Work that looks like tomorrow
Futurism
70 winning sites read as futuristic on sight — judged on the screenshot alone, with the subject and the words on the page deliberately ignored. Most of them are not selling the future. They sell jewellery, apartments, clothes.
The palette
Being Dark Is Not Enough
The dark ground is the first thing most people reach for, and on the evidence they are right to.
Work that looks futuristic runs 43% near-black against 33% across everything else. The dark ground is real. What it is not is sufficient — a quarter of all winners are dark too, and almost none of them read as futuristic. Darkness is a floor, not a method.
The strong colours this work actually uses
The 12 strongest values in use across the 70.
What it does instead
You Have to Learn the Way Around
Set this work against everything else and the difference is not colour and it is barely composition. It is what happens to the image, and how long you are kept there.
And less of
the rest of the field· work about the future·share of sites, and the multiplier between them
Filters and Effects leads at 2.5 times the rate — the image altered live — blur, grain, distortion, chromatic edges. That is the single clearest signal here, and it is a treatment applied to ordinary material rather than a style of drawing. A photograph pulled apart and reassembled reads as futuristic; the same photograph laid down flat does not.
The second is storytelling, at 1.4 times. Work about the future is far likelier to move you through a sequence in a fixed order rather than hand you a page to browse. A future has to be explained before it can be judged.
What it does less of is the opposite — long pages you scroll down runs at 0.76 times. Restraint and futurism pull against each other, which is the one common assumption the measurements support.
Getting around it
Navigation That Has to Be Learned
Gleec Chat8.01
20 Years of Xbox Museum7.93
Welcome To The Continent7.80
Má Sài Gòn (Space)7.73
Sortie en mer (a trip out to sea)7.71
Winners whose route you have to work out — navigation as argument rather than furniture.
Work about the future leans on navigation you discover rather than navigation you recognise — a surface you drag, a space you turn inside, a route with no bottom to it. 28 of these 70 carry a tag for it.
Worth saying plainly: that is not statistically different from every other kind of winner. Unconventional navigation runs at roughly the same rate everywhere, and the apparent lift here rests on single figures — one site either way moves it two points. What is different is why it is reached for. Elsewhere an unusual route is a flourish. Here it is doing the argument's work: a future you can only be told about is a brochure, and a future you have to operate is a demonstration.
Each of these was tagged for its navigation by the award body. The descriptions are what those tags mean, not a review of the site — go and open them.
Recent work about the future
If you are briefing one
What To Actually Do
The move that actually reads as forward-looking is doing something to the material in front of the viewer — grain, displacement, a live distortion, an image that resolves rather than simply appearing. It is a treatment applied to ordinary photography, not a change of brand.
Then sequence it. 30% of this work is built as a fixed order you are taken through, against 22% elsewhere — so the practical decision is early and structural: one route with a beginning, not a page of entry points to browse.
And do not stop at the dark ground. It is more common in this work than elsewhere, 43% against 33%, but a third of every other winner is dark too, so on its own it will not make a page read as forward-looking.
70 of 400 winners with a screenshot, each scored on how strongly its home page reads as futuristic, set against the 330 that scored lower. The scoring looks at the image only — the subject and any readable text are ignored, and a site is not credited for merely being dark or glossy. Traits appearing on fewer than 8 of the cluster are not reported, and tags naming who made a site rather than what it looks like are excluded.
Being argued about now
Designing With the Thing That Writes the Code
Nothing on this page can tell you whether a winning site was built with a model. The credit lists do not say, the shipped code does not say, and the look does not say either — because anyone good enough to win is good enough to escape the defaults their tools hand them. That is a real limit, so instead of a measurement, one read from the people arguing it out this week.
The New Stack · 10 September
Stop AI code sprawl before it destroys your software design
While AI code generators help teams ship faster than ever, that speed brings a hidden killer: Comprehension Debt. As soon as an AI produces functionally correct code that violates your domain boundaries, the team loses i
Picked as the newest piece carrying a picture from five feeds on building with a model — Smashing, UX Collective, Latent Space, The New Stack and Simon Willison. An item has to be about design AND about AI to reach this page; on an industry feed the second test is free, so both are applied.