Design engine bake-off / DawnBands lander

Five design engines, one identical brief. Round two.

Round one was rejected as plain. The fault was the brief: four sections and one cutout is a SaaS skeleton, and every engine converged on a template. Round two holds a twelve-section DTC lander constant — lit-LED hero, 41-buyer survey bars, mechanism, school-morning timeline, comparison table, buyer verbatims, stakes, four-colour picker, guarantee and FAQ, sticky mobile buy bar — with identical copy, images and facts. The only variable is still the design engine.

Each frame below is the real page rendered at the selected viewport and scaled down. Open any one full-size to judge it; thumbnails flatten typography. Every number on the pages is from the July–August post-purchase survey (41 responses). No reviews or star ratings appear anywhere because the store’s review widget returns none.

1 · minimalist-skilltaste-skill Open
warm paperNewsreader serifflat bento, 1px rulesspec 100% · 0 errors
2 · soft-skilltaste-skill Open
dawn mesh on blackdouble-bezel glassmask-slide barsspec 100% · 0 errors
3 · taste-skilltaste-skill Open
cool bone + cobaltreal dark modezero iconsspec 100% · 0 errors
4 · frontend-designanthropic · control Open
night → daylight rhythmfull-bleed timelineamber skew bandspec 100% · 0 errors
5 · awesome-design-htmllinear reference Open
Stripe mesh heroVercel surfacesApple 17px bodyspec 100% · 0 errors

What each engine did with the same twelve sections

EngineVisual system it imposedWhere it pushed the page
minimalist-skillWarm bone canvas, every divider 1px #EAEAEA, flat 12px bento, Newsreader serif against Geist, pastel pill tags as the only colour, motion invisible.Reads as a well-set document. Densest information per scroll of the five; lowest sales pressure. Calm, not plain, now that there is something to set.
soft-skillBlack ground with three radial mesh orbs sampled from the hero’s own sunrise gradient, film grain, nested double-bezel on every card and table, island CTAs with button-in-button arrows, magnetic hover.Most expensive-looking. Longest page (8046px). The survey bars and comparison table become jewellery. Furthest from minimal.
taste-skillCool bone + one cobalt accent, Geist/Geist Mono, zero icons, real mirrored dark-mode tokens, seven distinct layout families so no two sections share a grid, rank bars with no tracks.Most disciplined. Its doctrine bans em dashes, caps eyebrows at four and quotes at three lines; it kept the copy verbatim anyway and paid in font scale. Round one cut copy to fit; this round it did not.
frontend-designMidnight indigo and horizon amber, Bricolage Grotesque 800 display, section backgrounds that travel night → paper → sunrise → paper so the page itself moves from 10:30 pm to daylight. Timeline as four full-bleed colour panels.Biggest swing again, and this time the content carries it. The school-morning section is the single strongest moment across all five pages.
awesome-design-htmlStripe’s mesh-gradient hero and hairlines, Vercel’s inset-shadow surfaces and mono micro-labels, Apple’s 17px body and image-first tiles. Linear kept only for rhythm.Fixed round one’s B2B density. Now reads as a consumer product page. Swatches carry their own colour thumbnails. Safest of the five to put paid traffic on tomorrow.

What changed, and what to judge

Plain was the brief, not the engines. Given real content, all five separate cleanly: 1 and 3 are the minimal pair (paper vs. cool bone), 2 is polish, 4 is concept, 5 is the reference-faithful consumer page. Pick on the full-size pages, not the frames.

4 and 5 are the two to shortlist. v4 stops a scroll and its timeline panels are the best use of the constant content. v5 is the one a mom will not find odd and the easiest to hand to a theme. v1 and v3 are the choice if the brand direction is genuinely minimal; v2 is the choice if it is premium.

Hero photo note. The store’s lit-display product image had six caption claims baked into the corners. It was cropped to the band and gradient before use, so none of the pages carry claims the copy does not.

Not yet cleared for a live store page. Sections 2, 4, 5, 6, 7 and 9 carry new prose written for this round. Facts are real and the register was checked by hand, but none of it has been through the copywriter or QA gate. The survey quotes are reasons for ordering, labelled as such on-page, never presented as results or reviews.