mirror of
https://github.com/hanzo-apps/hanzo.ai.git
synced 2026-10-29 07:20:43 +00:00
The site paints in white, #a1a1aa and black. The deck was painting in warm orange for our numbers and cool blue for the field's, which is a palette this company does not have. The distinction the numbers need survives without it: what we measured is white, solid and half a weight heavier; what the field published is grey, hollow and light. Bars, squares, tiles, pills, ledger and tables all follow that one rule, so a reader learns it once on slide five and carries it to the appendix. Colour now appears exactly once, on marks that belong to other people — the labs we route to, the places we deploy — each in its own brand's hex, out of one dictionary that holds artwork and colour together. Where a lab's own mark is black, it is white here, which is what those labs do on a dark ground themselves. The faces come from @hanzo/font, the package where they are authored; the deck had been carrying @hanzo/design's copy of that package's output, one build behind. Declared as that package declares them, with the site's fallback tails and Hanzo's voice — ss01, the single-storey a. Three slides also stopped being sparse and started being evidence. The models slide is a wall of twelve labs ranked by how many of each we carry, counted off the pricing catalog at sync rather than remembered — which caught the deck saying 541 models where the site says 536. The score slide grew the price column its own headline promises: Enso Ultra at 98.0 for $25 against GPT-5.2 Pro at 93.2 for $168. The credibility slide says NVIDIA Inception and Techstars with their marks, and names the five clouds we deploy to with theirs.
70 KiB
70 KiB