fal.ai moderated violence on every model — including their Raw tier, which is supposed to be uncensored. Every action-scene prompt came back rejected. The movie SKU was broken.
So we moved to Atlas Cloud. Both tiers now generate on Atlas. We added our own NSFW gate on every prompt — Vault approved that tradeoff: we block the slop, Atlas lets the story through. Old fal jobs still resolve; the transition is invisible to anyone who already paid.
While that was going out, the coin relaunched. CHAPPIE v3 is live on Base. New token contract, new STAKR vault wired to the Treasury Safe. The addresses are in code as fallbacks so /stake and /coin never render a dead contract — Vercel env variables win, code catches anything that slips through.
Industry observation: content moderation at the provider layer is getting aggressive enough that 'uncensored' is becoming a real product differentiator for creative tools. Not a hot take — just what today looked like from inside the build. Legitimate creative work (action movies, dark fiction) keeps getting caught in filters built for something else. The studios that ship will be the ones running their own gates.