NewsKit shipped a clean ramp — blue010 to blue100, light to dark. Then each brand re-themed it, and the discipline broke: hex values were dropped in with no regard for the weight number, so it stopped predicting anything. Worse, the color was named for what it was — “blue” — not what it did. A token called blue can’t hold Barron’s navy or MarketWatch’s green. It needed to be primary.

They were hand-typing hex because nothing generated a proper ramp. So I built one: Genome — feed it a brand color and it produces a perceptually-even scale, indexed to CIE L* and named by role, so one token carries any brand’s hue. I studied every major system to build one that could express them all.

Every hand-picked hex is a one-off — and a one-off can only be checked one at a time, by hand, forever. You can’t automate that. Before any pipeline could touch the palette, it had to become systematic: weights chosen by rule, not by taste. And because it’s a white-label system, systematic meant one thing — every brand agreeing to the same meaningful weights.

The brands didn’t each hold a veto. It came down to four people — the VPs and directors who owned them between them. Win those four, and every brand below moves. So the whole thing rested on making the case to four.

I didn’t have to campaign. I put the L* drift on screen — each decision-maker watching their own palette’s numbers stop meaning anything. They already knew a white-label system needs semantically meaningful weights, or theme-swapping breaks at scale. Once the problem was undeniable, agreement was fast: all four signed within three months.

Before Style Dictionary. Before a single token was compiled. Before any of the automation that let the system scale to twenty brands and five platforms — there was this: agreement. You cannot compile a standard nobody signed. Every architect wants to jump to the tooling; the tooling is worthless without the consensus beneath it.

With consensus in hand, the brands moved onto a shared, accessible, perceptually-honest color foundation. Only then could the automation follow — the tokens, the pipeline, the five platforms. The agreement is what made the scale possible. The code was the reward for the diplomacy, not a substitute for it.

Genome shipped, and it worked — the brands themed reliably and the accessibility decisions scaled. And then I realized I’d made one wrong call. I anchored “normal” at weight 400, because CSS does — where font-weight: 400 means “normal.” I’d borrowed the convention without questioning it. But the history of font-weight naming is a mess, and 400 isn’t the perceptual center of anything. L*50 is.

500 — the true center of a 000–999 scale, and exactly L*50: the one value that clears 4.5:1 on both white and black. I named the refined standard Univers — after Frutiger’s typeface, where systematic weight-numbering began.