Click any cell to open that section's side-by-side page. Every version is rendered at the real width of the Wefunder body column, 720px, so what you see is the on-screen size.
Rather not click section by section → all versions flat on one page ↗
The four numbers beside each version — on-screen height, horizontal overflow, image sharpness, contrast of text over photography — are all measured live by script, not self-reported by whoever built the version. Contrast is computed by repainting the text layer transparent, re-shooting the plate underneath, and running WCAG on the actual pixels behind every piece of text. It is not eyeballed.
68 versions in total; 55 of them are green on all six checks.
- Trevor Blackwell's title. The source text said
co-founder of both Y Combinator and OpenAI— that is wrong; he is not an OpenAI co-founder. The live raise page has long carried the correct wording,Y Combinator cofounder · OpenAI founding member, and every section copies that line exactly. The difference is one word, and that one word is the difference between accurate and caught out on the spot. - The revenue projection in Section 06. The source line
scale from millions to tens of millions of dollars in revenueis a forward-looking revenue projection, which is a liability under Reg CF, so the whole sentence is cut. The source'sfirst $1Mis also out of date — actual cumulative revenue is $1,735,844, the figure you already approved as Q14 on the decision board and the one the live hero line uses. - The "high rating" claim in Section 02 — this one reversed mid-way, and the story
is worth keeping. I first ruled that "the fact base holds no Google or Yelp rating,
so this claim cannot go on a banner," and Section 02 was built through three versions on
that basis. The person building that section went back to the source of record and
showed me I had it wrong:
organic.ratingis inverifiedstate, with an archived screenshot captured 2026-07-05, and4.8★ / 623 reviewshas been running on the live raise page in Section 05 all along. Those three versions were reworked and the rating went back in. What was retired is only the older unsourced figure4.9★ / 146— not the idea of showing a rating.
Two more figures went into that section alongside it: $4,700 of marketing spend in 2025 and $0 of product marketing — "organic" stands up because of that auditable spend, not because of the adjective.
Section 01's build notes claim "#c89324 is 4.6:1 on warm paper, passes AA."
That number is wrong. Measured, it is 2.58:1. And the whole set of Daylight
secondary colours misses too:
| Token | Measured on warm paper #faf8f3 | AA requires |
|---|---|---|
--ignite #c89324 (kickers, emphasis words) |
2.58:1 | 4.5:1 |
--faint, 46% black (secondary notes) |
3.26:1 | 4.5:1 |
--steel-3 #c8c2b3 (numerals, separators) |
1.67:1 | 4.5:1 |
Across every version in every section, 351 pieces of text land below the AA line on warm paper — all of them caused by these three tokens, not by any one version being built badly.
I did not touch it this round. This sits at the brand-token layer: it affects the entire raise page and every Daylight panel, not just these banners. Letting each person mix their own "more compliant amber" would scatter the brand on the spot. So the re-measuring script only flags this class in yellow and does not block; text set over photography is still a hard block (darkening that scrim clears it, so there is no excuse).
If it is to be fixed, amber on warm paper has to come down to around #8a6414 to
reach 5.06:1. Current on the left, darkened on the right — same sentence, same type size:
Darkened, amber is still amber — it just stops being the kind you can see but cannot read. This is your call, not something an automated step can make for you — changing it affects the whole raise page and every Daylight panel, not only these banners.