dugout
Confidential 360° feedback app for a youth baseball team, where coaches and players complete assessments on a shared rubric and the head coach releases AI-synthesized growth summaries directly to players.
23 commits over 2 days
What it is
Dugout enables youth baseball coaches to collect structured feedback from players (self-assessment) and coaching staff, then synthesize it into personalized, encouraging summaries for each player. The app enforces confidentiality by design: players only see their own released summary—never raw coach ratings, numeric scores, or peer data. Coaches sign in via SMS codes sent to their mobile numbers, and players access private, unguessable links (or sign in with an SMS code to their coach-provided number). The seven-area rubric (Hitting, Fielding, Throwing, Baserunning, Effort & Hustle, Coachability, Teammate) is rated 1–4; AI synthesis (Claude Opus 4) drafts the summaries, which the head coach edits and releases. An optional GameChanger import enriches feedback with season stats.
Why I built it
Built to solve a real problem for a youth baseball team: delivering meaningful, actionable growth feedback to young players while protecting confidentiality and avoiding the pitfalls of numeric rankings or peer comparison. The README explicitly calls it an MVP and outlines a roadmap for hardening before real use with children's data (RLS, per-team isolation, consent workflows), suggesting the developer is building toward production use with a specific team or league.
Tech stack
Next.js 15 (App Router), React 19, TypeScript (strict), Tailwind CSS v4, Supabase (Postgres), server actions for mutations, Anthropic SDK (Claude Opus 4) for AI synthesis, Textbelt for SMS (OTP authentication). Deployed to Vercel.
Timeline & effort
Last active Jun 10, 2026 — 23 commits across 2 active days (spanning 0 days).
Intensive focused sprint: 23 commits across just 2 days (June 9–10, 2026). The developer built the entire feature set—multi-role auth (per-coach SMS login, passwordless player links and SMS), rubric forms, AI synthesis with template fallback, GameChanger stats import, and Supabase schema migrations—in a single weekend push.
Did it turn out useful?
Deployed to Vercel with a functioning feature set and real AI synthesis, but the README explicitly labels this an MVP and lists hardening tasks (per-team isolation, RLS policies, consent workflows) required "before real use with children's data." The project is production-ready in architecture but deliberately scoped as a prototype pending compliance and multi-tenancy work.