About X-Brain
We're building the prediction layer for personal taste.
Why
Choice is exhausting. Walk into a dispensary and there are 200 strains. A wine shop and there are 1,000 bottles. Reviews are written for the average person — but you're not the average person. Your friend recommends "Wedding Cake" and it's a let-down because they like sleepy and you like creative. The technology to fix this has been around for a decade. Nobody has bothered to point it at the categories that matter most to enthusiasts.
How
You sign up, answer 15 quick pairwise comparisons (this vs that), and the model builds your starter taste fingerprint. Cohort assignment slots you near other users with similar fingerprints. Content-based scoring fills gaps for products you haven't rated. Each new rating updates everything. Day 1 predictions are usable; day 30 predictions are uncannily accurate.
Verticals
BudBrain (cannabis) is live. WineBrain is launching alongside. The platform is tenant-aware — adding FoodBrain or CoffeeBrain or TeaBrain is a config change, not a rewrite. The cross-brand parts (friends, badges, shared ratings, knowledge graph) span every vertical you join.
Stack
Go API gateway, Python ML service, PostgreSQL with row-level security, Neo4j for the cross-brand knowledge graph. SwiftUI for iOS, Jetpack Compose for Android, Next.js for the web. We use Anthropic's Claude for vision identification — the waterfall short-circuits to local models when COA QR codes or terpene stickers are detected, so cost stays predictable at scale.
Operated by
BudBrain LLC, based in Pennsylvania. Reach us at hello@xbrain.dev or read the blog for technical posts on how the system works.