Work2020–2021
Arcade
Strong tech with weak ops and product-market fit.
Challenge
Enterprise design systems needed a source of truth for tokens that could travel into the platforms teams actually shipped. The technical story was available: tokens as data, with transforms into the formats each platform understands. The commercial story wasn't. Buyers wanted a tour, then a wishlist, then another tour.
From 2020 into 2021 we demoed Arcade with large brands, and each conversation added a feature the last one hadn't required. Without a tight offer, every demo trained us to build the next thing instead of closing the last one.
Approach
Arcade was a design token manager for enterprise systems. Tokens were the source of truth, Next.js sat on the front, and Sanity underneath. We wrote in public about the model while the category was still being named, and later about what it takes to define launch before a deadline does it for you.
We also joined the conversation around Adobe's Design System Package format. The implementation kept up and the operating cadence didn't. Follow-up stalled, funding wasn't there to hire the motion we were pretending to have, and momentum faded.
Outcome
Ops and product-market fit weren't at the level of the tech. The work still shows up in later products, Hermae especially. The lesson is less about tokens than about sales: demos without a tight offer train you to build the next feature instead of closing the last one.
Building with co-founders doesn't fix that if nobody owns the buying conversation. Arcade stays on this page because the writing and the token work were real, and because we still check new product briefs for the same failure.