Two things happened in the last two weeks.

Walmart redesigned Great Value, their flagship private label, for the first time in over a decade. Ten thousand SKUs. The new look is wonkier — visible texture, off-center compositions, type that looks like a person held a pencil. The largest retailer on earth, with the budget to commission anything, deliberately picked the messier option.

Around the same time, a Hacker News thread climbed the front page asking the question every B2B SaaS founder is privately asking: Why does every B2B SaaS have to look like Linear or Stripe? Six hundred comments. Almost no disagreement.

Two stories, same insight.

Slick is the new generic. The polish that used to mean "we hired good designers" now means "we hired no one." The work is no longer to look more refined. It's to look more decided.

The polish premium has inverted

Three years ago, "looking professional" had a fixed shape. Clean sans-serif headline. Abstract gradient hero. Four-up feature grid. Stock illustration of three people pointing at a laptop. Pastel-on-white for SaaS, neon-on-black for AI. If you nailed the template, customers and investors read you as legitimate.

Then the template became free.

Midjourney v8 will give you the gradient. v0 will give you the landing page. A $20 plan and 40 minutes will produce an identity that two years ago took a studio four weeks. The aesthetic that used to cost $40K is now a button.

When the polished look becomes a default, it stops signaling competence. It starts signaling absence. Independent design buyers are now paying a 10–50x premium for visibly hand-crafted work over AI-generated equivalents. That's not nostalgia. That's the market repricing the only signal that's still scarce: evidence that a human decided.

The Linear-Stripe convergence problem

The Hacker News thread is interesting because it isn't about Linear or Stripe. They each made specific design decisions a few years ago — Linear's monospace + dark-mode aesthetic, Stripe's restrained gradient + serif system — and those decisions worked. So three hundred companies copied them.

Now the entire B2B SaaS category is wearing the same uniform. Aeonik headline. Indigo accent. Subtle gradient that fades into a section break. Animated rotating words in the hero. A 4-up feature grid with abstract icons. Open Vercel's templates page and you can see the same site rendered 80 times.

The lesson isn't "don't look like Linear." Linear made good decisions for Linear. The lesson is what got Linear there in the first place: they picked a specific aesthetic and defended it against everyone who told them it was too aggressive, too dark, too programmer-coded. That defended specificity is the asset. Copying the output without copying the decision-making is how categories collapse into sameness.

The same pattern played out with AI startups. Look at any cohort of AI Series A logos from the last 18 months: hexagons, neural mesh, gradient orbs, abstract circuits. The companies that read as distinct in that field — Anthropic with a serif wordmark on cream, Hugging Face with an actual emoji, Perplexity with a teal nobody else will touch — all made decisions that someone in the room argued against. Those arguments are the brand. The ones that didn't have those arguments now look like a stock photo of an AI startup.

What the founder mistake actually is

The mistake isn't "we used AI to make our brand." Tools are tools.

The mistake is using AI to skip the step where you decide what you stand for.

When the visual identity decision is outsourced to a model, the output is statistically average — by definition, that's what models return. The mean of the training set. Average works fine if your category is empty. In a category with 200 companies chasing the same TAM, average is invisible.

The work isn't to look more handmade. Hand-drawn type on its own is a costume. By the end of this year there will be ten SaaS companies with wonky-illustration brand systems and you'll find them just as hard to tell apart as the Linear clones. The aesthetic isn't the moat.

The work is to look like a real choice was made — and to make that choice visible. A custom mark instead of a Figma template. A photographic treatment nobody else in your category uses. A color you'll defend against your own board. An off-grid layout that reads as deliberate, not lazy. The market is paying a premium for proof a human steered, and that proof has to live in the work itself.

The so-what for founders

Open your homepage. Then open three competitors' homepages in adjacent tabs.

Mask the logos. Could you swap them across the four sites and have anyone notice? If the answer is yes — same hero pattern, same color drift, same headline structure, same illustration style — you don't have a brand. You have a render of the average company in your category.

Two moves you can run this week before you do anything bigger:

1. Take your homepage hero. Cover the logo and the company name. Send a screenshot to three customers and ask them to identify the company. If they can't, your visual identity is doing zero work for you.

2. Open your last five marketing assets — deck, ad, blog header, lifecycle email, social tile. For each one, write down the single decision that made it specifically yours. If you can't write one for an asset, that asset was generated, not designed.

The companies that win this cycle aren't the ones using AI hardest. They're the ones who use AI to ship faster on decisions a human already made. Slick is the cheapest input in the system. The expensive thing — and the thing customers will pay for — is the decision behind the slick.

If you want a second opinion, reply with your homepage URL. I'll tell you which elements prove a person decided them and which read as prompts.

— Daniel