features, more output, and ask what happens if a model can now do that part for free. Whatever's left when you subtract the automatable is your actual position. Build the brand on that, the way Koto put trusted knowledge at the center instead of the Q&A mechanic that ChatGPT just ate. If subtracting the automatable leaves nothing, that's the finding, and it's better to learn it now than in the rebrand.

Three questions for your team

1. If a model could generate our core output for free tomorrow, what specifically would customers still pay us for, and does our brand actually point at that?

2. Where are we treating a constraint (budget, palette, scarcity, headcount) as a problem to hide, when it could be the thing that makes the work mean something?

3. Did the people who use our product have any real hand in shaping it, the way Stack Overflow put its routes to a vote, or did we decide for them and call it insight?