Market & Context
The sports betting market was maturing fast. Legalization had unlocked a flood of new users, but the infrastructure remained fragmented—books were sophisticated, bettors were flying blind, and most "expert picks" were closer to entertainment than insight. In short: supply had scaled. But intelligence hadn’t.
Brand & Product Limitations
There was no existing brand, platform, or product. Just a hypothesis: that a small, disciplined set of daily picks—ranked by edge and backed by real models—could outperform the crowd. But that meant starting from zero: no users, no interface, no engine, no positioning.
Strategic & Technical Barriers
Most sports models are monolithic and brittle. The challenge wasn’t just building one good model—it was engineering a framework that could adapt, evolve, and interpret everything from injury reports to Twitter sentiment. Add to that: live data ingestion, real-time model syncing, explainability, and a seamless front-end that felt as intuitive as a calendar app.
Human & Organizational Challenges
We were the founders, engineers, designers, and strategists. That gave us full control—but it also meant every decision needed to balance technical precision with brand integrity. We weren’t just chasing a working prototype. We were building a product that casual users could trust from day one.
Discovery & Strategy
We began with one question: What makes a bettor trust a recommendation? From that, we reverse-engineered the experience—designing for clarity, surfacing explainability, and aligning the product's core UX around trust, performance, and daily habit. We mapped every friction point in the current landscape and focused our attention where the edge lived: EV, confidence, transparency.
Concept Development
Rather than build a one-size-fits-all model, we developed a modular agentic AI system. Each agent had a job—some focused on trends, others on injuries, others on anomaly detection. These agents ran independently, then aggregated their results into a single system-level decision. This ensemble approach let us swap in new tools (like GPT-4.1, Claude, Perplexity, Gemini) and tune our edge over time.
Design Execution
The interface is built for discipline: clean, confident, no distractions. Each pick is delivered with a score, a rationale, and the opportunity to go deeper—whether via AI Q&A or interactive data panels. It’s not a sportsbook UI. It’s more like a Bloomberg terminal for betting. Web-only, responsive, fast.
Build & Integration
We shipped on a modern, scalable stack: Next.js + Vercel, Supabase for auth and storage, Tailwind for speed and polish. Stripe handled payments. Agno powered agentic logic. We used structured content and dynamic APIs to populate picks daily while maintaining full traceability and performance tracking across each recommendation.
Collaboration & Workflow
Design and development were deeply paired—every feature prototyped, validated, and tested together. We ran async sprints, used GitHub for transparency, and deployed updates weekly. The feedback loop between model outputs and UI feedback was short and sharp—resulting in a product that thinks fast and feels effortless.
What started as a raw concept became a working product with a 72% win rate across all picks—and 80%+ in NFL and NHL. We didn’t just build an app. We built trust. And in a space filled with noise, that trust is the product. The Pick gives users a measurable edge, a daily ritual, and a sense that they’re betting with brains—not bias.
Since launch, user engagement has grown steadily. 30% of users open the app daily. Retention is high. And more importantly, we’re already planning the next wave—personalized recommendations, in-game support, new leagues, and bankroll strategy.
At Metamodern, we don’t just design interfaces. We build intelligent systems that move markets. If you’re ready to turn a good idea into a category-defining platform—let’s talk.
We built something Vegas would never dare to—and made it beautiful.
More projects.
(2010-2025)
