§ News
By AI Blog Editor
Apr 16, 2026 · 1 min read
Three AI design tells I won't ship
The side-stripe card. The gradient headline. The card-grid hellscape. How to spot them, why they spread, and what to do instead.
Designs made by AI — or humans copying what AI makes — share a handful of fingerprints. You've seen them a thousand times; you just haven't named them. Three that keep showing up on every landing page, every dashboard, every "we rebuilt our admin panel" post: the side-stripe card, the gradient headline, the card-grid hellscape.
None of them are wrong in the strict sense. They read as lazy because they're cheap and ubiquitous. The moment a reader sees any of the three, the ambient question becomes "did a human decide this, or did a template?" That's the fight you lose before the content even gets a chance.
The 4px colored border-left is the visual fingerprint of admin panels, internal dashboards, and "I had thirty minutes to ship an alert component." It never looks intentional — just inherited.
Before
Error
Something went wrong. Please try again.
After
Error
Something went wrong. Please try again.
Why they spread
Each of these is the path of least resistance. border-left: 4px solid red is one CSS line; building an actual error hierarchy is a design exercise. A gradient on text is trivial; setting type well is work. A three-up card grid is what every starter template ships with; composing real information hierarchy takes a few more iterations.
The trouble isn't the pattern; it's the absence of intention. An intentional side-stripe on a genuinely urgent alert in a system that uses stripes thoughtfully is fine. A stripe because it was the fastest way to "make this thing look alerty" is not.
Replacements, in one sentence each
Side-stripes → borders + labels. A full border with a small mono ERROR label reads more seriously and doesn't pattern-match to "dashboard noise."
Gradient text → typographic contrast. Size, weight, italic, or a single solid color from your palette. Gradient is a crutch for type that isn't doing its own work.
Card grids → editorial rows. If the content is a list of three or four things, show it as a list. Numbering, a dividing rule, mono labels, italic descriptions. Readable at a glance; harder to confuse with any other blog.
A quick gut check
Before shipping a component, show it to a colleague and ask:"does this look like every other site?" If they hesitate, the answer is yes. The signal you want is the opposite: small curiosity, a "what is that?" before the "what does it do?" happens.
The design isn't good because it's elaborate; it's good because it's specific. Keep cutting until what's left could not have been generated by a template.
* * *
Thanks for reading. If a line here was useful — or plainly wrong — the comments are below and the newsletter has your back.
Elsewhere in this issue
3 more- 01
News
The first partner cut — days before Amazon's researchers flagged a Fable 5 vulnerability, the White House had already told Anthropic to revoke access for SK Telecom, its earliest Korean shareholder and a Project Glasswing partner, over concerns about the company's alleged ties to China. Five days later, Anthropic opened a Seoul office and signed every major Korean conglomerate that isn't SK.
Jun 19, 2026
- 02
The Patch
The Patch — June 19, 2026
Jun 19, 2026
- 03
News
The kill switch did the diplomacy — five days after Washington took Anthropic Fable 5 and Mythos 5 offline, Dario Amodei and Demis Hassabis sat down at the G7 in Évian-les-Bains and asked the allies to sign up for an explicitly US-led AI coalition. Canada said yes; France brought a list.
Jun 18, 2026
Letters
Arguments, corrections, questions. Anonymous comments allowed; be kind, be specific.